NVIDIA Accelerated Linux Graphics Driver README and Installation Guide NVIDIA Corporation Last Updated: Fri Mar 14 13:21:26 UTC 2025 Most Recent Driver Version: 570.123.06 Published by NVIDIA Corporation 2701 San Tomas Expressway Santa Clara, CA 95050 NOTICE: ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, "MATERIALS") ARE BEING PROVIDED "AS IS." NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE. Information furnished is believed to be accurate and reliable. However, NVIDIA Corporation assumes no responsibility for the consequences of use of such information or for any infringement of patents or other rights of third parties that may result from its use. No license is granted by implication or otherwise under any patent or patent rights of NVIDIA Corporation. Specifications mentioned in this publication are subject to change without notice. This publication supersedes and replaces all information previously supplied. NVIDIA Corporation products are not authorized for use as critical components in life support devices or systems without express written approval of NVIDIA Corporation. NVIDIA, the NVIDIA logo, NVIDIA nForce, GeForce, NVIDIA Quadro, Vanta, TNT2, TNT, RIVA, RIVA TNT, Tegra, and TwinView are registered trademarks or trademarks of NVIDIA Corporation in the United States and/or other countries. Linux is a registered trademark of Linus Torvalds. Fedora and Red Hat are trademarks of Red Hat, Inc. SuSE is a registered trademark of SuSE AG. Mandriva is a registered trademark of Mandriva S.A. Intel is a registered trademark of Intel Corporation. OpenGL is a registered trademark of Silicon Graphics Inc. PCI Express is a registered trademark and/or service mark of PCI-SIG. Windows is a registered trademark of Microsoft Corporation in the United States and other countries. Other company and product names may be trademarks or registered trademarks of the respective owners with which they are associated. Copyright 2006 - 2025 NVIDIA Corporation. All rights reserved. ______________________________________________________________________________ TABLE OF CONTENTS ______________________________________________________________________________ Chapter 1. Introduction Chapter 2. Minimum Requirements Chapter 3. Selecting and Downloading the NVIDIA Packages for Your System Chapter 4. Installing the NVIDIA Driver Chapter 5. Listing of Installed Components Chapter 6. Configuring X for the NVIDIA Driver Chapter 7. Frequently Asked Questions Chapter 8. Common Problems Chapter 9. Known Issues Chapter 10. Allocating DMA Buffers on 64-bit Platforms Chapter 11. Specifying OpenGL Environment Variable Settings Chapter 12. Configuring Multiple Display Devices on One X Screen Chapter 13. Configuring GLX in Xinerama Chapter 14. Configuring Multiple X Screens on One Card Chapter 15. Support for the X Resize and Rotate Extension Chapter 16. Configuring a Notebook Chapter 17. Using the NVIDIA Driver with Optimus Laptops Chapter 18. Programming Modes Chapter 19. Configuring Flipping and UBB Chapter 20. Using the /proc File System Interface Chapter 21. Configuring Power Management Support Chapter 22. PCI-Express Runtime D3 (RTD3) Power Management Chapter 23. Dynamic Boost on Linux Chapter 24. Using the X Composite Extension Chapter 25. Using the nvidia-settings Utility Chapter 26. NVIDIA Spectre V2 Mitigation Chapter 27. Using the nvidia-smi Utility Chapter 28. The NVIDIA Management Library Chapter 29. Using the nvidia-debugdump Utility Chapter 30. Using the nvidia-persistenced Utility Chapter 31. Configuring SLI Mosaic Chapter 32. Configuring Frame Lock and Genlock Chapter 33. Configuring Depth 30 Displays Chapter 34. Offloading Graphics Display with RandR 1.4 Chapter 35. PRIME Render Offload Chapter 36. Direct Rendering Manager Kernel Modesetting (DRM KMS) Chapter 37. Configuring External and Removable GPUs Chapter 38. NGX Chapter 39. OpenGL, Vulkan and VDPAU on Xwayland Chapter 40. GBM and GBM-based Wayland Compositors Chapter 41. Addressing Capabilities Chapter 42. GPUDirect RDMA Peer Memory Client Chapter 43. GSP Firmware Chapter 44. Open Linux Kernel Modules Chapter 45. NVIDIA Contact Info and Additional Resources Chapter 46. Acknowledgements Appendix A. Supported NVIDIA GPU Products Appendix B. X Config Options Appendix C. Display Device Names Appendix D. GLX Support Appendix E. Dots Per Inch Appendix F. i2c Bus Support Appendix G. VDPAU Support Appendix H. Audio Support Appendix I. Tips for New Linux Users Appendix J. Application Profiles Appendix K. GPU Names Appendix L. Wayland Known Issues ______________________________________________________________________________ Chapter 1. Introduction ______________________________________________________________________________ 1A. ABOUT THE NVIDIA ACCELERATED LINUX GRAPHICS DRIVER The NVIDIA Accelerated Linux Graphics Driver brings accelerated 2D functionality and high-performance OpenGL support to Linux x86_64 with the use of NVIDIA graphics processing units (GPUs). These drivers provide optimized hardware acceleration for OpenGL and X applications and support nearly all recent NVIDIA GPU products (see Appendix A for a complete list of supported GPUs). 1B. ABOUT THIS DOCUMENT This document provides instructions for the installation and use of the NVIDIA Accelerated Linux Graphics Driver. Chapter 3, Chapter 4 and Chapter 6 walk the user through the process of downloading, installing and configuring the driver. Chapter 7 addresses frequently asked questions about the installation process, and Chapter 8 provides solutions to common problems. The remaining chapters include details on different features of the NVIDIA Linux Driver. Frequently asked questions about specific tasks are included in the relevant chapters. These pages are posted on NVIDIA's web site (http://www.nvidia.com), and are installed in '/usr/share/doc/NVIDIA_GLX-1.0/'. 1C. ABOUT THE AUDIENCE It is assumed that the user and reader of this document has at least a basic understanding of Linux techniques and terminology. However, new Linux users can refer to Appendix I for details on parts of the installation process. 1D. ADDITIONAL INFORMATION In case additional information is required, Chapter 45 provides contact information for NVIDIA Linux driver resources, as well as a brief listing of external resources. ______________________________________________________________________________ Chapter 2. Minimum Requirements ______________________________________________________________________________ 2A. MINIMUM SOFTWARE REQUIREMENTS Software Element Supported versions Check With... --------------------- --------------------- --------------------- Linux kernel 4.15 and newer `cat /proc/version` X.Org xserver 1.7, 1.8, 1.9, 1.10, `Xorg -version` 1.11, 1.12, 1.13, 1.14, 1.15, 1.16, 1.17, 1.18, 1.19, 1.20, 21.1 Kernel modutils 2.1.121 and newer `insmod --version` glibc 2.11 `/lib/libc.so.6` libvdpau * 0.2 `pkg-config --modversion vdpau` libvulkan ** 1.0.61 `pkg-config --modversion vulkan` * Required for hardware-accelerated video playback. See Appendix G for more information. ** Required for applications which use the Vulkan API. If you need to build the NVIDIA kernel module: Software Element Min Requirement Check With... --------------------- --------------------- --------------------- binutils 2.9.5 `size --version` GNU make 3.77 `make --version` gcc 2.91.66 `gcc --version` All official stable kernel releases from 4.15 and up are supported; pre-release versions, such as 4.19-rc1, are not supported. The Linux kernel can be downloaded from http://www.kernel.org or one of its mirrors. binutils and gcc can be retrieved from http://www.gnu.org or one of its mirrors. Sometimes very recent X server versions are not supported immediately following release, but we aim to support all new versions as soon as possible. Support is not added for new X server versions until after the video driver ABI is frozen, which usually happens at the release candidate stage. Prerelease versions that are not release candidates, such as "1.10.99.1", are not supported. If you are setting up the X Window System for the first time, it is often easier to begin with one of the open source drivers that ships with X.Org (either "vga", "vesa", or "fbdev"). Once your system is operating properly with the open source driver, you may then switch to the NVIDIA driver. These software packages may also be available through your Linux distributor. ______________________________________________________________________________ Chapter 3. Selecting and Downloading the NVIDIA Packages for Your System ______________________________________________________________________________ NVIDIA drivers can be downloaded from the NVIDIA website (http://www.nvidia.com). The NVIDIA graphics driver uses a Unified Driver Architecture: the single graphics driver supports all modern NVIDIA GPUs. "Legacy" GPU support has been moved from the unified driver to special legacy GPU driver releases. See Appendix A for a list of legacy GPUs. The NVIDIA graphics driver is bundled in a self-extracting package named 'NVIDIA-Linux-x86_64-570.123.06.run'. On Linux-x86_64, that file contains both the 64-bit driver binaries as well as 32-bit compatibility driver binaries; the 'NVIDIA-Linux-x86_64-570.123.06-no-compat32.run' file only contains the 64-bit driver binaries. ______________________________________________________________________________ Chapter 4. Installing the NVIDIA Driver ______________________________________________________________________________ This chapter provides instructions for installing the NVIDIA driver. Note that after installation, but prior to using the driver, you must complete the steps described in Chapter 6. Additional details that may be helpful for the new Linux user are provided in Appendix I. 4A. BEFORE YOU BEGIN Before you begin the installation, exit the X server and terminate all OpenGL applications (note that it is possible that some OpenGL applications persist even after the X server has stopped). You should also set the default run level on your system such that it will boot to a VGA console, and not directly to X. Doing so will make it easier to recover if there is a problem during the installation process. See Appendix I for details. If you're installing on a system that is set up to use the Nouveau driver, then you should first disable it before attempting to install the NVIDIA driver. See Interaction with the Nouveau Driver for details. 4B. STARTING THE INSTALLER After you have downloaded the file 'NVIDIA-Linux-x86_64-570.123.06.run', change to the directory containing the downloaded file, and as the 'root' user run the executable: # cd yourdirectory # sh NVIDIA-Linux-x86_64-570.123.06.run The '.run' file is a self-extracting archive. When executed, it extracts the contents of the archive and runs the contained 'nvidia-installer' utility, which provides an interactive interface to walk you through the installation. 'nvidia-installer' will also install itself to '/usr/bin/nvidia-installer', which may be used at some later time to uninstall drivers, auto-download updated drivers, etc. The use of this utility is detailed later in this chapter. You may also supply command line options to the '.run' file. Some of the more common options are listed below. Common '.run' Options --info Print embedded info about the '.run' file and exit. --check Check integrity of the archive and exit. --extract-only Extract the contents of './NVIDIA-Linux-x86_64-570.123.06.run', but do not run 'nvidia-installer'. --help Print usage information for the common commandline options and exit. --advanced-options Print usage information for common command line options as well as the advanced options, and then exit. 4C. INSTALLING THE KERNEL INTERFACE The NVIDIA kernel module has a kernel interface layer that must be compiled specifically for each kernel. NVIDIA distributes the source code to this kernel interface layer. When the installer is run, it will check your system for the required kernel sources and compile the kernel interface. You must have the source code for your kernel installed for compilation to work. On most systems, this means that you will need to locate and install the correct kernel-source, kernel-headers, or kernel-devel package; on some distributions, no additional packages are required. After the correct kernel interface has been compiled, the kernel interface will be linked with the closed-source portion of the NVIDIA kernel module. This requires that you have a linker installed on your system. The linker, usually '/usr/bin/ld', is part of the binutils package. You must have a linker installed prior to installing the NVIDIA driver. 4D. REGISTERING THE NVIDIA KERNEL MODULE WITH DKMS The installer will check for the presence of DKMS on your system. If DKMS is found, you will be given the option of registering the kernel module with DKMS. On most systems with DKMS, DKMS will take care of automatically rebuilding registered kernel modules when installing a different Linux kernel. This is typically done by hooks that Linux distributions set up to be triggered as part of the kernel installation process. If your system is not configured to automatically run DKMS to rebuild kernel modules after installing a different kernel, you may run `dkms autoinstall` to trigger the process manually. Note that while DKMS is useful for rebuilding the kernel modules after a minor kernel update, such as typical maintenance and security updates that Linux distributions may publish on a regular basis, a major kernel update may require installing a newer NVIDIA driver, as changes to the kernel APIs may not be compatible with the existing driver version. On systems which require signed kernel modules, newer versions of DKMS may assist you in generating and enrolling a Machine Owner Key (MOK), so that future automated DKMS builds can reuse the already trusted signing keys. This is more convenient than having to create a new signature every time the kernel modules are rebuilt; however, it is slightly less secure, as the keys will be stored at a well-known location in your filesystem. 4E. SIGNING THE NVIDIA KERNEL MODULE Some kernels may require that kernel modules be cryptographically signed by a key trusted by the kernel in order to be loaded. In particular, many distributions require modules to be signed when loaded into kernels running on UEFI systems with Secure Boot enabled. 'nvidia-installer' includes support for signing the kernel module before installation, to ensure that it can be loaded on such systems. Note that not all UEFI systems have Secure Boot enabled, and not all kernels running on UEFI Secure Boot systems will require signed kernel modules, so if you are uncertain about whether your system requires signed kernel modules, you may try installing the driver without signing the kernel module, to see if the unsigned kernel module can be loaded. In order to sign the kernel module, you will need a private signing key, and an X.509 certificate for the corresponding public key. The X.509 certificate must be trusted by the kernel before the module can be loaded: we recommend ensuring that the signing key be trusted before beginning the driver installation, so that the newly signed module can be used immediately. If you do not already have a key pair suitable for module signing use, you must generate one. Please consult your distribution's documentation for details on the types of keys suitable for module signing, and how to generate them. 'nvidia-installer' can generate a key pair for you at install time, but it is preferable to have a key pair already generated and trusted by the kernel before installation begins. Once you have a key pair ready, you can use that key pair in 'nvidia-installer' by passing the keys to the installer on the command line with the --module-signing-secret-key and --module-signing-public-key options. As an example, it is possible to install the driver with a signed kernel module in silent mode (i.e., non-interactively) by running: # sh ./NVIDIA-Linux-x86_64-570.123.06.run -s \ --module-signing-secret-key=/path/to/signing.key \ --module-signing-public-key=/path/to/signing.x509 In the example above, 'signing.key' and 'signing.x509' are a private/public key pair, and the public key is already enrolled in one of the kernel's trusted module signing key sources. On UEFI systems with secure boot enabled, nvidia-installer will present a series of interactive prompts to guide users through the module signing process. As an alternative to setting the key paths on the command line, the paths can be provided interactively in response to the prompts. These prompts will also appear when building the NVIDIA kernel module against a kernel which has CONFIG_MODULE_SIG_FORCE enabled in its configuration, or if the installer is run in expert mode. KEY SOURCES TRUSTED BY THE KERNEL In order to load a kernel module into a kernel that requires module signatures, the module must be signed by a key that the kernel trusts. There are several sources that a kernel may draw upon to build its pool of trusted keys. If you have generated a key pair, but it is not yet trusted by your kernel, you must add a certificate for your public key to a trusted key source before it can be used to verify signatures of signed kernel modules. These trusted sources include: Certificates embedded into the kernel image On kernels with CONFIG_MODULE_SIG set, a certificate for the public key used to sign the in-tree kernel modules is embedded, along with any additional module signing certificates provided at build time, into the kernel image. Modules signed by the private keys that correspond to the embedded public key certificates will be trusted by the kernel. Since the keys are embedded at build time, the only way to add a new public key is to build a new kernel. On UEFI systems with Secure Boot enabled, the kernel image will, in turn, need to be signed by a key that is trusted by the bootloader, so users building their own kernels with custom embedded keys should have a plan for making sure that the bootloader will load the new kernel. Certificates stored in the UEFI firmware database On kernels with CONFIG_MODULE_SIG_UEFI, in addition to any certificates embedded into the kernel image, the kernel can use certificates stored in the 'db', 'KEK', or 'PK' databases of the computer's UEFI firmware to verify the signatures of kernel modules, as long as they are not in the UEFI 'dbx' blacklist. Any user who holds the private key for the Secure Boot 'PK', or any of the keys in the 'KEK' list should be able to add new keys that can be used by a kernel with CONFIG_MODULE_SIG_UEFI, and any user with physical access to the computer should be able to delete any existing Secure Boot keys, and install his or her own keys instead. Please consult the documentation for your UEFI-based computer system for details on how to manage the UEFI Secure Boot keys. Certificates stored in a supplementary key database Some distributions include utilities that allow for the secure storage and management of cryptographic keys in a database that is separate from the kernel's built-in key list, and the key lists in the UEFI firmware. A prominent example is the MOK (Machine Owner Key) database used by some versions of the 'shim' bootloader, and the associated management utilities, 'mokutil' and 'MokManager'. Such a system allows users to enroll additional keys without the need to build a new kernel or manage the UEFI Secure Boot keys. Please consult your distribution's documentation for details on whether such a supplementary key database is available, and if so, how to manage its keys. GENERATING SIGNING KEYS IN NVIDIA-INSTALLER 'nvidia-installer' can generate keys that can be used for module signing, if existing keys are not readily available. Note that a module signed by a newly generated key cannot be loaded into a kernel that requires signed modules until its key is trusted, and when such a module is installed on such a system, the installed driver will not be immediately usable, even if the installation was successful. When 'nvidia-installer' generates a key pair and uses it to sign a kernel module, an X.509 certificate containing the public key will be installed to disk, so that it can be added to one of the kernel's trusted key sources. 'nvidia-installer' will report the location of the installed certificate: make a note of this location, and of the certificate's SHA1 fingerprint, so that you will be able to enroll the certificate and verify that it is correct, after the installation is finished. By default, 'nvidia-installer' will attempt to securely delete the generated private key with 'shred -u' after the module is signed. This is to prevent the key from being exploited to sign a malicious kernel module, but it also means that the same key can't be used again to install a different driver, or even to install the same driver on a different kernel. 'nvidia-installer' can optionally install the private signing key to disk, as it does with the public certificate, so that the key pair can be reused in the future. If you elect to install the private key, please make sure that appropriate precautions are taken to ensure that it cannot be stolen. Some examples of precautions you may wish to take: Prevent the key from being read by anybody without physical access to the computer In general, physical access is required to install Secure Boot keys, including keys managed outside of the standard UEFI key databases, to prevent attackers who have remotely compromised the security of the operating system from installing malicious boot code. If a trusted key is available to remote users, even root, then it will be possible for an attacker to sign arbitrary kernel modules without first having physical access, making the system less secure. One way to ensure that the key is not available to remote users is to keep it on a removable storage medium, which is disconnected from the computer except when signing modules. Do not store the unencrypted private key Encrypting the private key can add an extra layer of security: the key will not be useful for signing modules unless it can be successfully decrypted, first. Make sure not to store unencrypted copies of the key on persistent storage: either use volatile storage (e.g. a RAM disk), or securely delete any unencrypted copies of the key when not in use (e.g. using 'shred' instead of 'rm'). Note that using 'shred' may not be sufficient to fully purge data from some storage devices, in particular, some types of solid state storage. ALTERNATIVES TO THE INSTALLER'S MODULE SIGNING SUPPORT It is possible to load the NVIDIA kernel module on a system that requires signed modules, without using the installer's module signing support. Depending on your particular use case, you may find one of these alternatives more suitable than signing the module with 'nvidia-installer': Disable UEFI Secure Boot, if applicable On some kernels, a requirement for signed modules is only enforced when booted on a UEFI system with Secure Boot enabled. Some users of such kernels may find it more convenient to disable Secure Boot; however, note that this will reduce the security of your system by making it easier for malicious users to install potentially harmful boot code, kernels, or kernel modules. Use a kernel that doesn't require signed modules The kernel can be configured not to check module signatures, or to check module signatures, but allow modules without a trusted signature to be loaded, anyway. Installing a kernel configured in such a way will allow the installation of unsigned modules. Note that on Secure Boot systems, you will still need to ensure that the kernel be signed with a key trusted by the bootloader and/or boot firmware, and that a kernel that doesn't enforce module signature verification may be slightly less secure than one that does. 4F. ADDING PRECOMPILED KERNEL INTERFACES TO THE INSTALLER PACKAGE When 'nvidia-installer' runs, it searches for a pre-compiled kernel interface layer for the target kernel: if one is found, then the complete kernel module can be produced by linking the precompiled interface with 'nv-kernel.o', instead of needing to compile the kernel interface on the target system. 'nvidia-installer' includes a feature which allows users to add a precompiled interface to the installer package. This is useful in many use cases; for example, an administrator of a large group of similarly configured computers can prepare an installer package with a precompiled interface for the kernel running on those computers, then deploy the customized installer, which will be able to install the NVIDIA kernel module without needing to have the kernel development headers or a compiler installed on the target systems. (A linker is still required.) To use this feature, simply invoke the '.run' installer package with the --add-this-kernel option; e.g. # sh ./NVIDIA-Linux-x86_64-570.123.06.run --add-this-kernel This will unpack 'NVIDIA-Linux-x86_64-570.123.06.run', compile a kernel interface layer for the currently running kernel (use the --kernel-source-path and --kernel-output-path options to specify a target kernel other than the currently running one), and create a new installer package with the kernel interface layer added. Administrators of large groups of similarly configured computers that are configured to require trusted signatures in order to load kernel modules may find this feature especially useful when combined with the built-in support for module signing in 'nvidia-installer'. To package a .run file with a precompiled kernel interface layer, plus a detached module signature for the linked module, just use the --module-signing-secret-key and --module-signing-public-key options alongside the --add-this-kernel option. The resulting package, besides being installable without kernel headers or a compiler on the target system, has the added benefit of being able to produce a signed module without needing access to the private key on the install target system. Note that the detached signature will only be valid if the result of linking the precompiled interface with 'nv-kernel.o' on the target system is exactly the same as the result of linking those two files on the system that was used to create the custom installer. To ensure optimal compatibility, the linker used on both the package preparation system and the install target system should be the same. 4G. OTHER FEATURES OF THE INSTALLER Without options, the '.run' file executes the installer after unpacking it. The installer can be run as a separate step in the process, or can be run at a later time to get updates, etc. Some of the more important commandline options of 'nvidia-installer' are: 'nvidia-installer' options --uninstall During installation, the installer will make backups of any conflicting files and record the installation of new files. The uninstall option undoes an install, restoring the system to its pre-install state. --ui=none The installer uses an ncurses-based user interface if it is able to locate the correct ncurses library. Otherwise, it will fall back to a simple commandline user interface. This option disables the use of the ncurses library. ______________________________________________________________________________ Chapter 5. Listing of Installed Components ______________________________________________________________________________ The NVIDIA Accelerated Linux Graphics Driver consists of the following components (filenames in parentheses are the full names of the components after installation). Some paths may be different on different systems (e.g., X modules may be installed in /usr/X11R6/ rather than /usr/lib/xorg/). o An X driver ('/usr/lib/xorg/modules/drivers/nvidia_drv.so'); this driver is needed by the X server to use your NVIDIA hardware. o A GLX extension module for X ('/usr/lib/xorg/modules/extensions/libglxserver_nvidia.so.570.123.06'); this module is used by the X server to provide server-side GLX support. o EGL and OpenGL ES libraries ( '/usr/lib/libEGL.so.1', '/usr/lib/libGLESv1_CM.so.570.123.06', and '/usr/lib/libGLESv2.so.570.123.06' ); these libraries provide the API entry points for all OpenGL ES and EGL function calls. They are loaded at run-time by applications. o A Wayland EGL external platform library ('/usr/lib/libnvidia-egl-wayland.so.1') and its corresponding configuration file ( '/usr/share/egl/egl_external_platform.d/10_nvidia_wayland.json' ); this library provides client-side Wayland EGL application support using either dmabuf or EGLStream buffer sharing protocols, as well as server-side protocol extensions enabling the client-side EGLStream-based buffer sharing path. The EGLStream path can only be used in combination with an EGLStream-enabled Wayland compositor, e.g: https://gitlab.freedesktop.org/ekurzinger/weston. More information can be found along with the EGL external interface and Wayland library source code at https://github.com/NVIDIA/eglexternalplatform and https://github.com/NVIDIA/egl-wayland. o A GBM EGL external platform library ('/usr/lib/libnvidia-egl-gbm.so.1') and its corresponding configuration file ( '/usr/share/egl/egl_external_platform.d/15_nvidia_gbm.json' ); this library provides GBM EGL application support. The source code for the GBM EGL external platform library can be found at https://github.com/NVIDIA/egl-gbm. See Chapter 40 for more details on GBM support in the NVIDIA driver. o Vendor neutral graphics libraries provided by libglvnd ('/usr/lib/libOpenGL.so.0', '/usr/lib/libGLX.so.0', and '/usr/lib/libGLdispatch.so.0'); these libraries are currently used to provide full OpenGL dispatching support to NVIDIA's implementation of EGL. Source code for libglvnd is available at https://github.com/NVIDIA/libglvnd o GLVND vendor implementation libraries for GLX ('/usr/lib/libGLX_nvidia.so.0') and EGL ('/usr/lib/libEGL_nvidia.so.0'); these libraries provide NVIDIA implementations of OpenGL functionality which may be accessed using the GLVND client-facing libraries. 'libGLX_nvidia.so.0' and 'libEGL_nvidia.so.0' can be used as Vulkan ICDs. The Vulkan ICD configuration file included with the driver references 'libGLX_nvidia.so.0', however in environments where X11 client libraries are not available, 'libEGL_nvidia.so.0' should be used. The Vulkan ICD configuration file is installed as '/etc/vulkan/icd.d/nvidia_icd.json'. An additional Vulkan layer configuration file is installed as '/etc/vulkan/implicit_layer.d/nvidia_layers.json'. These layers add functionality to the Vulkan loader. o A GLVND GLX client ICD loader library ('/usr/lib/libGL.so.1'), This library provides API entry points for all GLX function calls, and is loaded at run-time by applications. This library is included as a convenience to support systems that do not already have an existing libglvnd installation. On systems with libglvnd libraries already installed, the existing libraries should be used instead. o Various libraries that are used internally by other driver components. These include '/usr/lib/libnvidia-cfg.so.570.123.06', '/usr/lib/libnvidia-eglcore.so.570.123.06', '/usr/lib/libnvidia-glcore.so.570.123.06', '/usr/lib/libnvidia-glsi.so.570.123.06', '/usr/lib/libnvidia-glvkspirv.so.570.123.06', '/usr/lib/libnvidia-gpucomp.so.570.123.06', '/usr/lib/libnvidia-rtcore.so.570.123.06', and '/usr/lib/libnvidia-allocator.so.570.123.06'. o A VDPAU (Video Decode and Presentation API for Unix-like systems) library for the NVIDIA vendor implementation, ('/usr/lib/vdpau/libvdpau_nvidia.so.570.123.06'); see Appendix G for details. o The CUDA library ('/usr/lib/libcuda.so.570.123.06') which provides runtime support for CUDA (high-performance computing on the GPU) applications. o The CUDA Debugger library ('/usr/lib/libcudadebugger.so.570.123.06') which provides support for debugging CUDA applications. o Cryptography library wrappers, ('/usr/lib/libnvidia-pkcs11.so.570.123.06') and ('/usr/lib/libnvidia-pkcs11.openssl3.so.570.123.06'), which dynamically link to the system's OpenSSL libraries and provide cryptography operations when the GPU and driver are operating in Confidential Compute mode. The driver attempts to use OpenSSL 3 first and if OpenSSL 3 is not available then it will attempt to use OpenSSL 1.1. o The PTX JIT Compiler library ('/usr/lib/libnvidia-ptxjitcompiler.so.570.123.06') is a JIT compiler which compiles PTX into GPU machine code and is used by the CUDA driver. o The NVVM Compiler library ('/usr/lib/libnvidia-nvvm.so.4.0.0') is loaded by the CUDA driver to do JIT link-time-optimization. o Two OpenCL libraries ('/usr/lib/libOpenCL.so.1.0.0', '/usr/lib/libnvidia-opencl.so.570.123.06'); the former is a vendor-independent Installable Client Driver (ICD) loader, and the latter is the NVIDIA Vendor ICD. A config file '/etc/OpenCL/vendors/nvidia.icd' is also installed, to advertise the NVIDIA Vendor ICD to the ICD Loader. o The 'nvidia-cuda-mps-control' and 'nvidia-cuda-mps-server' applications, which allow MPI processes to run concurrently on a single GPU. o A kernel module ('/lib/modules/`uname -r`/kernel/drivers/video/nvidia-modeset.ko'); this kernel module is responsible for programming the display engine of the GPU. User-mode NVIDIA driver components such as the NVIDIA X driver, OpenGL driver, and VDPAU driver communicate with nvidia-modeset.ko through the /dev/nvidia-modeset device file. o A kernel module ('/lib/modules/`uname -r`/kernel/drivers/video/nvidia.ko'); this kernel module provides low-level access to your NVIDIA hardware for all of the above components. It is generally loaded into the kernel when the X server is started, and is used by the X driver and OpenGL. nvidia.ko consists of two pieces: the binary-only core, and a kernel interface that must be compiled specifically for your kernel version. Note that the Linux kernel does not have a consistent binary interface like the X server, so it is important that this kernel interface be matched with the version of the kernel that you are using. This can either be accomplished by compiling yourself, or using precompiled binaries provided for the kernels shipped with some of the more common Linux distributions. o NVIDIA Unified Memory kernel module ('/lib/modules/`uname -r`/kernel/drivers/video/nvidia-uvm.ko'); this kernel module provides functionality for sharing memory between the CPU and GPU in CUDA programs. It is generally loaded into the kernel when a CUDA program is started, and is used by the CUDA driver on supported platforms. o The nvidia-tls library ('/usr/lib/libnvidia-tls.so.570.123.06'); this file provides thread local storage support for the NVIDIA OpenGL libraries (libGLX_nvidia, libnvidia-glcore, and libglxserver_nvidia). o The nvidia-ml library ('/usr/lib/libnvidia-ml.so.570.123.06'); The NVIDIA Management Library provides a monitoring and management API. See Chapter 28 for more information. o The application nvidia-installer ('/usr/bin/nvidia-installer') is NVIDIA's tool for installing and updating NVIDIA drivers. See Chapter 4 for a more thorough description. Source code is available at https://download.nvidia.com/XFree86/nvidia-installer/. o The application nvidia-modprobe ('/usr/bin/nvidia-modprobe') is installed as setuid root and is used to load the NVIDIA kernel module, create the '/dev/nvidia*' device nodes and configure certain runtime settings in the kernel by processes (such as CUDA applications) that don't run with sufficient privileges to do those things themselves. Source code is available at https://download.nvidia.com/XFree86/nvidia-modprobe/. o The application nvidia-xconfig ('/usr/bin/nvidia-xconfig') is NVIDIA's tool for manipulating X server configuration files. See Chapter 6 for more information. Source code is available at https://download.nvidia.com/XFree86/nvidia-xconfig/. o The application nvidia-settings ('/usr/bin/nvidia-settings') is NVIDIA's tool for dynamic configuration while the X server is running. See Chapter 25 for more information. o The libnvidia-gtk libraries ('/usr/lib/libnvidia-gtk2.so.570.123.06' and '/usr/lib/libnvidia-gtk3.so.570.123.06'); these libraries are required to provide the nvidia-settings user interface. Source code is available at https://download.nvidia.com/XFree86/nvidia-settings/. o The application nvidia-smi ('/usr/bin/nvidia-smi') is the NVIDIA System Management Interface for management and monitoring functionality. See Chapter 27 for more information. o The application nvidia-debugdump ('/usr/bin/nvidia-debugdump') is NVIDIA's tool for collecting internal GPU state. It is normally invoked by the nvidia-bug-report.sh ('/usr/bin/nvidia-bug-report.sh') script. See Chapter 29 for more information. o The daemon nvidia-persistenced ('/usr/bin/nvidia-persistenced') is the NVIDIA Persistence Daemon for allowing the NVIDIA kernel module to maintain persistent state when no other NVIDIA driver components are running. See Chapter 30 for more information. Source code is available at https://download.nvidia.com/XFree86/nvidia-persistenced/. o The NVCUVID library ('/usr/lib/libnvcuvid.so.570.123.06'); The NVIDIA CUDA Video Decoder (NVCUVID) library provides an interface to hardware video decoding capabilities on NVIDIA GPUs with CUDA. o The NvEncodeAPI library ('/usr/lib/libnvidia-encode.so.570.123.06'); The NVENC Video Encoding library provides an interface to video encoder hardware on supported NVIDIA GPUs. o The NVAPI library ('/usr/lib/libnvidia-api.so.1;'); NVAPI provides an interface for managing properties of GPUs. The NVAPI library provides runtime support for NVAPI applications. o The NvFBC library ('/usr/lib/libnvidia-fbc.so.570.123.06'); The NVIDIA Framebuffer Capture library provides an interface to capture and optionally encode the framebuffer of an X server screen. o An X driver configuration file ('/usr/share/X11/xorg.conf.d/nvidia-drm-outputclass.conf'); If the X server is sufficiently new, this file will be installed to configure the X server to load the 'nvidia_drv.so' driver automatically if it is started after the NVIDIA DRM kernel module ('nvidia-drm.ko') is loaded. This feature is supported in X.Org xserver 1.16 and higher when running on Linux kernel 3.13 or higher with CONFIG_DRM enabled. o Predefined application profile keys and documentation for those keys can be found in the following files in the directory '/usr/share/nvidia/': 'nvidia-application-profiles-570.123.06-rc', 'nvidia-application-profiles-570.123.06-key-documentation'. See Appendix J for more information. o The OptiX library ('/usr/lib/libnvoptix.so.1'); This library implements the OptiX ray tracing engine. It is loaded by the 'liboptix.so.*' library bundled with applications that use the OptiX API. OptiX also includes a data file '/usr/share/nvidia/nvoptix.bin'. o The NVIDIA Optical Flow library ('/usr/lib/libnvidia-opticalflow.so.570.123.06'); The NVIDIA Optical Flow library can be used for hardware-accelerated computation of optical flow vectors and stereo disparity values on Turing and later NVIDIA GPUs. This is useful for some forms of computer vision and image analysis. The Optical Flow library depends on the NVCUVID library, which in turn depends on the CUDA library. o NGX ('/usr/lib/libnvidia-ngx.so.570.123.06'); and the NVIDIA NGX Updater ('/usr/bin/nvidia-ngx-updater'); NGX is a collection of software which provides AI features to applications. On Linux this is supported only with x86_64 applications. Developers interested in using NGX in their applications should visit the following resources: o https://developer.nvidia.com/rtx/ngx o https://developer.nvidia.com/dlss NGX comes with nvidia-ngx-updater, a binary which updates NGX features without requiring a full application update. By default this functionality is disabled on Linux and needs to be enabled via a configuration file; see Chapter 38 for details. o A kernel module ('/lib/modules/`uname -r`/kernel/drivers/video/nvidia-peermem.ko'); this kernel module allows Mellanox HCAs access to NVIDIA GPU memory read/write buffers without needing to copy data to host memory. See Chapter 42 for more information. o NGX for Proton and Wine ('/usr/lib/nvidia/wine/nvngx.dll'); ('/usr/lib/nvidia/wine/_nvngx.dll'); NGX for Proton and Wine is a Microsoft Windows dynamic-link library used by Microsoft Windows applications which support NVIDIA DLSS. In addition to driver-side support, changes to third-party software are required in order to support DLSS within Proton and Wine. o Firmware ('/lib/firmware/nvidia/570.123.06/gsp_*.bin') which offloads tasks from the CPU to the GPU. See Chapter 43 for more information. o VulkanSC ICD ('/usr/lib/libnvidia-vksc-core.so.1'), which provides NVIDIA's implementation of VulkanSC. More information about VulkanSC can be found at https://www.khronos.org/vulkansc/ . Pipeline Cache Compiler ('/usr/bin/nvidia-pcc'), which enables offline shader compilation for VulkanSC. The VulkanSC ICD configuration file is installed as '/etc/vulkansc/icd.d/nvidia_icd.json'. VulkanSC support is only provided in x86_64 driver packages. Note that the VulkanSC ICD distributed in this package is intended for development purposes only, and is NOT safety certified. o A JSON file ('/usr/share/nvidia/files.d/sandboxutils-filelist.json'), which lists all the driver files (library, firmware, binary, configuration, ...) used by container runtime environments such as nvidia-container-toolkit and enroot. Problems will arise if applications use the wrong version of a library. This can be the case if there are either old libGL libraries or stale symlinks left lying around. If you think there may be something awry in your installation, check that the following files are in place (these are all the files of the NVIDIA Accelerated Linux Graphics Driver, as well as their symlinks): /usr/lib/xorg/modules/drivers/nvidia_drv.so /usr/lib/xorg/modules/extensions/libglxserver_nvidia.so.570.123.06 /usr/lib/xorg/modules/extensions/libglxserver_nvidia.so -> libglxserver_nvidia.so.570.123.06 /usr/lib/libGL.so.1.0.0; /usr/lib/libGL.so.1 -> libGL.so.1.0.0; /usr/lib/libGL.so -> libGL.so.1 (libGL.so.1.0.0 is the name of the libglvnd client-side GLX ICD loader library included with the NVIDIA Linux driver. A compatible ICD loader provided by your distribution may have a slightly different filename.) /usr/lib/libnvidia-glcore.so.570.123.06 /usr/lib/libcuda.so.570.123.06 /usr/lib/libcuda.so -> libcuda.so.570.123.06 /lib/modules/`uname -r`/video/nvidia.{o,ko}, or /lib/modules/`uname -r`/kernel/drivers/video/nvidia.{o,ko} If there are other libraries whose "soname" conflicts with that of the NVIDIA libraries, ldconfig may create the wrong symlinks. It is recommended that you manually remove or rename conflicting libraries (be sure to rename clashing libraries to something that ldconfig will not look at -- we have found that prepending "XXX" to a library name generally does the trick), rerun 'ldconfig', and check that the correct symlinks were made. An example of a library that often creates conflicts is "/usr/lib/mesa/libGL.so*". If the libraries appear to be correct, then verify that the application is using the correct libraries. For example, to check that the application /usr/bin/glxgears is using the libglvnd GLX libraries, run: % ldd /usr/bin/glxgears linux-vdso.so.1 (0x00007ffc8a5d4000) libGL.so.1 => /usr/lib/x86_64-linux-gnu/libGL.so.1 (0x00007f6593896000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f65934f8000) libX11.so.6 => /usr/lib/x86_64-linux-gnu/libX11.so.6 (0x00007f65931c0000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f6592dcf000) libGLX.so.0 => /usr/lib/x86_64-linux-gnu/libGLX.so.0 (0x00007f6592b9e000) libGLdispatch.so.0 => /usr/lib/x86_64-linux-gnu/libGLdispatch.so.0 (0x00007f65928e8000) libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f65926c9000) /lib64/ld-linux-x86-64.so.2 (0x00007f6593d28000) libxcb.so.1 => /usr/lib/x86_64-linux-gnu/libxcb.so.1 (0x00007f65924a1000) libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f659229d000) libXau.so.6 => /usr/lib/x86_64-linux-gnu/libXau.so.6 (0x00007f6592099000) libXdmcp.so.6 => /usr/lib/x86_64-linux-gnu/libXdmcp.so.6 (0x00007f6591e93000) libbsd.so.0 => /lib/x86_64-linux-gnu/libbsd.so.0 (0x00007f6591c7e000) librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007f6591a76000) In the example above, the list of libraries reported by 'ldd' includes 'libGLX.so.0' and 'libGLdispatch.so.0'. If the GLX client library is something other than the GLVND 'libGL.so.1', then you will need to either remove the library that is getting in the way or adjust your dynamic loader search path using the 'LD_LIBRARY_PATH' environment variable. You may want to consult the man pages for 'ldconfig' and 'ldd'. ______________________________________________________________________________ Chapter 6. Configuring X for the NVIDIA Driver ______________________________________________________________________________ The X configuration file provides a means to configure the X server. This section describes the settings necessary to enable the NVIDIA driver. A comprehensive list of parameters is provided in Appendix B. The NVIDIA Driver includes a utility called nvidia-xconfig, which is designed to make editing the X configuration file easy. You can also edit it by hand. 6A. USING NVIDIA-XCONFIG TO CONFIGURE THE X SERVER nvidia-xconfig will find the X configuration file and modify it to use the NVIDIA X driver. In most cases, you can simply answer "Yes" when the installer asks if it should run it. If you need to reconfigure your X server later, you can run nvidia-xconfig again from a terminal. nvidia-xconfig will make a backup copy of your configuration file before modifying it. Note that the X server must be restarted for any changes to its configuration file to take effect. More information about nvidia-xconfig can be found in the nvidia-xconfig manual page by running. % man nvidia-xconfig 6B. MANUALLY EDITING THE CONFIGURATION FILE If you do not have a working X config file, there are a few different ways to obtain one. A sample config file is included both with the X.Org distribution and with the NVIDIA driver package (at '/usr/share/doc/NVIDIA_GLX-1.0/'). The 'nvidia-xconfig' utility, provided with the NVIDIA driver package, can generate a new X configuration file. Additional information on the X config syntax can be found in the xorg.conf manual page (`man xorg.conf`). If you have a working X config file for a different driver (such as the "vesa" or "fbdev" driver), then simply edit the file as follows. Remove the line: Driver "vesa" (or Driver "fbdev") and replace it with the line: Driver "nvidia" Remove the following lines: Load "dri" Load "GLCore" In the "Module" section of the file, add the line (if it does not already exist): Load "glx" If the X config file does not have a "Module" section, you can safely skip the last step. There are numerous options that may be added to the X config file to tune the NVIDIA X driver. See Appendix B for a complete list of these options. Once you have completed these edits to the X config file, you may restart X and begin using the accelerated OpenGL libraries. After restarting X, any OpenGL application should automatically use the new NVIDIA libraries. (NOTE: If you encounter any problems, see Chapter 8 for common problem diagnoses.) 6C. RESTORING THE X CONFIGURATION AFTER UNINSTALLING THE DRIVER If X is explicitly configured to use the NVIDIA driver, then the X config file should be edited to use a different X driver after uninstalling the NVIDIA driver. Otherwise, X may fail to start, since the driver it was configured to use will no longer be present on the system after uninstallation. If you edited the file manually, revert any edits you made. If you used the 'nvidia-xconfig' utility, either by answering "Yes" when prompted to configure the X server by the installer, or by running it manually later on, then you may restore the backed-up X config file, if it exists and reflects the X config state that existed before the NVIDIA driver was installed. If you do not recall any manual changes that you made to the file, or do not have a backed-up X config file that uses a non-NVIDIA X driver, you may want to try simply renaming the X configuration file, to see if your X server loads a sensible default. ______________________________________________________________________________ Chapter 7. Frequently Asked Questions ______________________________________________________________________________ This section provides answers to frequently asked questions associated with the NVIDIA Linux x86_64 Driver and its installation. Common problem diagnoses can be found in Chapter 8 and tips for new users can be found in Appendix I. Also, detailed information for specific setups is provided in the Appendices. NVIDIA-INSTALLER Q. How do I extract the contents of the '.run' without actually installing the driver? A. Run the installer as follows: # sh NVIDIA-Linux-x86_64-570.123.06.run --extract-only This will create the directory NVIDIA-Linux-x86_64-570.123.06, containing the uncompressed contents of the '.run' file. Q. How can I see the source code to the kernel interface layer? A. The source files to the kernel interface layer are in the kernel directory of the extracted .run file. To get to these sources, run: # sh NVIDIA-Linux-x86_64-570.123.06.run --extract-only # cd NVIDIA-Linux-x86_64-570.123.06/kernel/ Q. How and when are the NVIDIA device files created? A. When a user-space NVIDIA driver component needs to communicate with the NVIDIA kernel module, and the NVIDIA character device files do not yet exist, the user-space component will first attempt to load the kernel module and create the device files itself. Device file creation and kernel module loading generally require root privileges. The X driver, running within a setuid root X server, will have these privileges, but not, e.g., the CUDA driver within the environment of a normal user. If the user-space NVIDIA driver component cannot load the kernel module or create the device files itself, it will attempt to invoke the setuid root nvidia-modprobe utility, which will perform these operations on behalf of the non-privileged driver. See the nvidia-modprobe(1) man page, or its source code, available here: https://download.nvidia.com/XFree86/nvidia-modprobe/ When possible, it is recommended to use your Linux distribution's native mechanisms for managing kernel module loading and device file creation. nvidia-modprobe is provided as a fallback to work out-of-the-box in a distribution-independent way. Whether a user-space NVIDIA driver component does so itself, or invokes nvidia-modprobe, it will default to creating the device files with the following attributes: UID: 0 - 'root' GID: 0 - 'root' Mode: 0666 - 'rw-rw-rw-' Existing device files are changed if their attributes don't match these defaults. If you want the NVIDIA driver to create the device files with different attributes, you can specify them with the "NVreg_DeviceFileUID" (user), "NVreg_DeviceFileGID" (group) and "NVreg_DeviceFileMode" NVIDIA Linux kernel module parameters. For example, the NVIDIA driver can be instructed to create device files with UID=0 (root), GID=44 (video) and Mode=0660 by passing the following module parameters to the NVIDIA Linux kernel module: NVreg_DeviceFileUID=0 NVreg_DeviceFileGID=44 NVreg_DeviceFileMode=0660 The "NVreg_ModifyDeviceFiles" NVIDIA kernel module parameter will disable dynamic device file management, if set to 0. Q. Why does NVIDIA not provide RPMs? A. Not every Linux distribution uses RPM, and NVIDIA provides a single solution that works across all Linux distributions. NVIDIA encourages Linux distributions to repackage and redistribute the NVIDIA Linux driver in their native package management formats. These repackaged NVIDIA drivers are likely to inter-operate best with the Linux distribution's package management technology. For this reason, NVIDIA encourages users to use their distribution's repackaged NVIDIA driver, where available. Q. What is the significance of the '-no-compat32' suffix on Linux-x86_64 '.run' files? A. To distinguish between Linux-x86_64 driver package files that do or do not also contain 32-bit compatibility libraries, "-no-compat32" is be appended to the latter. 'NVIDIA-Linux-x86_64-570.123.06.run' contains both 64-bit and 32-bit driver binaries; but 'NVIDIA-Linux-x86_64-570.123.06-no-compat32.run' omits the 32-bit compatibility libraries. Q. Can I add my own precompiled kernel interfaces to a '.run' file? A. Yes, the --add-this-kernel '.run' file option will unpack the '.run' file, build a precompiled kernel interface for the currently running kernel, and repackage the '.run' file, appending '-custom' to the filename. This may be useful, for example. if you administer multiple Linux computers, each running the same kernel. Q. Where can I find the source code for the 'nvidia-installer' utility? A. The 'nvidia-installer' utility is released under the GPL. The source code for the version of nvidia-installer built with driver 570.123.06 is in 'nvidia-installer-570.123.06.tar.bz2' available here: https://download.nvidia.com/XFree86/nvidia-installer/ NVIDIA DRIVER Q. Where should I start when diagnosing display problems? A. One of the most useful tools for diagnosing problems is the X log file in '/var/log'. Lines that begin with "(II)" are information, "(WW)" are warnings, and "(EE)" are errors. You should make sure that the correct config file (i.e. the config file you are editing) is being used; look for the line that begins with: (==) Using config file: Also make sure that the NVIDIA driver is being used, rather than another driver. Search for (II) LoadModule: "nvidia" Lines from the driver should begin with: (II) NVIDIA(0) Q. How can I increase the amount of data printed in the X log file? A. By default, the NVIDIA X driver prints relatively few messages to stderr and the X log file. If you need to troubleshoot, then it may be helpful to enable more verbose output by using the X command line options -verbose and -logverbose, which can be used to set the verbosity level for the 'stderr' and log file messages, respectively. The NVIDIA X driver will output more messages when the verbosity level is at or above 5 (X defaults to verbosity level 1 for 'stderr' and level 3 for the log file). So, to enable verbose messaging from the NVIDIA X driver to both the log file and 'stderr', you could start X with the verbosity level set to 5, by doing the following % startx -- -verbose 5 -logverbose 5 Q. What is NVIDIA's policy towards development series Linux kernels? A. NVIDIA does not officially support development series kernels. However, all the kernel module source code that interfaces with the Linux kernel is available in the 'kernel/' directory of the '.run' file. NVIDIA encourages members of the Linux community to develop patches to these source files to support development series kernels. A web search will most likely yield several community supported patches. Q. Where can I find the tarballs? A. Plain tarballs are not available. The '.run' file is a tarball with a shell script prepended. You can execute the '.run' file with the --extract-only option to unpack the tarball. Q. How do I tell if I have my kernel sources installed? A. If you are running on a distro that uses RPM (Red Hat, Mandriva, SuSE, etc), then you can use 'rpm' to tell you. At a shell prompt, type: % rpm -qa | grep kernel and look at the output. You should see a package that corresponds to your kernel (often named something like kernel-2.6.15-7) and a kernel source package with the same version (often named something like kernel-devel-2.6.15-7). If none of the lines seem to correspond to a source package, then you will probably need to install it. If the versions listed mismatch (e.g., kernel-2.6.15-7 vs. kernel-devel-2.6.15-10), then you will need to update the kernel-devel package to match the installed kernel. If you have multiple kernels installed, you need to install the kernel-devel package that corresponds to your RUNNING kernel (or make sure your installed source package matches the running kernel). You can do this by looking at the output of 'uname -r' and matching versions. Q. What is SELinux and how does it interact with the NVIDIA driver ? A. Security-Enhanced Linux (SELinux) is a set of modifications applied to the Linux kernel and utilities that implement a security policy architecture. When in use it requires that the security type on all shared libraries be set to 'shlib_t'. The installer detects when to set the security type, and sets it on all shared libraries it installs. The option --force-selinux passed to the '.run' file overrides the detection of when to set the security type. Q. Why does X use so much memory? A. When measuring any application's memory usage, you must be careful to distinguish between physical system RAM used and virtual mappings of shared resources. For example, most shared libraries exist only once in physical memory but are mapped into multiple processes. This memory should only be counted once when computing total memory usage. In the same way, the video memory on a graphics card or register memory on any device can be mapped into multiple processes. These mappings do not consume normal system RAM. This has been a frequently discussed topic on XFree86 mailing lists; see, for example: http://marc.theaimsgroup.com/?l=xfree-xpert&m=96835767116567&w=2 Additionally, the nvidia X driver reserves virtual memory ranges that will never be backed by physical memory to use for addressing video memory that isn't directly mapped to the CPU. The amount of virtual memory reserved for this purpose will vary depending on GPU capabilities, and is printed to the X log: [ 1859.106] (II) NVIDIA: Reserving 24576.00 MB of virtual memory for [ 1859.106] (II) NVIDIA: indirect memory access. The 'pmap' utility described in the above thread is available in the "procps" package shipped with most recent Linux distributions, and is a useful tool in distinguishing between types of memory mappings. For example, while 'top' may indicate that X is using several GB of memory, the last line of output from the output of pmap (note that pmap may need to be run as root): # pmap -d `pidof Xorg` | tail -n 1 mapped: 25435416K writeable/private: 20908K shared: 10768K reveals that X is really only using roughly 21MB of system RAM (the "writeable/private" value). Note, also, that X must allocate resources on behalf of X clients (the window manager, your web browser, etc); the X server's memory usage will increase as more clients request resources such as pixmaps, and decrease as you close X applications. Q. Why do applications that use DGA graphics fail? A. The NVIDIA driver does not support the graphics component of the XFree86-DGA (Direct Graphics Access) extension. Applications can use the XDGASelectInput() function to acquire relative pointer motion, but graphics-related functions such as XDGASetMode() and XDGAOpenFramebuffer() will fail. The graphics component of XFree86-DGA is not supported because it requires a CPU mapping of framebuffer memory. As graphics cards ship with increasing quantities of video memory, the NVIDIA X driver has had to switch to a more dynamic memory mapping scheme that is incompatible with DGA. Furthermore, DGA does not cooperate with other graphics rendering libraries such as Xlib and OpenGL because it accesses GPU resources directly. NVIDIA recommends that applications use OpenGL or Xlib, rather than DGA, for graphics rendering. Using rendering libraries other than DGA will yield better performance and improve interoperability with other X applications. Q. My kernel log contains messages that are prefixed with "Xid"; what do these messages mean? A. "Xid" messages indicate that a general GPU error occurred, most often due to the driver misprogramming the GPU or to corruption of the commands sent to the GPU. These messages provide diagnostic information that can be used by NVIDIA to aid in debugging reported problems. Some information on how to interpret Xid messages is available here: http://docs.nvidia.com/deploy/xid-errors/index.html Q. My kernel log contains the message "NVRM: Xid (...): 81, VGA Subsystem Error." How can I fix this? A. In some extreme cases, the VGA console can hang if messages are printed to a legacy VGA text console concurrently with applications that generate high GPU memory traffic. The solution to this problem is to not use a legacy VGA text console. Instead, on capable systems, use pure UEFI mode (not Compatibility Support Module (CSM)). On legacy SBIOS systems, use a framebuffer console such as vesafb. Q. I use the Coolbits overclocking interface to adjust my graphics card's clock frequencies, but the defaults are reset whenever X is restarted. How do I make my changes persistent? A. Clock frequency settings are not saved/restored automatically by default to avoid potential stability and other problems that may be encountered if the chosen frequency settings differ from the defaults qualified by the manufacturer. You can add an 'nvidia-settings' command to '~/.xinitrc' to automatically apply custom clock frequency settings when the X server is started. See the 'nvidia-settings(1)' manual page for more information on setting clock frequency settings on the command line. Q. Why is the refresh rate not reported correctly by utilities that use the XF86VidMode X extension and/or RandR X extension versions prior to 1.2 (e.g., `xrandr --q1`)? A. These extensions are not aware of multiple display devices on a single X screen; they only see the MetaMode bounding box, which may contain one or more actual modes. This means that if multiple MetaModes have the same bounding box, these extensions will not be able to distinguish between them. In order to support dynamic display configuration, the NVIDIA X driver must make each MetaMode appear to be unique and accomplishes this by using the refresh rate as a unique identifier. You can use `nvidia-settings -q RefreshRate` to query the actual refresh rate on each display device. Q. Why does starting certain applications result in Xlib error messages indicating extensions like "XFree86-VidModeExtension" or "SHAPE" are missing? A. If your X config file has a "Module" section that does not list the "extmod" module, some X server extensions may be missing, resulting in error messages of the form: Xlib: extension "SHAPE" missing on display ":0.0" Xlib: extension "XFree86-VidModeExtension" missing on display ":0.0" Xlib: extension "XFree86-DGA" missing on display ":0.0" You can solve this problem by adding the line below to your X config file's "Module" section: Load "extmod" Q. Where can I find older driver versions? A. Please visit https://download.nvidia.com/XFree86/Linux-x86_64/ Q. What is the format of a PCI Bus ID? A. Different tools have different formats for the PCI Bus ID of a PCI device. The X server's "BusID" X configuration file option interprets the BusID string in the format "bus@domain:device:function" (the "@domain" portion is only needed if the PCI domain is non-zero), in decimal. More specifically, "%d@%d:%d:%d", bus, domain, device, function in printf(3) syntax. NVIDIA X driver logging, nvidia-xconfig, and nvidia-settings match the X configuration file BusID convention. The lspci(8) utility, in contrast, reports the PCI BusID of a PCI device in the format "domain:bus:device.function", printing the values in hexadecimal. More specifically, "%04x:%02x:%02x.%x", domain, bus, device, function in printf(3) syntax. The "Bus Location" reported in the information file matches the lspci format. Also, the name of per-GPU directory in /proc/driver/nvidia/gpus is the same as the corresponding GPU's PCI BusID in lspci format. On systems where both an integrated GPU and a PCI slot are present, setting the "BusID" option to "AXI" selects the integrated GPU. By default, not specifying this option or setting it to an empty string selects a discrete GPU if available, the integrated GPU otherwise. Q. Why doesn't the NVIDIA X driver make more display resolutions and refresh rates available via RandR? A. Prior to the 302.* driver series, the list of modes reported to applications by the NVIDIA X driver was not limited to the list of modes natively supported by a display device. In order to expose the largest possible set of modes on digital flat panel displays, which typically do not accept arbitrary mode timings, the driver maintained separate sets of "front-end" and "back-end" mode timings, and scaled between them to simulate the availability of more modes than would otherwise be supported. Front-end timings were the values reported to applications, and back-end timings were what was actually sent to the display. Both sets of timings went through the full mode validation process, with the back-end timings having the additional constraint that they must be provided by the display's EDID, as only EDID-provided modes can be safely assumed to be supported by the display hardware. Applications could request any available front-end timings, which the driver would implicitly scale to either the "best fit" or "native" mode timings. For example, an application might request an 800x600 @ 60 Hz mode and the driver would provide it, but the real mode sent to the display would be 1920x1080 @ 30 Hz. While the availability of modes beyond those natively supported by a display was convenient for some uses, it created several problems. For example: o The complete front-end timings were reported to applications, but only the width and height were actually used. This could cause confusion because in many cases, changing the front-end timings did not change the back-end timings. This was especially confusing when trying to change the refresh rate, because the refresh rate in the front-end timings was ignored, but was still reported to applications. o The front-end timings reported to the user could be different from the backend timings reported in the display device's on screen display, leading to user confusion. Finding out the back-end timings (e.g. to find the real refresh rate) required using the NVIDIA-specific NV-CONTROL X extension. o The process by which back-end timings were selected for use with any given front-end timings was not transparent to users, and this process could only be explicitly configured with NVIDIA-specific xorg.conf options or the NV-CONTROL X extension. Confusion over how changing front-end timings could affect the back-end timings was especially problematic in use cases that were sensitive to the timings the display device receives, such as NVIDIA 3D Vision. o User-specified modes underwent normal mode validation, even though the timings in those modes were not used. For example, a 1920x1080 @ 100 Hz mode might fail the VertRefresh check, even though the back-end timings might actually be 1920x1080 @ 30 Hz. Version 1.2 of the X Resize and Rotate extension (henceforth referred to as "RandR 1.2") allows configuration of display scaling in a much more flexible and standardized way. The protocol allows applications to choose exactly which (back-end) mode timing is used, and exactly how the screen is scaled to fill that mode. It also allows explicit control over which displays are enabled, and which portions of the screen they display. This also provides much-needed transparency: the mode timings reported by RandR 1.2 are the actual mode timings being sent to the display. However, this means that only modes actually supported by the display are reported in the RandR 1.2 mode list. Scaling configurations, such as the 800x600 to 1920x1080 example above, need to be configured via the RandR 1.2 transform feature. Adding implicitly scaled modes to the mode list would conflict with the transform configuration options and reintroduce the same problems that the previous front-end/back-end timing system had. With the introduction of RandR 1.2 support to the 302.* driver series, the front-end/back-end timing system was abandoned, and the list of mode timings exposed by the NVIDIA X driver was simplified to include only those modes which would actually be driven by the hardware. Although it remained possible to manually configure all of the scaling configurations that were previously possible, and many scaling configurations which were previously impossible, this change resulted in some inconvenient losses of functionality: o Applications which used RandR 1.1 or earlier or XF86VidMode to set modes no longer had the implicitly scaled front-end timings available to them. Many displays have EDIDs which advertise only the display's native resolution, or a list of resolutions that is otherwise small, compared to the list that would previously have been exposed as front-end timings, preventing these applications from setting modes that were possible with previous versions of the NVIDIA driver. o The 'nvidia-settings' control panel, which formerly listed all available front-end modes for displays in its X Server Display Configuration page, only listed the actual back-end modes. Subsequent driver releases restored some of this functionality without reverting to the front-end/back-end system: o The NVIDIA X driver now builds a list of "Implicit MetaModes", which implicitly scale many common resolutions to a mode that is supported by the display. These modes are exposed to applications which use RandR 1.1 and XF86VidMode, as neither supports the scaling or other transform capabilities of RandR 1.2. o The resolution list in the 'nvidia-settings' X Server Display Configuration page now includes explicitly scaled modes for many common resolutions which are not directly supported by the display. To reduce confusion, the scaled modes are identified as being scaled, and it is not possible to set a refresh rate for any of the scaled modes. As mentioned previously, the RandR 1.2 mode list contains only modes which are supported by the display. Modern applications that wish to set modes other than those available in the RandR 1.2 mode list are encouraged to use RandR 1.2 transformations to program any required scaling operations. For example, the 'xrandr' utility can program RandR scaling transformations, and the following command can scale a 1280x720 mode to a display connected to output DVI-I-0 that does not support the desired mode, but does support 1920x1080: xrandr --output DVI-I-0 --mode 1920x1080 --scale-from 1280x720 ______________________________________________________________________________ Chapter 8. Common Problems ______________________________________________________________________________ This section provides solutions to common problems associated with the NVIDIA Linux x86_64 Driver. Q. My X server fails to start, and my X log file contains the error: (EE) NVIDIA(0): The NVIDIA kernel module does not appear to (EE) NVIDIA(0): be receiving interrupts generated by the NVIDIA graphics (EE) NVIDIA(0): device PCI:x:x:x. Please see the COMMON PROBLEMS (EE) NVIDIA(0): section in the README for additional information. A. This can be caused by a variety of problems, such as PCI IRQ routing errors, I/O APIC problems, conflicts with other devices sharing the IRQ (or their drivers), or MSI compatibility problems. If possible, configure your system such that your graphics card does not share its IRQ with other devices (try moving the graphics card to another slot if applicable, unload/disable the driver(s) for the device(s) sharing the card's IRQ, or remove/disable the device(s)). Depending on the nature of the problem, one of (or a combination of) these kernel parameters might also help: Parameter Behavior -------------- --------------------------------------------------- pci=noacpi don't use ACPI for PCI IRQ routing pci=biosirq use PCI BIOS calls to retrieve the IRQ routing table noapic don't use I/O APICs present in the system acpi=off disable ACPI The problem may also be caused by MSI compatibility problems. See "MSI Interrupts" for details. Q. X starts for me, but OpenGL applications terminate immediately. A. If X starts but you have trouble with OpenGL, you most likely have a problem with other libraries in the way, or there are stale symlinks. See Chapter 5 for details. Sometimes, all it takes is to rerun 'ldconfig'. You should also check that the correct extensions are present; % xdpyinfo should show the "GLX" and "NV-GLX" extensions present. If these two extensions are not present, then there is most likely a problem loading the glx module, or it is unable to implicitly load GLcore. Check your X config file and make sure that you are loading glx (see Chapter 6). If your X config file is correct, then check the X log file for warnings/errors pertaining to GLX. Also check that all of the necessary symlinks are in place (refer to Chapter 5). Q. When Xinerama is enabled, my stereo glasses are shuttering only when the stereo application is displayed on one specific X screen. When the application is displayed on the other X screens, the stereo glasses stop shuttering. A. This problem occurs with DDC and "blue line" stereo glasses, that get the stereo signal from one video port of the graphics card. When a X screen does not display any stereo drawable the stereo signal is disabled on the associated video port. Forcing stereo flipping allows the stereo glasses to shutter continuously. This can be done by enabling the OpenGL control "Force Stereo Flipping" in nvidia-settings, or by setting the X configuration option "ForceStereoFlipping" to "1". Q. Stereo is not in sync across multiple displays. A. There are two cases where this may occur. If the displays are attached to the same GPU, and one of them is out of sync with the stereo glasses, you will need to reconfigure your monitors to drive identical mode timings; see Chapter 18 for details. If the displays are attached to different GPUs, the only way to synchronize stereo across the displays is with a Quadro Sync device, which is only supported by certain NVIDIA RTX/Quadro cards. See Chapter 32 for details. Q. I just upgraded my kernel, and now the NVIDIA kernel module will not load. A. The kernel interface layer of the NVIDIA kernel module must be compiled specifically for the configuration and version of your kernel. If you upgrade your kernel, then the simplest solution is to reinstall the driver. ADVANCED: You can install the NVIDIA kernel module for a non running kernel (for example: in the situation where you just built and installed a new kernel, but have not rebooted yet) with a command line such as this: # sh NVIDIA-Linux-x86_64-570.123.06.run --kernel-name='KERNEL_NAME' Where 'KERNEL_NAME' is what 'uname -r' would report if the target kernel were running. Q. Installing the driver fails with: Unable to load the kernel module 'nvidia.ko'. My X server fails to start, and my X log file contains the error: (EE) NVIDIA(0): Failed to load the NVIDIA kernel module! A. `nvidia-installer` attempts to load the NVIDIA kernel module before installing the driver, and will abort if this test load fails. Similarly, if the kernel module fails to load when starting the an X server with the NVIDIA X driver, the X server will fail to start. If the NVIDIA kernel module fails to load, you should check the output of `dmesg` for kernel error messages and/or attempt to load the kernel module explicitly with `modprobe nvidia`. There are a number of common failure cases: o Some symbols that the kernel module depends on failed to be resolved. If this happens, then the kernel module was most likely built against a Linux kernel source tree (or kernel headers) for a kernel revision or configuration that doesn't match the running kernel. In some cases, the NVIDIA kernel module may fail to resolve symbols due to those symbols being provided by modules that were built as part of the configuration of the currently running kernel, but which are not installed. For example, some distributions, such as Ubuntu 14.04, provide the DRM kernel module in an optionally installed package (in the case of Ubuntu 14.04, linux-image-extra), but the kernel headers will reflect the availability of DRM regardless of whether the module that provides it is actually installed. The NVIDIA kernel module build will detect the availability of DRM when building, but will fail at load time with messages such as: nvidia: Unknown symbol drm_open (err 0) If any of the NVIDIA kernel modules fail to load due to unresolved symbols, and you are certain that the modules were built against the correct kernel source tree (or headers), check to see if there are any optionally installable modules that might provide these symbols which are not currently installed. If you believe that you might not be using the correct kernel sources/headers, you can specify their location when you install the NVIDIA driver using the --kernel-source-path command line option (see `sh NVIDIA-Linux-x86_64-570.123.06.run --advanced-options` for details). o Nouveau, or another driver, is already using the GPU. See Interaction with the Nouveau Driver for more information on Nouveau and how to disable it. o The kernel requires that kernel modules carry a valid signature from a trusted key, and the NVIDIA kernel module is unsigned, or has an invalid or untrusted signature. This may happen, for example, on some systems with UEFI Secure Boot enabled. See "Signing the NVIDIA Kernel Module" in Chapter 4 for more information about signing the kernel module. o No supported GPU is detected, either because no NVIDIA GPUs are detected in the system, or because none of the NVIDIA GPUs which are present are supported by this version of the NVIDIA kernel module. See Appendix A for information on which GPUs are supported by which driver versions. Q. Installing the NVIDIA kernel module gives an error message like: #error Modules should never use kernel-headers system headers #error but headers from an appropriate kernel-source A. You need to install the source for the Linux kernel. In most situations you can fix this problem by installing the kernel-source or kernel-devel package for your distribution Q. OpenGL applications crash and print out the following warning: WARNING: Your system is running with a buggy dynamic loader. This may cause crashes in certain applications. If you experience crashes you can try setting the environment variable __GL_SINGLE_THREADED to 1. For more information, consult the FREQUENTLY ASKED QUESTIONS section in the file /usr/share/doc/NVIDIA_GLX-1.0/README.txt. A. The dynamic loader on your system has a bug which will cause applications linked with pthreads, and that dlopen() libGL multiple times, to crash. This bug is present in older versions of the dynamic loader. Distributions that shipped with this loader include but are not limited to Red Hat Linux 6.2 and Mandrake Linux 7.1. Version 2.2 and later of the dynamic loader are known to work properly. If the crashing application is single threaded then setting the environment variable '__GL_SINGLE_THREADED' to "1" will prevent the crash. In the bash shell you would enter: % export __GL_SINGLE_THREADED=1 and in csh and derivatives use: % setenv __GL_SINGLE_THREADED 1 Previous releases of the NVIDIA Accelerated Linux Graphics Driver attempted to work around this problem. Unfortunately, the workaround caused problems with other applications and was removed after version 1.0-1541. Q. Quake3 crashes when changing video modes. A. You are probably experiencing a problem described above. Please check the text output for the "WARNING" message described in the previous hint. Setting '__GL_SINGLE_THREADED' to "1" as will fix the problem. Q. I cannot build the NVIDIA kernel module, or, I can build the NVIDIA kernel module, but modprobe/insmod fails to load the module into my kernel. A. These problems are generally caused by the build using the wrong kernel header files (i.e. header files for a different kernel version than the one you are running). The convention used to be that kernel header files should be stored in '/usr/include/linux/', but that is deprecated in favor of '/lib/modules/RELEASE/build/include' (where RELEASE is the result of 'uname -r'. The 'nvidia-installer' should be able to determine the location on your system; however, if you encounter a problem you can force the build to use certain header files by using the --kernel-include-dir option. For this to work you will of course need the appropriate kernel header files installed on your system. Consult the documentation that came with your distribution; some distributions do not install the kernel header files by default, or they install headers that do not coincide properly with the kernel you are running. Q. Compiling the NVIDIA kernel module gives this error: You appear to be compiling the NVIDIA kernel module with a compiler different from the one that was used to compile the running kernel. This may be perfectly fine, but there are cases where this can lead to unexpected behavior and system crashes. If you know what you are doing and want to override this check, you can do so by setting IGNORE_CC_MISMATCH. In any other case, set the CC environment variable to the name of the compiler that was used to compile the kernel. A. You should compile the NVIDIA kernel module with the same compiler version that was used to compile your kernel. Some Linux kernel data structures are dependent on the version of gcc used to compile it; for example, in 'include/linux/spinlock.h': ... * Most gcc versions have a nasty bug with empty initializers. */ #if (__GNUC__ > 2) typedef struct { } rwlock_t; #define RW_LOCK_UNLOCKED (rwlock_t) { } #else typedef struct { int gcc_is_buggy; } rwlock_t; #define RW_LOCK_UNLOCKED (rwlock_t) { 0 } #endif If the kernel is compiled with gcc 2.x, but gcc 3.x is used when the kernel interface is compiled (or vice versa), the size of rwlock_t will vary, and things like ioremap will fail. To check what version of gcc was used to compile your kernel, you can examine the output of: % cat /proc/version To check what version of gcc is currently in your '$PATH', you can examine the output of: % gcc -v Q. I recently updated various libraries on my system using my Linux distributor's update utility, and the NVIDIA graphics driver no longer works. A. Conflicting libraries may have been installed by your distribution's update utility; see Chapter 5 for details on how to diagnose this. Q. I have rebuilt the NVIDIA kernel module, but when I try to insert it, I get a message telling me I have unresolved symbols. A. Unresolved symbols are most often caused by a mismatch between your kernel sources and your running kernel. They must match for the NVIDIA kernel module to build correctly. Make sure your kernel sources are installed and configured to match your running kernel. Q. OpenGL applications leak significant amounts of memory on my system! A. If your kernel is making use of the -rmap VM, the system may be leaking memory due to a memory management optimization introduced in -rmap14a. The -rmap VM has been adopted by several popular distributions, the memory leak is known to be present in some of the distribution kernels; it has been fixed in -rmap15e. If you suspect that your system is affected, try upgrading your kernel or contact your distribution's vendor for assistance. Q. When I try to install the driver, the installer claims that X is running, even though I have exited X. A. The installer detects the presence of an X server by checking for the X server's lock files: '/tmp/.Xn-lock', where 'n' is the number of the X Display (the installer checks for X Displays 0-7). If you have exited X, but one of these files has been left behind, then you will need to manually delete the lock file. DO NOT remove this file if X is still running! Q. Why does the VBIOS fail to load on my Optimus system? A. On some notebooks with Optimus graphics, the NVIDIA driver may not be able to retrieve the Video BIOS due to interactions between the System BIOS and the Linux kernel's ACPI subsystem. On affected notebooks, applications that require the GPU will fail, and messages like the following may appear in the system log: NVRM: failed to copy vbios to system memory. NVRM: RmInitAdapter failed! (0x30:0xffffffff:858) NVRM: rm_init_adapter(0) failed Such problems are typically beyond the control of the NVIDIA driver, which relies on proper cooperation of ACPI and the System BIOS to retrieve important information about the GPU, including the Video BIOS. Q. OpenGL applications do not work with driver version 364.xx and later, which worked with previous driver versions A. Release 361 of the NVIDIA Linux driver introduced OpenGL libraries built upon the libglvnd (GL Vendor Neutral Dispatch) architecture, to allow for the coexistence of multiple OpenGL implementations on the same system. Beginning with release 364, the GLVND libraries are installed by default. By design, GLVND conforms with the Linux OpenGL ABI version 1.0 as defined at https://www.opengl.org/registry/ABI/ and exposes all required entry points; however, applications which depend upon specifics of the NVIDIA OpenGL implementation which fall outside of the OpenGL ABI may be incompatible with a GLVND-based OpenGL implementation. If you encounter an application which is incompatible with GLVND, please contact the application vendor to inform them that their application will need to be updated to ensure compatibility with GLVND. Q. Vulkan applications crash when entering or leaving fullscreen, or when resized A. Resizing a Vulkan application generates events that trigger an out-of-date swapchain. Fullscreen Vulkan applications are optimized for performance by the driver. This optimization also generates events that trigger an out-of-date swapchain upon entering or leaving fullscreen mode. This is commonly encountered when using the alt-tab key combination, for example. Applications that do not properly respond to the VK_ERROR_OUT_OF_DATE_KHR return code may not function properly when these events occur. The expected behavior is documented in section 30.8 of the Vulkan specification. Q. The Vulkan ICD has dependencies on X libraries A. By default, nvidia-installer creates a /etc/vulkan/icd.d/nvidia_icd.json that points to libGLX_nvidia.so.0. This DSO has dependencies on X libraries. It is possible to avoid those dependencies by hand editing that file to point to libEGL_nvidia.so.0 instead. However in that case, an application will only be able to create non-X swapchains if it wants to present frames. Q. Vulkan applications show up as a blank window A. The Vulkan application may be choosing your iGPU instead of your NVIDIA dGPU. Starting with the 435 driver series, you can set the following environment variable to force the Vulkan loader to list NVIDIA GPUs before other devices: __NV_PRIME_RENDER_OFFLOAD=1 This can be used even if the system is not otherwise configured to use PRIME Render Offload. In order to prefer discrete GPUs over integrated GPUs, application developers can check the VkPhysicalDeviceProperties::deviceType field when querying available physical devices. Q. My Wayland compositor or Vulkan Direct-To-Display application does not start as a regular user. A. Starting in the 570 release series, the nvidia-drm driver enables DRM fbdev support by default, when the nvidia-drm "modeset=1" kernel module parameter is also enabled. In this configuration, applications that require modeset capability will need read/write access to the "/dev/dri/cardN" device file corresponding to the display GPU, or they may fail to start. Common applications include Wayland compositors and Vulkan direct-to-display. Many Linux distributions use the "video" UNIX group to provide read/write access to the DRI device nodes. If that is the case, then adding your user to the "video" group may solve this problem. Q. OpenGL applications are running slowly A. The application is probably using a different library that still remains on your system, rather than the NVIDIA supplied OpenGL library. See Chapter 5 for details. Q. Fonts are incorrectly sized after installing the NVIDIA driver. A. Incorrectly sized fonts are generally caused by incorrect DPI (Dots Per Inch) information. You can check what X thinks the physical size of your monitor is, by running: % xdpyinfo | grep dimensions This will report the size in pixels, and in millimeters. If these numbers are wrong, you can correct them by modifying the X server's DPI setting. See Appendix E for details. Q. OpenGL applications don't work, and my X log file contains the error: (EE) NVIDIA(0): Unable to map device node /dev/zero with read and write (EE) NVIDIA(0): privileges. The GLX extension will be disabled on this (EE) NVIDIA(0): X screen. Please see the COMMON PROBLEMS section in the (EE) NVIDIA(0): README for more information. A. The NVIDIA OpenGL driver must be able to map anonymous memory with read and write execute privileges in order to function correctly. The driver needs this ability to allocate aligned memory, which is used for certain optimizations. Currently, GLX cannot run without these optimizations. Q. My log file contains a message like the following: (WW) NVIDIA(GPU-0): Unable to enter interactive mode, because non-interactive (WW) NVIDIA(GPU-0): mode has been previously requested. The most common (WW) NVIDIA(GPU-0): cause is that a GPU compute application is currently (WW) NVIDIA(GPU-0): running. Please see the README for details. A. This indicates that the X driver was not able to put the GPU in interactive mode, because another program has requested non-interactive mode. The GPU watchdog will not run, and long-running GPU compute programs may cause the X server and OpenGL programs to hang. If you intend to run long-running GPU compute programs, set the "Interactive" option to "off" to disable interactive mode. Q. I see a blank screen or an error message instead of a login screen or desktop session A. Installation or configuration problems may prevent the X server, a login/session manager, or a desktop environment from starting correctly. If your system is failing to display a login screen, or failing to start a desktop session, try the following troubleshooting steps: o Make sure that you are using the correct X driver for your configuration. Recent X servers will be able to automatically select the correct X driver in many cases, but if your X server does not automatically select the correct driver, you may need to manually configure it. For example, systems with multiple GPUs will likely require a PCI BusID in the "Device" section of the X configuration file, in order to specify which GPU is to be used. If you are planning to use NVIDIA GPUs for graphics, you can run the 'nvidia-xconfig' utility to automatically generate a simple X configuration file that uses the NVIDIA X driver. If you are not using NVIDIA GPUs for graphics (e.g. on a server system where displays are driven by an onboard graphics controller, and NVIDIA GPUs are used for non-graphical computational purposes only), DO NOT run 'nvidia-xconfig'. o Some recent desktop environments (e.g. GNOME 3, Unity), window managers (e.g. mutter, compiz), and session managers (e.g. gdm3) require a working OpenGL driver in order to function correctly. In addition to making sure that the X server is configured to use the correct X driver for your configuration, please ensure that you are using the correct OpenGL driver to match your X driver. If you are not using NVIDIA GPUs for graphical purposes, try installing the driver with the --no-opengl-files option on the installer's command line to prevent the installer from overwriting any existing OpenGL installation, which may be needed for proper OpenGL functionality on whichever graphics controller is to be used on the system. o Some desktop environments (e.g. GNOME 3, Unity) and window managers (e.g. mutter) do not properly support multiple X screens, leaving you with a blank screen displaying only a cursor on the non-primary X screen. If you encounter such a problem, try configuring X with a single X screen, or switching to a different desktop environment or window manager. o Desktop environments, window managers, and session managers that require OpenGL typically also require the X Composite extension. If you have disabled the Composite extension, either explicitly, or by enabling a feature that is not compatible with it, try re-enabling the extension (possibly by disabling any incompatible features). If you are unable to satisfy your desired use case with the Composite extension enabled, try switching to a different desktop environment, window manager, and/or session manager that does not require Composite. o Check the X log (e.g. '/var/log/Xorg.0.log') for additional errors not covered above. Warning or error messages in the log may highlight a specific problem that can be fixed with a configuration adjustment. Q. The display settings I configured in 'nvidia-settings' do not persist. A. Depending on the type of configuration being performed, 'nvidia-settings' will save configuration changes to one of several places: o Static X server configuration changes are saved to the X configuration file (e.g. '/etc/X11/xorg.conf'). These settings are loaded by the X server when it starts, and cannot be changed without restarting X. o Dynamic, user-specific configuration changes are saved to '~/.nvidia-settings-rc'. 'nvidia-settings' loads this file and applies any settings contained within. These settings can be changed without restarting the X server, and can typically be configured through the 'nvidia-settings' command line interface as well, or via the RandR and/or NV-CONTROL APIs. o User-specific application profiles edited in 'nvidia-settings' are saved to '~/.nv/nvidia-application-profiles-rc'. This file is loaded along with the other files in the application profile search path by the NVIDIA OpenGL driver when it is loaded by an OpenGL application. The driver evaluates the application profiles to determine which settings apply to the application. Changes made to this configuration file while an application is already running will be applied when the application is next restarted. See Appendix J for more information about application profiles. Settings in '~/.nvidia-settings-rc' only take effect when processed by 'nvidia-settings', and therefore will not be loaded by default when starting a new X session. To load settings from '~/.nvidia-settings-rc' without actually opening the 'nvidia-settings' control panel, use the --load-config-only option on the 'nvidia-settings' command line. 'nvidia-settings --load-config-only' can be added to your login scripts to ensure that your settings are restored when starting a new desktop session. Even after 'nvidia-settings' has been run to restore any settings set in '~/.nvidia-settings-rc', some desktop environments (e.g. GNOME, KDE, Unity, Xfce) include advanced display configuration tools that may override settings that were configured via 'nvidia-settings'. These tools may attempt to restore their own display configuration when starting a new desktop session, or when events such as display hotplugs, resolution changes, or VT switches occur. These tools may also override some types of settings that are stored in and loaded from the X configuration file, such as any MetaMode strings that may specify the initial display layouts of NVIDIA X screens. Although the configuration of the initial MetaMode is static, it is possible to dynamically switch to a different MetaMode after X has started. This can have the effect of making the set of active displays, their resolutions, and layout positions as configured in the 'nvidia-settings' control panel appear to be ineffective, when in reality, this configuration was active when starting X and then overridden later by the desktop environment. If you believe that your desktop environment is overriding settings that you configured in 'nvidia-settings', some possible solutions are: o Use the display configuration tools provided as part of the desktop environment (e.g. 'gnome-control-center display', 'gnome-display-properties', 'kcmshell4 display', 'unity-control-center display', 'xfce4-display-settings') to configure your displays, instead of the 'nvidia-settings' control panel or the 'xrandr' command line tool. Setting your desired configuration using the desktop environment's tools should cause that configuration to be the one which is restored when the desktop environment overrides the existing configuration from 'nvidia-settings'. If you are not sure which tools your desktop environment uses for display configuration, you may be able to discover them by navigating any available system menus for "Display" or "Monitor" control panels. o For settings loaded from '~/.nvidia-settings-rc' which have been overridden, run 'nvidia-settings --load-config-only' as needed to reload the settings from '~/.nvidia-settings-rc'. o Disable any features your desktop environment may have for managing displays. (Note: this may disable other features, such as display configuration tools that are integrated into the desktop.) o Use a different desktop environment which does not actively manage display configuration, or do not use any desktop environment at all. Some systems may have multiple different display configuration utilities, each with its own way of managing settings. In addition to conflicting with 'nvidia-settings', such tools may conflict with each other. If your system uses more than one tool for configuring displays, make sure to check the configuration of each tool when attempting to determine the source of any unexpected display settings. Q. My displays are reconfigured in unexpected ways when I plug in or unplug a display, or power a display off and then power it on again. A. This is a special case of the issues described in "Q. The display settings I configured in nvidia-settings do not persist." in Chapter 8. Some desktop environments which include advanced display configuration tools will automatically configure the display layout in response to detected configuration changes. For example, when a new display is plugged in, such a desktop environment may attempt to restore the previous layout that was used with the set of currently connected displays, or may configure a default layout based upon its own policy. On X servers with support for RandR 1.2 or later, the NVIDIA X driver reports display hotplug events to the X server via RandR when displays are connected and disconnected. These hotplug events may trigger a desktop environment with advanced display management capabilities to change the display configuration. These changes may affect settings such as the set of active displays, their resolutions and positioning relative to each other, per-display color correction settings, and more. In addition to hotplug events generated by connecting or disconnecting displays, DisplayPort displays will generate a hot unplug event when they power off, and a hotplug event when they power on, even if no physical plugging in or unplugging takes place. This can lead to hotplug-induced display configuration changes without any actual hotplug action taking place. Upon suspend, the NVIDIA X driver will incur an implicit VT switch. If a DisplayPort monitor is powered off when a VT switch or modeset occurs, RandR will forget the configuration of that monitor. As a result, the display will be left without a mode once powered back on. In the absence of an RandR-aware window manager, bringing back the display will require manually configuring it with RandR. If display hotplug events are resulting in undesired configuration changes, try the solutions and workarounds listed in "Q. The display settings I configured in nvidia-settings do not persist." in Chapter 8. Another workaround would be to disable the NVIDIA X driver's reporting of hotplug events with the "UseHotplugEvents" X configuration option. Note that this option will have no effect on DisplayPort devices, which must report all hotplug events to ensure proper functionality. Q. My remote desktop software doesn't work when no displays are attached. A. Some desktop environments or window managers might not function correctly if no displays are exposed via the RandR API. When the NVIDIA X driver is configured for headless operation, either with 'Option "AllowEmptyInitialConfiguration"' and no connected displays, with 'Option "MetaModes" "NULL"', or with 'Option "UseDisplayDevice" "none"', it will not expose any displays via RandR. To work around software that requires RandR displays on a machine that is only intended to be used remotely, you can connect a physical display, or configure the X driver to behave as if a physical display were connected using the "ConnectedMonitor" option. See Appendix B for additional information about the "ConnectedMonitor" option. INTERACTION WITH THE NOUVEAU DRIVER Q. What is Nouveau, and why do I need to disable it? A. Nouveau is a display driver for NVIDIA GPUs, developed as an open-source project through reverse-engineering of the NVIDIA driver. It ships with many current Linux distributions as the default display driver for NVIDIA hardware. It is not developed or supported by NVIDIA, and is not related to the NVIDIA driver, other than the fact that both Nouveau and the NVIDIA driver are capable of driving NVIDIA GPUs. Only one driver can control a GPU at a time, so if a GPU is being driven by the Nouveau driver, Nouveau must be disabled before installing the NVIDIA driver. Nouveau performs modesets in the kernel. This can make disabling Nouveau difficult, as the kernel modeset is used to display a framebuffer console, which means that Nouveau will be in use even if X is not running. As long as Nouveau is in use, its kernel module cannot be unloaded, which will prevent the NVIDIA kernel module from loading. It is therefore important to make sure that Nouveau's kernel modesetting is disabled before installing the NVIDIA driver. Q. How do I prevent Nouveau from loading and performing a kernel modeset? A. A simple way to prevent Nouveau from loading and performing a kernel modeset is to add configuration directives for the module loader to a file in one of the system's module loader configuration directories: for example, '/etc/modprobe.d/' or '/usr/local/modprobe.d'. These configuration directives can technically be added to any file in these directories, but many of the existing files in these directories are provided and maintained by your distributor, which may from time to time provide updated configuration files which could conflict with your changes. Therefore, it is recommended to create a new file, for example, '/etc/modprobe.d/disable-nouveau.conf', rather than editing one of the existing files, such as the popular '/etc/modprobe.d/blacklist.conf'. Note that some module loaders will only look for configuration directives in files whose names end with '.conf', so if you are creating a new file, make sure its name ends with '.conf'. Whether you choose to create a new file or edit an existing one, the following two lines will need to be added: blacklist nouveau options nouveau modeset=0 The first line will prevent Nouveau's kernel module from loading automatically at boot. It will not prevent manual loading of the module, and it will not prevent the X server from loading the kernel module; see "How do I prevent the X server from loading Nouveau?" below. The second line will prevent Nouveau from doing a kernel modeset. Without the kernel modeset, it is possible to unload Nouveau's kernel module, in the event that it is accidentally or intentionally loaded. You will need to reboot your system after adding these configuration directives in order for them to take effect. If nvidia-installer detects Nouveau is in use by the system, it will offer to create such a modprobe configuration file to disable Nouveau. Q. What if my initial ramdisk image contains Nouveau? A. Some distributions include Nouveau in an initial ramdisk image (henceforth referred to as "initrd" in this document, and sometimes also known as "initramfs"), so that Nouveau's kernel modeset can take place as early as possible in the boot process. This poses an additional challenge to those who wish to prevent the modeset from occurring, as the modeset will occur while the system is executing within the initrd, before any directives in the module loader configuration files are processed. If you have an initrd which loads the Nouveau driver, you will additionally need to ensure that Nouveau is disabled in the initrd. In most cases, rebuilding the initrd will pick up the module loader configuration files, including any which may disable Nouveau. Please consult your distribution's documentation on how to rebuild the initrd, as different distributions have different tools for building and modifying the initrd. Some popular distro initrd tools include: 'dracut', 'mkinitrd', and 'update-initramfs'. Some initrds understand the rdblacklist parameter. On these initrds, as an alternative to rebuilding the initrd, you can add the option rdblacklist=nouveau to your kernel's boot parameters. On initrds that do not support rdblacklist, it is possible to prevent Nouveau from performing a kernel modeset by adding the option nouveau.modeset=0 to your kernel's boot parameters. Note that nouveau.modeset=0 will prevent a kernel modeset, but it may not prevent Nouveau from being loaded, so rebuilding the initrd or using rdblacklist may be more effective than using nouveau.modeset=0. Any changes to the default kernel boot parameters should be made in your bootloader's configuration file(s), so that the options get passed to your kernel every time the system is booted. Please consult your distribution's documentation on how to configure your bootloader, as different distributions use different bootloaders and configuration files. Q. How do I prevent the X server from loading Nouveau? A. Blacklisting Nouveau will only prevent it from being loaded automatically at boot. If an X server is started as part of the normal boot process, and that X server uses the Nouveau X driver, then the Nouveau kernel module will still be loaded. Should this happen, you will be able to unload Nouveau with `modprobe -r nouveau` after stopping the X server, as long as you have taken care to prevent it from doing a kernel modeset; however, it is probably better to just make sure that X does not load Nouveau in the first place. If your system is not configured to start an X server at boot, then you can simply run the NVIDIA driver installer after rebooting. Otherwise, the easiest thing to do is to edit your X server's configuration file so that your X server uses a non-modesetting driver that is compatible with your card, such as the 'vesa' driver. You can then stop X and install the driver as usual. Please consult your X server's documentation to determine where your X server configuration file is located. ______________________________________________________________________________ Chapter 9. Known Issues ______________________________________________________________________________ The following problems still exist in this release and are in the process of being resolved. Known Issues Cache Aliasing Cache aliasing occurs when multiple mappings to a physical page of memory have conflicting caching states, such as cached and uncached. Due to these conflicting states, data in that physical page may become corrupted when the processor's cache is flushed. If that page is being used for DMA by a driver such as NVIDIA's graphics driver, this can lead to hardware stability problems and system lockups. NVIDIA has encountered bugs with some Linux kernel versions that lead to cache aliasing. Although some systems will run perfectly fine when cache aliasing occurs, other systems will experience severe stability problems, including random lockups. Users experiencing stability problems due to cache aliasing will benefit from updating to a kernel that does not cause cache aliasing to occur. Valgrind The NVIDIA OpenGL implementation makes use of self modifying code. To force Valgrind to retranslate this code after a modification you must run using the Valgrind command line option: --smc-check=all Without this option Valgrind may execute incorrect code causing incorrect behavior and reports of the form: ==30313== Invalid write of size 4 Driver fails to initialize when MSI interrupts are enabled The Linux NVIDIA driver uses Message Signaled Interrupts (MSI) by default. This provides compatibility and scalability benefits, mainly due to the avoidance of IRQ sharing. Some systems have been seen to have problems supporting MSI, while working fine with virtual wire interrupts. These problems manifest as an inability to start X with the NVIDIA driver, or CUDA initialization failures. The NVIDIA driver will then report an error indicating that the NVIDIA kernel module does not appear to be receiving interrupts generated by the GPU. Problems have also been seen with suspend/resume while MSI is enabled. All known problems have been fixed, but if you observe problems with suspend/resume that you did not see with previous drivers, disabling MSI may help you. NVIDIA is working on a long-term solution to improve the driver's out of the box compatibility with system configurations that do not fully support MSI. MSI interrupts can be disabled via the NVIDIA kernel module parameter "NVreg_EnableMSI=0". This can be set on the command line when loading the module, or more appropriately via your distribution's kernel module configuration files (such as those under /etc/modprobe.d/). Console restore behavior The Linux NVIDIA driver uses the nvidia-modeset module for console restore whenever it can. Currently, the improved console restore mechanism is used on systems that boot with the UEFI Graphics Output Protocol driver, and on systems that use supported VESA linear graphical modes. Note that VGA text, color index, planar, banked, and some linear modes cannot be supported, and will use the older console restore method instead. When the new console restore mechanism is in use and the nvidia-modeset module is initialized (e.g. because an X server is running on a different VT, nvidia-persistenced is running, or the nvidia_drm module is loaded with the "modeset=1" parameter), then nvidia-modeset will respond to hot plug events by displaying the console on as many displays as it can. Note that to save power, it may not display the console on all connected displays. Vulkan and device enumeration Starting with the X.Org X server version 1.20.7, it is possible to enumerate all the NVIDIA devices in the system if the application is able to open a connection to the X server. However, such applications will only be able to create an Xlib or XCB swapchain on the device driving the X screen. Such a device can be identified by using the vkGetPhysicalDeviceSurfaceSupportKHR() API. Prior to the X.Org X server version 1.20.7, it is not possible to enumerate multiple devices if one of them will be used to present to an X11 swapchain. It is still possible to enumerate multiple devices even if one of them is driving an X screen, if the devices will be used for Vulkan offscreen rendering or presenting to a display swapchain. For that, make sure that the application cannot open a display connection to an X server by, for example, unsetting the DISPLAY environment variable. Restricting access to GPU performance counters NVIDIA Developer Tools allow developers to debug, profile, and develop software for NVIDIA GPUs. GPU performance counters are integral to these tools. By default, access to the GPU performance counters is restricted to root, and other users with the CAP_SYS_ADMIN capability, for security reasons. If developers require access to the NVIDIA Developer Tools, a system administrator can accept the security risk and allow access to users without the CAP_SYS_ADMIN capability. Wider access to GPU performance counters can be granted by setting the kernel module parameter "NVreg_RestrictProfilingToAdminUsers=0" in the nvidia.ko kernel module. This can be set on the command line when loading the module, or more appropriately via your distribution's kernel module configuration files (such as those under /etc/modprobe.d/). Driver fails to initialize with some versions of RHEL 8 Some versions of Red Hat Enterprise Linux 8 kernels have a bug that causes driver initialization to fail with an error such as: NVRM: Xid (PCI:0000:09:00): 79, pid=2172, GPU has fallen off the bus. NVRM: GPU 0000:09:00.0: GPU has fallen off the bus. NVRM: GPU 0000:09:00.0: RmInitAdapter failed! (0x26:0x65:1239) NVRM: GPU 0000:09:00.0: rm_init_adapter failed, device minor number 0 See the Red Hat knowledge base article https://access.redhat.com/solutions/5825061 to find the specific affected and fixed kernel versions. Driver fails to load on Linux kernel versions 5.18 through 5.18.19 with CONFIG_X86_KERNEL_IBT enabled The NVIDIA driver fails to load on IBT (Indirect Branch Tracking) supported CPUs running Linux kernel versions 5.18 to 5.18.19, when IBT is enabled, with the following error: error "traps: Missing ENDBR:" This issue is not seen with Linux kernels having the following commit: commit 3c6f9f77e618 (objtool: Rework ibt and extricate from stack validation) The aforementioned commit is available in Linux kernel versions 5.19 and later. The NVIDIA driver's IBT support works with Linux kernels containing commit 3c6f9f77e618 (5.19 and later). Please use the kernel boot parameter "ibt=off" as a workaround on kernels without that commit. Notebooks If you are using a notebook see the "Known Notebook Issues" in Chapter 16. Texture seams in Quake 3 engine Many games based on the Quake 3 engine set their textures to use the "GL_CLAMP" clamping mode when they should be using "GL_CLAMP_TO_EDGE". This was an oversight made by the developers because some legacy NVIDIA GPUs treat the two modes as equivalent. The result is seams at the edges of textures in these games. To mitigate this, older versions of the NVIDIA display driver remap "GL_CLAMP" to "GL_CLAMP_TO_EDGE" internally to emulate the behavior of the older GPUs, but this workaround has been disabled by default. To re-enable it, uncheck the "Use Conformant Texture Clamping" checkbox in nvidia-settings before starting any affected applications. FSAA When FSAA is enabled (the __GL_FSAA_MODE environment variable is set to a value that enables FSAA and a multisample visual is chosen), the rendering may be corrupted when resizing the window. libGL DSO finalizer and pthreads When a multithreaded OpenGL application exits, it is possible for libGL's DSO finalizer (also known as the destructor, or "_fini") to be called while other threads are executing OpenGL code. The finalizer needs to free resources allocated by libGL. This can cause problems for threads that are still using these resources. Setting the environment variable "__GL_NO_DSO_FINALIZER" to "1" will work around this problem by forcing libGL's finalizer to leave its resources in place. These resources will still be reclaimed by the operating system when the process exits. Note that the finalizer is also executed as part of dlclose(3), so if you have an application that dlopens(3) and dlcloses(3) libGL repeatedly, "__GL_NO_DSO_FINALIZER" will cause libGL to leak resources until the process exits. Using this option can improve stability in some multithreaded applications, including Java3D applications. Thread cancellation Canceling a thread (see pthread_cancel(3)) while it is executing in the OpenGL driver causes undefined behavior. For applications that wish to use thread cancellation, it is recommended that threads disable cancellation using pthread_setcancelstate(3) while executing OpenGL or GLX commands. This section describes problems that will not be fixed. Usually, the source of the problem is beyond the control of NVIDIA. Following is the list of problems: Problems that Will Not Be Fixed NV-CONTROL versions 1.8 and 1.9 Version 1.8 of the NV-CONTROL X Extension introduced target types for setting and querying attributes as well as receiving event notification on targets. Targets are objects like X Screens, GPUs and Quadro Sync devices. Previously, all attributes were described relative to an X Screen. These new bits of information (target type and target id) were packed in a non-compatible way in the protocol stream such that addressing X Screen 1 or higher would generate an X protocol error when mixing NV-CONTROL client and server versions. This packing problem has been fixed in the NV-CONTROL 1.10 protocol, making it possible for the older (1.7 and prior) clients to communicate with NV-CONTROL 1.10 servers. Furthermore, the NV-CONTROL 1.10 client library has been updated to accommodate the target protocol packing bug when communicating with a 1.8 or 1.9 NV-CONTROL server. This means that the NV-CONTROL 1.10 client library should be able to communicate with any version of the NV-CONTROL server. NVIDIA recommends that NV-CONTROL client applications relink with version 1.10 or later of the NV-CONTROL client library (libXNVCtrl.a, in the nvidia-settings-570.123.06.tar.bz2 tarball). The version of the client library can be determined by checking the NV_CONTROL_MAJOR and NV_CONTROL_MINOR definitions in the accompanying nv_control.h. The only web released NVIDIA Linux driver that is affected by this problem (i.e., the only driver to use either version 1.8 or 1.9 of the NV-CONTROL X extension) is 1.0-8756. CPU throttling reducing memory bandwidth on IGP systems For some models of CPU, the CPU throttling technology may affect not only CPU core frequency, but also memory frequency/bandwidth. On systems using integrated graphics, any reduction in memory bandwidth will affect the GPU as well as the CPU. This can negatively affect applications that use significant memory bandwidth, such as video decoding using VDPAU, or certain OpenGL operations. This may cause such applications to run with lower performance than desired. To work around this problem, NVIDIA recommends configuring your CPU throttling implementation to avoid reducing memory bandwidth. This may be as simple as setting a certain minimum frequency for the CPU. Depending on your operating system and/or distribution, this may be as simple as writing to a configuration file in the /sys or /proc filesystems, or other system configuration file. Please read, or search the Internet for, documentation regarding CPU throttling on your operating system. VDPAU initialization failures on supported GPUs If VDPAU gives the VDP_STATUS_NO_IMPLEMENTATION error message on a GPU which was labeled or specified as supporting PureVideo or PureVideo HD, one possible reason is a hardware defect. After ruling out any other software problems, NVIDIA recommends returning the GPU to the manufacturer for a replacement. Some applications, such as Quake 3, crash after querying the OpenGL extension string Some applications have bugs that are triggered when the extension string is longer than a certain size. As more features are added to the driver, the length of this string increases and can trigger these sorts of bugs. You can limit the extensions listed in the OpenGL extension string to the ones that appeared in a particular version of the driver by setting the "__GL_ExtensionStringVersion" environment variable to a particular version number. For example, __GL_ExtensionStringVersion=17700 quake3 will run Quake 3 with the extension string that appeared in the 177.* driver series. Limiting the size of the extension string can work around this sort of application bug. gnome-shell doesn't update until a window is moved Versions of libcogl prior to 1.10.x have a bug which causes glBlitFramebuffer() calls used to update the window to be clipped by a 0x0 scissor (see https://bugzilla.gnome.org/show_bug.cgi?id=690451 for more details). To work around this bug, the scissor test can be disabled by setting the "__GL_ConformantBlitFramebufferScissor" environment variable to 0. Note this version of the NVIDIA driver comes with an application profile which automatically disables this test if libcogl is detected in the process. Some X servers ignore the RandR transform filter during a modeset request The RandR layer of the X server attempts to ignore redundant RRSetCrtcConfig requests. If the only property changed by an RRSetCrtcConfig request is the transform filter, some X servers will ignore the request as redundant. This can be worked around by also changing other properties, such as the mode, transformation matrix, etc. ______________________________________________________________________________ Chapter 10. Allocating DMA Buffers on 64-bit Platforms ______________________________________________________________________________ NVIDIA GPUs have limits on how much physical memory they can address. This directly impacts DMA buffers, as a DMA buffer allocated in physical memory that is unaddressable by the NVIDIA GPU cannot be used (or may be truncated, resulting in bad memory accesses). See Chapter 41 for details on the addressing limitations of specific GPUs. When an NVIDIA GPU has an addressing limit less than the maximum possible physical system memory address, the NVIDIA Linux driver will use the __GFP_DMA32 flag to limit system memory allocations to the lowest 4 GB of physical memory in order to guarantee accessibility. This restriction applies even if there is hardware capable of remapping physical memory into an accessible range present, such as an IOMMU, because the NVIDIA Linux driver cannot determine at the time of memory allocation whether the memory can be remapped. This limitation can significantly reduce the amount of physical memory available to the NVIDIA GPU in some configurations. The Linux kernel requires that device drivers use the DMA remapping APIs to make physical memory accessible to devices, even when no remapping hardware is present. The NVIDIA Linux driver generally adheres to this requirement, except when it detects that the remapping is implemented using the SWIOTLB, which is not supported by the NVIDIA Linux driver. When the NVIDIA Linux driver detects that the SWIOTLB is in use, it will instead calculate the correct bus address needed to access a physical allocation instead of calling the kernel DMA remapping APIs to do so, as SWIOTLB space is very limited and exhaustion can result in a kernel panic. The NVIDIA Linux driver does not generally limit its usage of the Linux kernel DMA remapping APIs, and this can result in IOMMU space exhaustion when large amounts of physical memory are remapped for use by the NVIDIA GPU. Most modern IOMMU drivers generally fail gracefully when IOMMU space is exhausted, but NVIDIA recommends configuring the IOMMU in such a way to avoid resource exhaustion if possible, either by increasing the size of the IOMMU or disabling the IOMMU. On AMD's AMD64 platform, the size of the IOMMU can be configured in the system BIOS or, if no IOMMU BIOS option is available, using the 'iommu=memaper' kernel parameter. This kernel parameter expects an order and instructs the Linux kernel to create an IOMMU of size 32 MB^order overlapping physical memory. If the system's default IOMMU is smaller than 64 MB, the Linux kernel automatically replaces it with a 64 MB IOMMU. ______________________________________________________________________________ Chapter 11. Specifying OpenGL Environment Variable Settings ______________________________________________________________________________ 11A. FULL SCENE ANTIALIASING Antialiasing is a technique used to smooth the edges of objects in a scene to reduce the jagged "stairstep" effect that sometimes appears. By setting the appropriate environment variable, you can enable full-scene antialiasing in any OpenGL application on these GPUs. Several antialiasing methods are available and you can select between them by setting the __GL_FSAA_MODE environment variable appropriately. Note that increasing the number of samples taken during FSAA rendering may decrease performance. To see the available values for __GL_FSAA_MODE along with their descriptions, run: nvidia-settings --query=fsaa --verbose The __GL_FSAA_MODE environment variable uses the same integer values that are used to configure FSAA through nvidia-settings and the NV-CONTROL X extension. In other words, these two commands are equivalent: export __GL_FSAA_MODE=5 nvidia-settings --assign FSAA=5 Note that there are three FSAA related configuration attributes (FSAA, FSAAAppControlled and FSAAAppEnhanced) which together determine how a GL application will behave. If FSAAAppControlled is 1, the FSAA specified through nvidia-settings will be ignored, in favor of what the application requests through FBConfig selection. If FSAAAppControlled is 0 but FSAAAppEnhanced is 1, then the FSAA value specified through nvidia-settings will only be applied if the application selected a multisample FBConfig. Therefore, to be completely correct, the nvidia-settings command line to unconditionally assign FSAA should be: nvidia-settings --assign FSAA=5 --assign FSAAAppControlled=0 --assign FSAAAppEnhanced=0 The driver may not be able to support a particular FSAA mode for a given application due to video or system memory limitations. In that case, the driver will silently fall back to a less demanding FSAA mode. 11B. FAST APPROXIMATE ANTIALIASING (FXAA) Fast approximate antialiasing is an antialiasing mode supported by the NVIDIA graphics driver that offers advantages over traditional multisampling and supersampling methods. This mode is incompatible with UBB, triple buffering, and other antialiasing methods. To enable this mode, run: nvidia-settings --assign FXAA=1 nvidia-settings will automatically disable incompatible features when this command is run. Users may wish to disable use of FXAA for individual applications when FXAA is globally enabled. This can be done by setting the environment variable __GL_ALLOW_FXAA_USAGE to 0. __GL_ALLOW_FXAA_USAGE has no effect when FXAA is globally disabled. 11C. ANISOTROPIC TEXTURE FILTERING Automatic anisotropic texture filtering can be enabled by setting the environment variable __GL_LOG_MAX_ANISO. The possible values are: __GL_LOG_MAX_ANISO Filtering Type ---------------------------------- ---------------------------------- 0 No anisotropic filtering 1 2x anisotropic filtering 2 4x anisotropic filtering 3 8x anisotropic filtering 4 16x anisotropic filtering 11D. VBLANK SYNCING The __GL_SYNC_TO_VBLANK (boolean) environment variable can be used to control whether swaps are synchronized to a display device's vertical refresh. o Setting __GL_SYNC_TO_VBLANK=0 allows glXSwapBuffers to swap without waiting for vblank. o Setting __GL_SYNC_TO_VBLANK=1 forces glXSwapBuffers to synchronize with the vertical blanking period. This is the default behavior. When sync to vblank is enabled with TwinView, OpenGL can only sync to one of the display devices; this may cause tearing corruption on the display device to which OpenGL is not syncing. You can use the environment variable __GL_SYNC_DISPLAY_DEVICE to specify which display device should be synchronized to for OpenGL or Vulkan. You should set this environment variable to the name of a display device; for example "CRT-1". Look for the line "Connected display device(s):" in your X log file for a list of the display devices present and their names. You may also find it useful to review Chapter 12 "Configuring Twinview" and the section on Ensuring Identical Mode Timings in Chapter 18. This environment variable is only applicable to Vulkan while using either VK_PRESENT_MODE_FIFO_KHR or VK_PRESENT_MODE_FIFO_RELAXED_KHR. If a display device is being provided by a synchronized RandR 1.4 Output Sink, it will not be listed under "Connected display device(s):", but can still be used with __GL_SYNC_DISPLAY_DEVICE. The names of these display devices can be found using the "xrandr" command line tool. See Synchronized RandR 1.4 Outputs for information on synchronized RandR 1.4 Output Sinks. 11E. CONTROLLING THE SORTING OF OPENGL FBCONFIGS The NVIDIA GLX implementation sorts FBConfigs returned by glXChooseFBConfig() as described in the GLX specification. To disable this behavior set __GL_SORT_FBCONFIGS to 0 (zero), then FBConfigs will be returned in the order they were received from the X server. To examine the order in which FBConfigs are returned by the X server run: nvidia-settings --glxinfo This option may be be useful to work around problems in which applications pick an unexpected FBConfig. 11F. OPENGL YIELD BEHAVIOR There are several cases where the NVIDIA OpenGL driver needs to wait for external state to change before continuing. To avoid consuming too much CPU time in these cases, the driver will sometimes yield so the kernel can schedule other processes to run while the driver waits. For example, when waiting for free space in a command buffer, if the free space has not become available after a certain number of iterations, the driver will yield before it continues to loop. By default, the driver calls sched_yield() to do this. However, this can cause the calling process to be scheduled out for a relatively long period of time if there are other, same-priority processes competing for time on the CPU. One example of this is when an OpenGL-based composite manager is moving and repainting a window and the X server is trying to update the window as it moves, which are both CPU-intensive operations. You can use the __GL_YIELD environment variable to work around these scheduling problems. This variable allows the user to specify what the driver should do when it wants to yield. The possible values are: __GL_YIELD Behavior --------------- ------------------------------------------------------ By default, OpenGL will call sched_yield() to yield. "NOTHING" OpenGL will never yield. "USLEEP" OpenGL will call usleep(0) to yield. 11G. CONTROLLING WHICH OPENGL FBCONFIGS ARE AVAILABLE The NVIDIA GLX implementation will hide FBConfigs that are associated with a 32-bit ARGB visual when the XLIB_SKIP_ARGB_VISUALS environment variable is defined. This matches the behavior of libX11, which will hide those visuals from XGetVisualInfo and XMatchVisualInfo. This environment variable is useful when applications are confused by the presence of these FBConfigs. 11H. USING UNOFFICIAL GLX PROTOCOL By default, the NVIDIA GLX implementation will not expose GLX protocol for GL commands if the protocol is not considered complete. Protocol could be considered incomplete for a number of reasons. The implementation could still be under development and contain known bugs, or the protocol specification itself could be under development or going through review. If users would like to test the client-side portion of such protocol when using indirect rendering, they can set the __GL_ALLOW_UNOFFICIAL_PROTOCOL environment variable to a non-zero value before starting their GLX application. When an NVIDIA GLX server is used, the related X Config option "AllowUnofficialGLXProtocol" will need to be set as well to enable support in the server. 11I. OVERRIDING DRIVER DETECTION OF SELINUX POLICY BOOLEANS On Linux, the NVIDIA GLX implementation will attempt to detect whether SELinux is enabled and modify its behavior to respect SELinux policy. By default, the driver adheres to SELinux policy boolean settings at the beginning of a client process's execution; due to shared library limitations, these settings remain fixed throughout the lifetime of the driver instance. Additionally, the driver will adhere to policy boolean settings regardless of whether SELinux is running in permissive mode or enforcing mode. The __GL_SELINUX_BOOLEANS environment variable allows the user to override driver detection of specified SELinux booleans so the driver acts as if these booleans were set or unset. This allows the user, for example, to run the driver under a more restrictive policy than specified by SELinux, or to work around problems when running the driver under SELinux while operating in permissive mode. __GL_SELINUX_BOOLEANS should be set to a comma-separated list of key/value pairs: __GL_SELINUX_BOOLEANS="key1=val1,key2=val2,key3=val3,..." Valid keys are any SELinux booleans specified by "getsebool -a", and valid values are 1, true, yes, or on to enable the boolean, and 0, false, no, or off to disable it. There should be no whitespace between any key, value, or delimiter. If this environment variable is set, the driver assumes that SELinux is enabled on the system. Currently, the driver only uses the "allow_execmem" and "deny_execmem" booleans to determine whether it can apply optimizations that use writable, executable memory. Users can explicitly request that these optimizations be turned off by using the __GL_WRITE_TEXT_SECTION environment variable (see "Disabling executable memory optimizations" below). By default, if the driver cannot detect the value of one or both of these booleans, it assumes the most permissive setting (i.e. executable memory is allowed). 11J. LIMITING HEAP ALLOCATIONS IN THE OPENGL DRIVER The NVIDIA OpenGL implementation normally does not enforce limits on dynamic system memory allocations (i.e., memory allocated by the driver from the C library via the malloc(3) memory allocation package). The __GL_HEAP_ALLOC_LIMIT environment variable enables the user to specify a per-process heap allocation limit for as long as libGL is loaded in the application. __GL_HEAP_ALLOC_LIMIT is specified in the form BYTES SUFFIX, where BYTES is a nonnegative integer and SUFFIX is an optional multiplicative suffix: kB = 1000, k = 1024, MB = 1000*1000, M = 1024*1024, GB = 1000*1000*1000, and G = 1024*1024*1024. SUFFIX is not case-sensitive. For example, to specify a heap allocation limit of 20 megabytes: __GL_HEAP_ALLOC_LIMIT="20 MB" If SUFFIX is not specified, the limit is assumed to be given in bytes. The minimum heap allocation limit is 12 MB. If a lower limit is specified, the limit is clamped to the minimum. The GNU C library provides several hooks that may be used by applications to modify the behavior of malloc(3), realloc(3), and free(3). In addition, an application or library may specify allocation symbols that the driver will use in place of those exported by libc. Heap allocation tracking is incompatible with these features, and the driver will disable the heap allocation limit if it detects that they are in use. WARNING: Enforcing a limit on heap allocations may cause unintended behavior and lead to application crashes, data corruption, and system instability. ENABLE AT YOUR OWN RISK. 11K. OPENGL SHADER DISK CACHE The NVIDIA OpenGL driver utilizes a shader disk cache. This optimization benefits some applications, by reusing shader binaries instead of compiling them repeatedly. The related environment variables __GL_SHADER_DISK_CACHE, __GL_SHADER_DISK_CACHE_PATH, and __GL_SHADER_DISK_CACHE_SIZE as well as the GLShaderDiskCache X configuration option, allow fine-grained configuration of the shader cache behavior. The shader disk cache: 1. is always disabled for indirect rendering 2. is always disabled for setuid and setgid binaries 3. by default, is disabled for direct rendering when the OpenGL application is run as the root user 4. by default, is enabled for direct rendering when the OpenGL application is run as a non-root user The GLShaderDiskCache X configuration option forcibly enables or disables the shader disk cache, for direct rendering as a non-root user. The following environment variables configure shader disk cache behavior, and override the GLShaderDiskCache configuration option: Environment Variable Description ---------------------------------- ---------------------------------- __GL_SHADER_DISK_CACHE (boolean) Enables or disables the shader cache for direct rendering. __GL_SHADER_DISK_CACHE_PATH Enables configuration of where (string) shader caches are stored on disk. __GL_SHADER_DISK_CACHE_SIZE (integer) If __GL_SHADER_DISK_CACHE_PATH is unset, caches will be stored in $XDG_CACHE_HOME/nvidia/GLCache/ if XDG_CACHE_HOME is set, or in $HOME/.cache/nvidia/GLCache/ if HOME is set. If none of the environment variables __GL_SHADER_DISK_CACHE_PATH, XDG_CACHE_HOME, or HOME is set, the shader cache will be disabled. Caches are persistent across runs of an application. Cached shader binaries are specific to each driver version; changing driver versions will cause binaries to be recompiled. Note that earlier driver versions kept the default caches in $XDG_CACHE_HOME/.nv/GLCache/ or $HOME/.nv/GLCache/ instead of $XDG_CACHE_HOME/nvidia/GLCache/ or $HOME/.cache/nvidia/GLCache/, respectively. If these earlier locations exist they will be reused instead. It is safe to remove any .nv/GLCache/ directory, after which the driver will populate a fresh cache in the new locations as described in the previous paragraph. If the default cache is used and __GL_SHADER_DISK_CACHE_SIZE is set locally: any applications that subsequently run without setting it will revert the cache to the default size and cause the cache to be wiped if it now goes over. To increase the cache size globally, __GL_SHADER_DISK_CACHE_SIZE can be set in e.g. ~/.bashrc or ~/.profile. 11L. THREADED OPTIMIZATIONS The NVIDIA OpenGL driver supports offloading its CPU computation to a worker thread. These optimizations typically benefit CPU-intensive applications, but might cause a decrease of performance in applications that heavily rely on synchronous OpenGL calls such as glGet*. Because of this, they are currently disabled by default. Setting the __GL_THREADED_OPTIMIZATIONS environment variable to "1" before loading the NVIDIA OpenGL driver library will enable these optimizations for the lifetime of the application. Please note that these optimizations will only work if the target application dynamically links against pthreads. If this isn't the case, the dynamic loader can be instructed to do so at runtime by setting the LD_PRELOAD environment variable to include the pthreads library. Additionally, these optimizations require Xlib to function in thread-safe mode. The NVIDIA OpenGL driver cannot reliably enable Xlib thread-safe mode itself, therefore the application needs to call XInitThreads() before making any other Xlib call. Otherwise, the threaded optimizations in the NVIDIA driver will not be enabled. 11M. CONFORMANT GLBLITFRAMEBUFFER() SCISSOR TEST BEHAVIOR This option enables the glBlitFramebuffer() scissor test, which must be enabled for glBlitFramebuffer() to behave in a conformant manner. Setting the __GL_ConformantBlitFramebufferScissor environment variable to 0 disables the glBlitFramebuffer() scissor test, and setting it to 1 enables it. By default, the glBlitFramebuffer() scissor test is enabled. Some applications have bugs which cause them to not display properly with a conformant glBlitFramebuffer(). See Chapter 9 for more details. 11N. G-SYNC When a G-SYNC or G-SYNC Compatible monitor is attached, and G-SYNC or G-SYNC Compatible mode was allowed when the mode was set, this option controls whether these "variable refresh rate", or VRR, features can be used. Setting the __GL_VRR_ALLOWED environment variable to 0 disables VRR flipping for the application on any G-SYNC or G-SYNC Compatible monitors. Setting the __GL_VRR_ALLOWED environment variable to 1 has no effect on non-VRR monitors, or monitors that disallowed G-SYNC or G-SYNC Compatible mode when the mode was set. When G-SYNC is active on a G-SYNC or G-SYNC Compatible display and __GL_SYNC_TO_VBLANK is disabled, applications rendering faster than the maximum refresh rate will tear. This eliminates tearing for frame rates below the monitor's maximum refresh rate while minimizing latency for frame rates above it. When __GL_SYNC_TO_VBLANK is enabled, the frame rate is limited to the monitor's maximum refresh rate to eliminate tearing completely. When a G-SYNC Compatible display is in use, applications rendering slower than the minimum refresh rate may tear when __GL_SYNC_TO_VBLANK is disabled, and their swaps may not complete until the next vblank when __GL_SYNC_TO_VBLANK is enabled. G-SYNC and G-SYNC Compatible mode cannot be used when workstation stereo or workstation overlays are enabled, or when there is more than one X screen. 11O. DISABLING EXECUTABLE MEMORY OPTIMIZATIONS By default, the NVIDIA driver will attempt to use optimizations which rely on being able to write to executable memory. This may cause problems in certain system configurations (e.g., on SELinux when the "allow_execmem" boolean is disabled or "deny_execmem" boolean is enabled, and on grsecurity kernels configured with CONFIG_PAX_MPROTECT). When possible, the driver will attempt to detect when it is running on an unsupported configuration and disable these optimizations automatically. If the __GL_WRITE_TEXT_SECTION environment variable is set to 0, the driver will unconditionally disable these optimizations. 11P. IGNORING GLSL (OPENGL SHADING LANGUAGE) EXTENSION CHECKS Some applications may use GLSL shaders that reference global variables defined only in an OpenGL extension without including a corresponding #extension directive in their source code. Additionally, some applications may use GLSL shaders version 150 or greater that reference global variables defined in a compatibility profile, without specifying that a compatibility profile should be used in their #version directive. Setting the __GL_IGNORE_GLSL_EXT_REQS environment variable to 1 will cause the driver to ignore this class of errors, which may allow these shaders to successfully compile. 11Q. SHOWING THE GRAPHICS API VISUAL INDICATOR The __GL_SHOW_GRAPHICS_OSD (boolean) environment variable can be used to control whether the graphics API visual indicator is rendered on top of OpenGL and Vulkan applications. This indicator displays various information such as the graphics API in use, instantaneous frame rate, whether the application is synced to vblank, and whether the application is blitting or flipping. 11R. IMAGE SHARPENING The __GL_SHARPEN_ENABLE environment variable can be used to enable image sharpening for OpenGL and Vulkan applications. Setting __GL_SHARPEN_ENABLE=1 enables image sharpening, while setting __GL_SHARPEN_ENABLE=0 (default) disables image sharpening. The amount of sharpening can be controlled by setting the __GL_SHARPEN_VALUE environment variable to a value between 0 and 100, with 0 being no sharpening, 100 being maximum sharpening, and 50 being the default. The amount of denoising done on the sharpened image can be controlled with the __GL_SHARPEN_IGNORE_FILM_GRAIN environment variable, with 0 being no denoising, 100 being maximum denoising, and 17 being the default. ______________________________________________________________________________ Chapter 12. Configuring Multiple Display Devices on One X Screen ______________________________________________________________________________ Multiple display devices (digital flat panels, CRTs, and TVs) can display the contents of a single X screen in any arbitrary configuration. Configuring multiple display devices on a single X screen has several distinct advantages over other techniques (such as Xinerama): o A single X screen is used. The NVIDIA driver conceals all information about multiple display devices from the X server; as far as X is concerned, there is only one screen. o Both display devices share one frame buffer. Thus, all the functionality present on a single display (e.g., accelerated OpenGL) is available with multiple display devices. o No additional overhead is needed to emulate having a single desktop. If you are interested in using each display device as a separate X screen, see Chapter 14. 12A. RELEVANT X CONFIGURATION OPTIONS When the NVIDIA X driver starts, by default it will enable as many display devices as are connected and as the GPU supports driving simultaneously. Most NVIDIA GPUs support driving up to four display devices simultaneously. If multiple X screens are configured on the GPU, the NVIDIA X driver will attempt to reserve display devices and GPU resources for those other X screens (honoring the "UseDisplayDevice" and "MetaModes" X configuration options of each X screen) and then allocate all remaining resources to the first X screen configured on the GPU. There are several X configuration options that influence how multiple display devices are used by an X screen: Option "MetaModes" "" Option "HorizSync" "" Option "VertRefresh" "" Option "MetaModeOrientation" "" Option "ConnectedMonitor" "" See detailed descriptions of each option below. 12B. DETAILED DESCRIPTION OF OPTIONS HorizSync VertRefresh With these options, you can specify a semicolon-separated list of frequency ranges, each optionally prepended with a display device name. In addition, if SLI Mosaic mode is enabled, a GPU specifier can be used. For example: Option "HorizSync" "CRT-0: 50-110; DFP-0: 40-70" Option "VertRefresh" "CRT-0: 60-120; GPU-0.DFP-0: 60" See Appendix C on Display Device Names for more information. These options are normally not needed: by default, the NVIDIA X driver retrieves the valid frequency ranges from the display device's EDID (see the UseEdidFreqs option). The "HorizSync" and "VertRefresh" options override any frequency ranges retrieved from the EDID. MetaModes MetaModes are "containers" that store information about what mode should be used on each display device. Multiple MetaModes list the combinations of modes and the sequence in which they should be used. In MetaMode syntax, modes within a MetaMode are comma separated, and multiple MetaModes are separated by semicolons. For example: ", ; , " Where is the name of the mode to be used on display device 0 concurrently with used on display device 1. A mode switch will then cause to be used on display device 0 and to be used on display device 1. Here is an example MetaMode: Option "MetaModes" "1280x1024,1280x1024; 1024x768,1024x768" If you want a display device to not be active for a certain MetaMode, you can use the mode name "NULL", or simply omit the mode name entirely: "1600x1200, NULL; NULL, 1024x768" or "1600x1200; , 1024x768" Optionally, mode names can be followed by offset information to control the positioning of the display devices within the virtual screen space; e.g., "1600x1200 +0+0, 1024x768 +1600+0; ..." Offset descriptions follow the conventions used in the X "-geometry" command line option; i.e., both positive and negative offsets are valid, though negative offsets are only allowed when a virtual screen size is explicitly given in the X config file. When no offsets are given for a MetaMode, the offsets will be computed following the value of the MetaModeOrientation option (see below). Note that if offsets are given for any one of the modes in a single MetaMode, then offsets will be expected for all modes within that single MetaMode; in such a case offsets will be assumed to be +0+0 when not given. When not explicitly given, the virtual screen size will be computed as the bounding box of all MetaMode bounding boxes. MetaModes with a bounding box larger than an explicitly given virtual screen size will be discarded. A MetaMode string can be further modified with a "Panning Domain" specification; e.g., "1024x768 @1600x1200, 800x600 @1600x1200" A panning domain is the area in which a display device's viewport will be panned to follow the mouse. Panning actually happens on two levels with MetaModes: first, an individual display device's viewport will be panned within its panning domain, as long as the viewport is contained by the bounding box of the MetaMode. Once the mouse leaves the bounding box of the MetaMode, the entire MetaMode (i.e., all display devices) will be panned to follow the mouse within the virtual screen, unless the "PanAllDisplays" X configuration option is disabled. Note that individual display devices' panning domains default to being clamped to the position of the display devices' viewports, thus the default behavior is just that viewports remain "locked" together and only perform the second type of panning. The most beneficial use of panning domains is probably to eliminate dead areas -- regions of the virtual screen that are inaccessible due to display devices with different resolutions. For example: "1600x1200, 1024x768" produces an inaccessible region below the 1024x768 display. Specifying a panning domain for the second display device: "1600x1200, 1024x768 @1024x1200" provides access to that dead area by allowing you to pan the 1024x768 viewport up and down in the 1024x1200 panning domain. Offsets can be used in conjunction with panning domains to position the panning domains in the virtual screen space (note that the offset describes the panning domain, and only affects the viewport in that the viewport must be contained within the panning domain). For example, the following describes two modes, each with a panning domain width of 1900 pixels, and the second display is positioned below the first: "1600x1200 @1900x1200 +0+0, 1024x768 @1900x768 +0+1200" Because it is often unclear which mode within a MetaMode will be used on each display device, mode descriptions within a MetaMode can be prepended with a display device name. For example: "CRT-0: 1600x1200, DFP-0: 1024x768" If no MetaMode string is specified, then the X driver uses the modes listed in the relevant "Display" subsection, attempting to place matching modes on each display device. Each mode of the MetaMode may also have extra attributes associated with it, specified as a comma-separated list of token=value pairs inside curly brackets. The value for each token can optionally be enclosed in parentheses, to prevent commas within the value from being interpreted as token=value pair separators. Currently, the only token that requires a parentheses-enclosed value is "Transform". The possible tokens within the curly bracket list are: o "Stereo": possible values are "PassiveLeft" or "PassiveRight". When used in conjunction with stereo mode "4", this allows each display to be configured independently to show any stereo eye. For example: "CRT-0: 1600x1200 +0+0 { Stereo = PassiveLeft }, CRT-1: 1600x1200 +1600+0 { Stereo=PassiveRight }" If the X screen is not configured for stereo mode "4", these options are ignored. See the Stereo X configuration option for more details about stereo configurations. o "Rotation": this rotates the content of an individual display device. Possible values are "0" (with synonyms "no", "off" and "normal"), "90" (with synonyms "left" and "CCW"), "180" (with synonyms "invert" and "inverted") and "270" (with synonyms "right" and "CW"). For example: "DFP-0: nvidia-auto-select { Rotation=left }, DFP-1: nvidia-auto-select { Rotation=right }" Independent rotation configurability of each display device is also possible through RandR. See Chapter 15 for details. o "Reflection": this reflects the content of an individual display device about either the X axis, the Y axis, or both the X and Y axes. Possible values are "X", "Y" and "XY". For example: "DFP-0: nvidia-auto-select { Reflection=X }, DFP-1: nvidia-auto-select" Independent reflection configurability of each display device is also possible through RandR. See Chapter 15 for details. o "Transform": this is a 3x3 matrix of floating point values that defines a transformation from the ViewPortOut for a display device to a region within the X screen. This is equivalent to the transformation matrix specified through the RandR 1.3 RRSetCrtcTransform request. As in RandR, the transform is applied before any specified rotation and reflection values to compute the complete transform. The 3x3 matrix is represented in the MetaMode syntax as a comma-separated list of nine floating point values, stored in row-major order. This is the same as the value passed to the xrandr(1) '--transform' command line option. Note that the transform value must be enclosed in parentheses, so that the commas separating the nine floating point values are interpreted correctly. For example: "DFP-0: nvidia-auto-select { Transform=(43.864288330078125, 21.333328247070312, -16384, 0, 43.864288330078125, 0, 0, 0.0321197509765625, 19.190628051757812) }" o "PixelShiftMode": This allows a display to be configured in pixel shift mode, in which a display overlays multiple downscaled images to simulate a higher effective resolution. This is used in certain JVC e-shift projectors. All pixel shift modes require an NVIDIA RTX/Quadro GPU. Possible values are "4kTopLeft", "4kBottomRight", and "8k". In 4K pixel shift mode, two cloned displays are configured in pixel shift mode, and either display is configured to display either the top left or bottom right pixels of every pixel quad. Note that the mode timings used by each display are one quarter of the resolution read from the X screen and one quarter of the effective resolution displayed (e.g., "1920x1080" rather than "3840x2160"). For example, here is the configuration of a 4K pixel shift mode, with an effective desktop resolution of 3840x2160: "DFP-0: 1920x1080 +0+0 { PixelShiftMode = 4kTopLeft, ViewPortIn = 3840x2160 }, DFP-1: 1920x1080 +0+0 { PixelShiftMode = 4kBottomRight, ViewPortIn = 3840x2160 }" In 8K pixel shift mode, the image is downscaled from the ViewPortIn resolution to the mode timing resolution, to produce two different images: one for the top left pixel of every pixel quad and one for the bottom right of every pixel quad. The display alternates between the two images each vblank. This requires an NVIDIA RTX/Quadro graphics card with a 3-pin DIN stereo connector. For example, here is the configuration for an 8K pixel shift mode, with an effective desktop resolution and refresh rate of 8192x4800 @30Hz, split across 4 1024x2400@60Hz displays. Note that the panning offsets of each display are in X screen (ViewPortIn) coordinates: "DFP-0: 1024x2400 +0+0 { PixelShiftMode=8k, ViewPortIn = 2048x4800 }, DFP-1: 1024x2400 +2048+0 { PixelShiftMode=8k, ViewPortIn = 2048x4800 }, DFP-2: 1024x2400 +4096+0 { PixelShiftMode=8k, ViewPortIn = 2048x4800 }, DFP-4: 1024x2400 +6144+0 { PixelShiftMode=8k, ViewPortIn = 2048x4800 }" In both examples above, the ViewPortIn is provided here for illustrative purposes only. When PixelShiftMode is used, the ViewPortIn and ViewPortOut are always inferred from the mode timings: the ViewPortOut will match the mode timing resolution, which is half the intended resolution. The ViewPortIn will be twice the ViewPortOut, in order to achieve the pixel shift effect. o "ViewPortOut": this specifies the region within the mode sent to the display device that will display pixels from the X screen. The region of the mode outside the ViewPortOut will contain black. The format is "WIDTH x HEIGHT +X +Y". This is useful, for example, for configuring overscan compensation. E.g., if the mode sent to the display device is 1920x1080, to configure a 10 pixel border on all four sides: "DFP-0: 1920x1080 { ViewPortOut=1900x1060+10+10 }" Or, to only display an image in the lower right quarter of the 1920x1080 mode: "DFP-0: 1920x1080 { ViewPortOut=960x540+960+540 }" When not specified, the ViewPortOut defaults to the size of the mode. o "ViewPortIn": this defines the size of the region of the X screen which will be displayed within the ViewPortOut. The format is "WIDTH x HEIGHT". ViewPortIn is useful for configuring scaling between the X screen and the display device. For example, to display an 800x600 region from the X screen on a 1920x1200 mode: "DFP-0: 1920x1200 { ViewPortIn=800x600 }" Or, to display a 2560x1600 region from the X screen on a 1920x1200 mode: "DFP-0: 1920x1200 { ViewPortIn=2560x1600 }" Or, in conjunction with ViewPortOut, to scale an 800x600 region of the X screen within a 1920x1200 mode while preserving the aspect ratio: "DFP-0: 1920x1200 { ViewPortIn=800x600, ViewPortOut=1600x1200+160+0 }" Scaling from ViewPortIn to ViewPortOut is also expressible through the "Transform" attribute. In fact, ViewPortIn is just a shortcut for populating the transformation matrix. If both ViewPortIn and Transform are specified in the MetaMode for a display device, ViewPortIn is ignored. ViewPortIn is also ignored if PixelShiftMode is enabled, as PixelShiftMode implies a transformation of double width and height. o "PanningTrackingArea": this defines the region of the MetaMode inside which cursor movement will influence panning of the display device. The format is "WIDTH x HEIGHT + X + Y", to describe the size and offset of the region within the X screen. E.g., "DFP-0: 1920x1200 +0+0 { PanningTrackingArea = 1920x1200 +0+0 }" If not specified in the MetaMode, this will default to the entire X screen. If the "PanAllDisplays" X configuration option is explicitly set to False, then PanningTrackingArea will default to the panning domain of the display device. This is equivalent to the panning tracking_area region in the RRSetPanning RandR 1.3 protocol request. o "PanningBorder": this defines the distances from the edges of the ViewPortIn that will activate panning if the pointer hits them. If the borders are 0, the display device will pan when the pointer hits the edge of the ViewPortIn (the default). If the borders are positive, the display device will pan when the pointer gets close to the edge of the ViewPortIn. If the borders are negative, the display device will pan when the pointer is beyond the edge of the ViewPortIn. The format is "LeftBorder/TopBorder/RightBorder/BottomBorder". E.g., "DFP-0: 1920x1200 +0+0 { PanningBorder = 10/10/10/10 }" This is equivalent to the panning border in the RRSetPanning RandR 1.3 protocol request. o "ForceCompositionPipeline": possible values are "On" or "Off". The NVIDIA X driver can use a composition pipeline to apply X screen transformations and rotations. "ForceCompositionPipeline" can be used to force the use of this pipeline, even when no transformations or rotations are applied to the screen. o "ForceFullCompositionPipeline": possible values are "On" or "Off". This option implicitly enables "ForceCompositionPipeline" and additionally makes use of the composition pipeline to apply ViewPortOut scaling. o "WarpMesh", "BlendTexture", "OffsetTexture": these string attributes control the operation of Warp and Blend, an advanced transformation feature available on select NVIDIA RTX/Quadro GPUs. Warp and Blend can adjust a display's geometry (warp) with a greater level of control than a simple matrix transformation: for example, to facilitate projecting an image onto a non-planar surface, and its intensity (blend) per pixel: for example, to seamlessly combine images from multiple overlapping projectors into a single large image. Each of the "WarpMesh", "BlendTexture", and "OffsetTexture" MetaMode tokens can be set to the name of a pixmap which has already been bound to a name via the XNVCtrlBindWarpPixmapName() NV_CONTROL call. See the 'nv-control-warpblend' sample application distributed with the 'nvidia-settings' source code for a more detailed description of the functionality of each of these pixmaps, how to lay data out into the pixmaps, and an example implementation of an X application that loads and binds pixmaps so that they are ready to use in a MetaMode. If driving the current mode in the RGB 4:4:4 color space would require a pixel clock that exceeds the display's or GPU's capabilities, and the display and GPU are capable of driving that mode in the YCbCr 4:2:0 color space, then the color space will be overridden to YCbCr 4:2:0. In this case, if a Turing or earlier GPU is in use, or if a DisplayPort display device is in use regardless of GPU, then Warp and Blend will be disabled. Warp and Blend is not yet supported with the GLX_NV_swap_group OpenGL extension. o "BlendOrder": Controls the order of warping and blending when using Warp and Blend. By default, warping is performed first, followed by blending; setting the "BlendOrder" MetaMode token to "BlendAfterWarp" will reverse the default order. o "ResamplingMethod": Controls the filtering method used to smooth the display image when scaling screen transformations (such as a WarpMesh or scaling ViewPortOut) are in use. Possible values are "Bilinear" (default), "BicubicTriangular", "BicubicBellShaped", "BicubicBspline", "BicubicAdaptiveTriangular", "BicubicAdaptiveBellShaped", "BicubicAdaptiveBspline", and "Nearest". Bicubic resampling is only available on NVIDIA RTX/Quadro GPUs. If a mode requiring a YCbCr 4:2:0 color space is in use on a DisplayPort display device or a Turing or earlier GPU, then bicubic resampling is unavailable. Bicubic resampling is not supported with the GLX_NV_swap_group OpenGL extension. o "AllowGSYNC" or "AllowGSYNCCompatible": Controls whether capable monitors are put into G-SYNC or G-SYNC Compatible mode or left in continuous refresh mode. By default, G-SYNC monitors are put into G-SYNC mode when a mode is set, and transition into and out of variable refresh mode seamlessly. However, this prevents certain other features, such as Ultra Low Motion Blur and Frame Lock, from working. Set this to "Off" to use continuous refresh rates that are compatible with these features. In addition, G-SYNC Compatible monitors are supported via DisplayPort on Pascal and newer GPUs, and via HDMI on Turing and newer GPUs. Supported monitors will enable variable refresh mode by default. This behavior may be overridden by setting the AllowGSYNCCompatible=Off MetaMode attribute. Pascal GPUs only support G-SYNC Compatible monitors in systems containing no more than one monitor in variable refresh mode. Monitors that have not been validated as G-SYNC Compatible have G-SYNC Compatible mode disabled by default, and can be forced into G-SYNC Compatible mode by setting the AllowGSYNCCompatible=On MetaMode attribute. A list of supported G-SYNC and G-SYNC Compatible monitors can be found at https://www.nvidia.com/en-us/geforce/products/g-sync-monitors/specs/ Note: Vulkan direct-to-display applications may allow or disallow G-SYNC or G-SYNC Compatible at modeset time using the VKDirectGSYNCAllowed and VKDirectGSYNCCompatibleAllowed environment variables. VKDirectGSYNCAllowed is set to true by default (allowing VRR on all G-SYNC monitors) and VKDirectGSYNCCompatibleAllowed may be set to 0 (disallowing VRR on all G-SYNC Compatible monitors), 1 (default, allowing VRR only on monitors validated as G-SYNC Compatible), or 2 (allowing VRR regardless of whether a monitor has been validated as G-SYNC Compatible). o "VRRMinRefreshRate": Overrides a G-SYNC Compatible monitor's minimum refresh rate. Some monitors that have not been validated as G-SYNC Compatible may flicker when applications are swapping close to the monitor's minimum refresh rate, and this MetaMode flag may be used to compensate for this. This MetaMode flag has no effect on G-SYNC monitors. Note that the current MetaMode can also be configured through the NV-CONTROL X extension and the nvidia-settings utility. For example: nvidia-settings --assign CurrentMetaMode="DFP-0: 1920x1200 { ViewPortIn=800x600, ViewPortOut=1600x1200+160+0 }" MetaModeOrientation This option controls the positioning of the display devices within the virtual X screen, when offsets are not explicitly given in the MetaModes. The possible values are: "RightOf" (the default) "LeftOf" "Above" "Below" "SamePositionAs" When "SamePositionAs" is specified, all display devices will be assigned an offset of 0,0. For backwards compatibility, "Clone" is a synonym for "SamePositionAs". Because it is often unclear which display device relates to which, MetaModeOrientation can be confusing. You can further clarify the MetaModeOrientation with display device names to indicate which display device is positioned relative to which display device. For example: "CRT-0 LeftOf DFP-0" ConnectedMonitor With this option you can override what the NVIDIA kernel module detects is connected to your graphics card. This may be useful, for example, if any of your display devices do not support detection using Display Data Channel (DDC) protocols. Valid values are a comma-separated list of display device names; for example: "CRT-0, CRT-1" "CRT" "CRT-1, DFP-0" WARNING: this option overrides what display devices are detected by the NVIDIA kernel module, and is very seldom needed. You really only need this if a display device is not detected, either because it does not provide DDC information, or because it is on the other side of a KVM (Keyboard-Video-Mouse) switch. In most other cases, it is best not to specify this option. Just as in all X config entries, spaces are ignored and all entries are case insensitive. FREQUENTLY ASKED TWINVIEW QUESTIONS Q. Nothing gets displayed on my second monitor; what is wrong? A. Monitors that do not support monitor detection using Display Data Channel (DDC) protocols (this includes most older monitors) are not detectable by your NVIDIA card. You need to explicitly tell the NVIDIA X driver what you have connected using the "ConnectedMonitor" option; e.g., Option "ConnectedMonitor" "CRT, CRT" Q. Will window managers be able to appropriately place windows (e.g., avoiding placing windows across both display devices, or in inaccessible regions of the virtual desktop)? A. Yes. Window managers can query the layout of display devices through either RandR 1.2 or Xinerama. The NVIDIA X driver provides a Xinerama extension that X clients (such as window managers) can use to discover the current layout of display devices. Note that the Xinerama protocol provides no way to notify clients when a configuration change occurs, so if you modeswitch to a different MetaMode, your window manager may still think you have the previous configuration. Using RandR 1.2, or the Xinerama extension in conjunction with the XF86VidMode extension to get modeswitch events, window managers should be able to determine the display device configuration at any given time. Unfortunately, the data provided by XineramaQueryScreens() appears to confuse some window managers; to work around such broken window managers, you can disable communication of the display device layout with the nvidiaXineramaInfo X configuration option. The order that display devices are reported in via the NVIDIA Xinerama information can be configured with the nvidiaXineramaInfoOrder X configuration option. Be aware that the NVIDIA driver cannot provide the Xinerama extension if the X server's own Xinerama extension is being used. Explicitly specifying Xinerama in the X config file or on the X server commandline will prohibit NVIDIA's Xinerama extension from installing, so make sure that the X server's log file does not contain: (++) Xinerama: enabled if you want the NVIDIA driver to be able to provide the Xinerama extension while in TwinView. Another solution is to use panning domains to eliminate inaccessible regions of the virtual screen (see the MetaMode description above). A third solution is to use two separate X screens, rather than use TwinView. See Chapter 14. Q. How are virtual screen dimensions determined in TwinView? A. After all requested modes have been validated, and the offsets for each MetaMode's viewports have been computed, the NVIDIA driver computes the bounding box of the panning domains for each MetaMode. The maximum bounding box width and height is then found. Note that one side effect of this is that the virtual width and virtual height may come from different MetaModes. Given the following MetaMode string: "1600x1200,NULL; 1024x768+0+0, 1024x768+0+768" the resulting virtual screen size will be 1600 x 1536. Q. Can I play full screen games across both display devices? A. Yes. While the details of configuration will vary from game to game, the basic idea is that a MetaMode presents X with a mode whose resolution is the bounding box of the viewports for that MetaMode. For example, the following: Option "MetaModes" "1024x768,1024x768; 800x600,800x600" Option "MetaModeOrientation" "RightOf" produce two modes: one whose resolution is 2048x768, and another whose resolution is 1600x600. Games such as Quake 3 Arena use the VidMode extension to discover the resolutions of the modes currently available. To configure Quake 3 Arena to use the above MetaMode string, add the following to your q3config.cfg file: seta r_customaspect "1" seta r_customheight "600" seta r_customwidth "1600" seta r_fullscreen "1" seta r_mode "-1" Note that, given the above configuration, there is no mode with a resolution of 800x600 (remember that the MetaMode "800x600, 800x600" has a resolution of 1600x600"), so if you change Quake 3 Arena to use a resolution of 800x600, it will display in the lower left corner of your screen, with the rest of the screen grayed out. To have single head modes available as well, an appropriate MetaMode string might be something like: "800x600,800x600; 1024x768,NULL; 800x600,NULL; 640x480,NULL" More precise configuration information for specific games is beyond the scope of this document, but the above examples coupled with numerous online sources should be enough to point you in the right direction. ______________________________________________________________________________ Chapter 13. Configuring GLX in Xinerama ______________________________________________________________________________ The NVIDIA Linux Driver supports GLX when Xinerama is enabled on similar GPUs. The Xinerama extension takes multiple physical X screens (possibly spanning multiple GPUs), and binds them into one logical X screen. This allows windows to be dragged between GPUs and to span across multiple GPUs. The NVIDIA driver supports hardware accelerated OpenGL rendering across all NVIDIA GPUs when Xinerama is enabled. To configure Xinerama 1. Configure multiple X screens (refer to the xorg.conf man page for details). 2. Enable Xinerama by adding the line Option "Xinerama" "True" to the "ServerFlags" section of your X config file. Requirements: o Using identical GPUs is recommended. Some combinations of non-identical, but similar, GPUs are supported. If a GPU is incompatible with the rest of a Xinerama desktop then no OpenGL rendering will appear on the screens driven by that GPU. Rendering will still appear normally on screens connected to other supported GPUs. In this situation the X log file will include a message of the form: (WW) NVIDIA(2): The GPU driving screen 2 is incompatible with the rest of (WW) NVIDIA(2): the GPUs composing the desktop. OpenGL rendering will (WW) NVIDIA(2): be disabled on screen 2. o NVIDIA's GLX implementation only supports Xinerama when physical X screen 0 is driven by the NVIDIA X driver. This is because the X.Org X server bases the visuals of the logical Xinerama X screen on the visuals of physical X screen 0. When physical X screen 0 is not being driven by the NVIDIA X driver and Xinerama is enabled, then GLX will be disabled. If physical X screens other than screen 0 are not being driven by the NVIDIA X driver, OpenGL rendering will be disabled on them. o Only the intersection of capabilities across all GPUs will be advertised. The maximum OpenGL viewport size is 16384x16384 pixels. If an OpenGL window is larger than the maximum viewport, regions beyond the viewport will be blank. o X configuration options that affect GLX operation (e.g.: stereo, overlays) should be set consistently across all X screens in the X server. ______________________________________________________________________________ Chapter 14. Configuring Multiple X Screens on One Card ______________________________________________________________________________ GPUs that support TwinView (Chapter 12) can also be configured to treat each connected display device as a separate X screen. While there are several disadvantages to this approach as compared to TwinView (e.g.: windows cannot be dragged between X screens, hardware accelerated OpenGL cannot span the two X screens), it does offer one advantage over TwinView: If each display device is a separate X screen, then properties that may vary between X screens may vary between displays (e.g.: depth, root window size, etc). To configure two separate X screens to share one graphics card, here is what you will need to do: First, create two separate Device sections, each listing the BusID of the graphics card to be shared and listing the driver as "nvidia", and assign each a separate screen: Section "Device" Identifier "nvidia0" Driver "nvidia" # Edit the BusID with the location of your graphics card BusID "PCI:2:0:0" Screen 0 EndSection Section "Device" Identifier "nvidia1" Driver "nvidia" # Edit the BusID with the location of your graphics card BusId "PCI:2:0:0" Screen 1 EndSection Then, create two Screen sections, each using one of the Device sections: Section "Screen" Identifier "Screen0" Device "nvidia0" Monitor "Monitor0" DefaultDepth 24 Subsection "Display" Depth 24 Modes "1600x1200" "1024x768" "800x600" "640x480" EndSubsection EndSection Section "Screen" Identifier "Screen1" Device "nvidia1" Monitor "Monitor1" DefaultDepth 24 Subsection "Display" Depth 24 Modes "1600x1200" "1024x768" "800x600" "640x480" EndSubsection EndSection (Note: You'll also need to create a second Monitor section) Finally, update the ServerLayout section to use and position both Screen sections: Section "ServerLayout" ... Screen 0 "Screen0" Screen 1 "Screen1" leftOf "Screen0" ... EndSection For further details, refer to the xorg.conf man page. ______________________________________________________________________________ Chapter 15. Support for the X Resize and Rotate Extension ______________________________________________________________________________ This NVIDIA driver release contains support for the X Resize and Rotate (RandR) Extension versions 1.1, 1.2, and 1.3. The version of the RandR extension advertised to X clients is controlled by the X server: the RandR extension and protocol are provided by the X server, which routes protocol requests to the NVIDIA X driver. Run `xrandr --version` to check the version of RandR provided by the X server. 15A. RANDR SUPPORT Specific supported features include: o Modes can be set per-screen, and the X screen can be resized through the RRSetScreenConfig request (e.g., with xrandr(1)'s '--size' and '--rate' command line options). o The X screen can be resized with the RandR 1.2 RRSetScreenSize request (e.g., with xrandr(1)'s '--fb' command line option). o The state of the display hardware can be queried with the RandR 1.2 and 1.3 RRGetScreenResources, RRGetScreenResourcesCurrent, RRGetOutputInfo, and RRGetCrtcInfo requests (e.g., with xrandr(1)'s '--query' command line option). o Modes can be set with RandR CRTC granularity with the RandR 1.2 RRSetCrtcConfig request. E.g., `xrandr --output DVI-I-2 --mode 1920x1200`. o Rotation can be set with RandR CRTC granularity with the RandR 1.2 RRSetCrtcConfig request. E.g., `xrandr --output DVI-I-2 --mode 1920x1200 --rotation left`. o Per-CRTC transformations can be manipulated with the RandR 1.3 RRSetCrtcTransform and RRGetCrtcTransform requests. E.g., `xrandr --output DVI-I-3 --mode 1920x1200 --transform 43.864288330078125,21.333328247070312,-16384,0,43.864288330078125,0,0,0.0321197509765625,19.190628051757812`. o The RandR 1.0/1.1 requests RRGetScreenInfo and RRSetScreenConfig manipulate MetaModes. The MetaModes that the X driver uses (either user-requested or implicitly generated) are reported through the RRGetScreenInfo request (e.g., `xrandr --query --q1`) and chosen through the RRSetScreenConfig request (e.g., `xrandr --size 1920x1200 --orientation left`). The configurability exposed through RandR is also available through the MetaMode syntax, independent of X server version. See Chapter 12 for more details. As an example, these two commands are equivalent: xrandr --output DVI-I-2 --mode 1280x1024 --pos 0x0 --rotate left \ --output DVI-I-3 --mode 1920x1200 --pos 0x0 nvidia-settings --assign CurrentMetaMode="DVI-I-2: 1280x1024 +0+0 \ { Rotation=left }, DVI-I-3: 1920x1200 +0+0" 15B. RANDR 1.1 ROTATION BEHAVIOR On X servers that support RandR 1.2 or later, when an RandR 1.1 rotation request is received (e.g., `xrandr --orientation left`), the NVIDIA X driver will apply that request to an entire MetaMode. E.g., if you configure multiple monitors, either through a MetaMode or through RandR 1.2: xrandr --output DVI-I-2 --mode 1280x1024 --pos 0x0 \ --output DVI-I-3 --mode 1920x1200 --pos 1280x0 nvidia-settings --assign CurrentMetaMode="DVI-I-2: 1280x1024 +0+0, \ DVI-I-3: 1920x1200 +1280+0" Requesting RandR 1.1 rotation through `xrandr --orientation left`, will rotate the entire MetaMode, producing the equivalent of either: xrandr --output DVI-I-2 --mode 1280x1024 --pos 176x0 --rotate left \ --output DVI-I-3 --mode 1920x1200 --rotate left --pos 0x1280 nvidia-settings --assign CurrentMetaMode="DVI-I-2: 1280x1024 +176+0, \ { Rotation=left }, DVI-I-3: 1920x1200 +0+1280 { Rotation=left }" On X servers that do not support RandR 1.2 or later, the NVIDIA X driver does not advertise RandR rotation support. On such X servers, it is recommended to configure rotation through MetaModes, instead. 15C. OUTPUT PROPERTIES The NVIDIA Linux driver supports a number of output device properties. OFFICIAL PROPERTIES Properties that do not start with an underscore are officially documented in the file "randrproto.txt" in X.Org's randrproto package. See that file for a full description of these properties. o "ConnectorNumber" This property groups RandR outputs by their physical connectors. For example, DVI-I ports have both an analog and a digital output, which is represented in RandR by two different output objects. One DVI-I port may be represented by RandR outputs "DVI-I-0" with "SignalFormat" "TMDS" (transition-minimized differential signaling, a digital signal format) and "DVI-I-1" with "SignalFormat" "VGA", representing the analog part. In this case, both RandR outputs would have the same value of "ConnectorNumber". o "ConnectorType" This property lists the physical type of the connector. For example, in the DVI-I example above, both "DVI-I-0" and "DVI-I-1" would have a "ConnectorType" of "DVI-I". o "EDID" This property contains the raw bytes of the display's extended display identification data. This data is intended for applications to use to glean information about the monitor connected. o "SignalFormat" This property describes the type of signaling used to send image data to the display device. For example, an analog device connected to a DVI-I port might use VGA as its signaling format. o "Border" This property is a list of integers specifying adjustments for the edges of the displayed image. How this property is applied depends on the number of elements in the list: o 0 = No border is applied. o 1 = A border of Border[0] is applied to all four sides of the image. o 2 = A border of Border[0] is applied to the left and right sides of the image, and a border of Border[1] is applied to the top and bottom. o 4 = The border dimensions are as follows: Border[0]: left, Border[1]: top, Border[2]: right, Border[3]: bottom This property is functionally equivalent to the ViewPortOut MetaMode token. o "BorderDimensions" This property lists how many Border adjustment parameters can actually be used. The NVIDIA implementation supports independently configuring all four Border values. o "GUID" DisplayPort 1.2 specifies that all devices must have a globally-unique identifier, referred to as a GUID. When a GUID is available, the "GUID" property contains its raw bytes. o "CscMatrix" This property controls the color-space conversion matrix applied to the pixels being displayed. The matrix is 3 rows and 4 columns, stored in row-major order. Each entry is a 32-bit fixed-point number with 3 integer bits and 16 fractional bits. Each entry in the X colormap is treated as a 4-component column vector C = [ R, G, B, 1 ]. The resulting components of the color vector [ R', G', B' ] = CscMatrix * C are used as indices into the gamma ramp. For example, using xrandr version 1.5.0 or higher, you can exchange the red channel with the green channel using this command: xrandr --output DP-6 --set CscMatrix \ 0,0x10000,0,0,0x10000,0,0,0,0,0,0x10000,0 To return to the default identity matrix, use xrandr --output DP-6 --set CscMatrix \ 0x10000,0,0,0,0,0x10000,0,0,0,0,0x10000,0 UNOFFICIAL PROPERTIES Properties whose names begin with an underscore are not specified by X.Org. They may be removed or modified in future driver releases. The NVIDIA Linux driver supports the following unofficial properties: o "_ConnectorLocation" This property describes the physical location of the connector. On add-in graphics boards, connector location 0 should generally be the position closest to the motherboard, with increasing location numbers indicating connectors progressively farther away. Type INTEGER Format 32 # Items 1 Flags Immutable, Static Range 0- 15D. DISPLAYPORT 1.2 When display devices are connected via DisplayPort 1.2 branch devices, additional RandR outputs will be created, one for each connected display device. These dynamic outputs will remain as long as the display device is connected or used in a MetaMode, even if they are not named in the current MetaMode. They will be deleted automatically when the display is disconnected and no MetaModes use them. See Appendix C for a description of how the names of these outputs are generated. If you are developing applications that use the RandR extension, please be aware that outputs can be created and destroyed dynamically. You should be sure to use the XRRSelectInput function to watch for events that indicate when this happens. 15E. MONITOR CONFIGURATION OPTIONS This NVIDIA Linux driver honors the "Enable", "Ignore", "Primary", and "Rotate" options in the Monitor section of the X configuration file. These options will apply to a monitor if the Identifier of the Monitor section matches one of the display device's names (see Appendix C for a description of how a display's names are generated). For example, a Monitor section with Identifier "DFP" will apply to all digitally-connected displays, while a Monitor section with Identifier "DPY-EDID-ee6cecc0-fa46-0c33-94e0-274313f9e7eb" will apply only to a display device with a specific EDID-based identification hash. You can also specify the name of the Monitor section to use in the Screen section: Option "Monitor-" "" See the xorg.conf(5) man page for a description of these options. 15F. KNOWN ISSUES o Rotation and Transformations (configured either through RandR or MetaModes) are not yet supported with the GLX_NV_swap_group OpenGL extension. o Some of the RandR 1.2 X configuration options provided by the XFree86 DDX implementation and documented in xorg.conf(5) are not yet supported. o Transformations (configured either through RandR or MetaModes) are not yet correctly clipped. ______________________________________________________________________________ Chapter 16. Configuring a Notebook ______________________________________________________________________________ 16A. INSTALLATION AND CONFIGURATION Installation and configuration of the NVIDIA Linux Driver Set on a notebook is the same as for any desktop environment, with a few additions, as described below. 16B. POWER MANAGEMENT All notebook NVIDIA GPUs support power management, both S3 (also known as "Standby" or "Suspend to RAM") and S4 (also known as "Hibernate", "Suspend to Disk" or "SWSUSP"). Power management is system-specific and is dependent upon all the components in the system; some systems may be more problematic than other systems. Most recent notebook NVIDIA GPUs also support PowerMizer, which monitors application work load to adjust system parameters to deliver the optimal balance of performance and battery life. However, PowerMizer is only enabled by default on some notebooks. Please see the known issues below for more details. 16C. HOTKEY SWITCHING OF DISPLAY DEVICES Most laptops generate keyboard events when the display change hotkey is pressed. On some laptops, these are simply normal keyboard keys. On others, they generate ACPI events that may be translated into keyboard events by other system components. The NVIDIA driver does not handle ACPI display change hotkeys itself. Instead, it is expected for desktop environments to listen for these key-press events and respond by reconfiguring the display devices as necessary. 16D. DOCKING EVENTS All notebook NVIDIA GPUs support docking, however support may be limited by the OS or system. There are three types of notebook docking (hot, warm, and cold), which refer to the state of the system when the docking event occurs. hot refers to a powered on system with a live desktop, warm refers to a system that has entered a suspended power management state, and cold refers to a system that has been powered off. Only warm and cold docking are supported by the NVIDIA driver. 16E. KNOWN NOTEBOOK ISSUES There are a few known issues associated with notebooks: o In many cases, suspending and/or resuming will fail. As mentioned above, this functionality is very system-specific. There are still many cases that are problematic. Here are some tips that may help: o In some cases, hibernation can have bad interactions with the PCI Express bus clocks, which can lead to system hangs when entering hibernation. This issue is still being investigated, but a known workaround is to leave an OpenGL application running when hibernating. o On notebooks with relatively little system memory, repetitive hibernation attempts may fail due to insufficient free memory. This problem can be avoided by running `echo 0 > /sys/power/image_size`, which reduces the image size to be stored during hibernation. o Some distributions use a tool called vbetool to save and restore VGA adapter state. This tool is incompatible with NVIDIA GPUs' Video BIOSes and is likely to lead to problems restoring the GPU and its state. Disabling calls to this tool in your distribution's init scripts may improve power management reliability. o This driver release does not support RTD3 on the Blackwell GPU architecture. o On some notebooks, PowerMizer is not enabled by default. This issue is being investigated, and there is no known workaround. o When available, the NVIDIA driver installs a backlight handler that allows access to the driver's backlight controller through /sys/class/backlight/nvidia_0. This option can be disabled by passing the NVreg_EnableBacklightHandler=0 parameter to the nvidia kernel module. ______________________________________________________________________________ Chapter 17. Using the NVIDIA Driver with Optimus Laptops ______________________________________________________________________________ Some laptops with NVIDIA GPUs make use of Optimus technology to allow switching between an integrated GPU and a discrete NVIDIA GPU. The NVIDIA Linux driver can be used on these systems. 17A. INSTALLING THE NVIDIA DRIVER ON AN OPTIMUS LAPTOP The driver may be installed normally on Optimus systems, but the NVIDIA X driver and the NVIDIA OpenGL driver may not be able to display to the laptop's internal display panel unless a means to connect the panel to the NVIDIA GPU (for example, a hardware multiplexer, or "mux", often controllable by a BIOS setting) is available. On systems without a mux, the NVIDIA GPU can still be useful for offscreen rendering, PRIME render offload, running CUDA applications, and other uses that don't require driving a display. On muxless Optimus laptops, or on laptops where a mux is present, but not set to drive the internal display from the NVIDIA GPU, the internal display is driven by the integrated GPU. On these systems, it's important that the X server not be configured to use the NVIDIA X driver after the driver is installed. Instead, the correct driver for the integrated GPU should be used. Often, this can be determined automatically by the X server, and no explicit configuration is required, especially on newer X server versions. If your X server autoselects the NVIDIA X driver after installation, you may need to explicitly select the driver for your integrated GPU. As an alternative to using only the integrated graphics device, support for the display output source functionality provided by the X Resize and Rotate extension version 1.4 is available. This functionality allows for graphics to be rendered on the NVIDIA GPU and displayed on the integrated graphics device. For information on how to use this functionality, see Chapter 34. A second alternative is to use PRIME render offload, such that the integrated graphics device is used to drive the X screen, but the NVIDIA GPU is used on a per-application basis to accelerate rendering of specific applications. For details, see Chapter 35. 17B. LOADING THE KERNEL MODULE AND CREATING THE DEVICE FILES WITHOUT X In order for programs that use the NVIDIA driver to work correctly (e.g.: X, OpenGL, and CUDA applications), the kernel module must be loaded, and the device files '/dev/nvidiactl' and '/dev/nvidia[0-9]+' must exist with read and write permissions for any users of such applications. If the setuid root nvidia-modprobe(1) utility is installed (the default when the driver is installed from .run file), this should be handled automatically. Otherwise, the kernel module will need to be loaded, and the device files created, through your Linux distribution's mechanisms. See "Q. How and when are the NVIDIA device files created?" in Chapter 7 for more information. Note that on some Optimus notebooks the driver may fail to initialize the GPU due to system-specific ACPI interaction problems: see "Q. Why does the VBIOS fail to load on my Optimus system?" in Chapter 8 for more information. ______________________________________________________________________________ Chapter 18. Programming Modes ______________________________________________________________________________ The NVIDIA Accelerated Linux Graphics Driver supports all standard VGA and VESA modes, as well as most user-written custom mode lines. To request one or more standard modes for use in X, you can simply add a "Modes" line such as: Modes "1600x1200" "1024x768" "640x480" in the appropriate Display subsection of your X config file (see the xorg.conf man page for details). Or, the nvidia-xconfig(1) utility can be used to request additional modes; for example: nvidia-xconfig --mode 1600x1200 See the nvidia-xconfig(1) man page for details. 18A. DEPTH, BITS PER PIXEL, AND PITCH While not directly a concern when programming modes, the bits used per pixel is an issue when considering the maximum programmable resolution; for this reason, it is worthwhile to address the confusion surrounding the terms "depth" and "bits per pixel". Depth is how many bits of data are stored per pixel. Supported depths are 8, 15, 16, 24, and 30. Most video hardware, however, stores pixel data in sizes of 8, 16, or 32 bits; this is the amount of memory allocated per pixel. When you specify your depth, X selects the bits per pixel (bpp) size in which to store the data. Below is a table of what bpp is used for each possible depth: Depth BPP ---------------------------------- ---------------------------------- 8 8 15 16 16 16 24 32 30 32 Lastly, the "pitch" is how many bytes in the linear frame buffer there are between one pixel's data, and the data of the pixel immediately below. You can think of this as the horizontal resolution multiplied by the bytes per pixel (bits per pixel divided by 8). In practice, the pitch may be more than this product due to alignment constraints. 18B. MAXIMUM RESOLUTIONS The NVIDIA Accelerated Linux Graphics Driver and NVIDIA GPU-based graphics cards support resolutions up to 32767x32767 pixels for Pascal and newer GPUs, and up to 16384x16384 pixels for older GPUs, though the maximum resolution your system can support is also limited by the amount of video memory (see Useful Formulas for details) and the maximum supported resolution of your display device (monitor/flat panel/television). Also note that while use of a video overlay does not limit the maximum resolution or refresh rate, video memory bandwidth used by a programmed mode does affect the overlay quality. Using HDMI 2.0 4K@60Hz modes in the RGB 4:4:4 color space requires a high single-link pixel clock that is only available on GM20x or later GPUs. In addition, using a mode requiring the YCbCr 4:2:0 color space over a DisplayPort connection requires a GP10x or later GPU. If driving the current mode in the RGB 4:4:4 color space would require a pixel clock that exceeds the display's or GPU's capabilities, and the display and GPU are capable of driving that mode in the YCbCr 4:2:0 color space, then the color space will be overridden to YCbCr 4:2:0. On all Turing and earlier GPUs, and on all DisplayPort display devices regardless of GPU, YCbCr 4:2:0 mode is not supported on depth 8 X screens, and is not currently supported with stereo or the GLX_NV_swap_group OpenGL extension. 18C. USEFUL FORMULAS The maximum resolution is a function both of the amount of video memory and the bits per pixel you elect to use: HR * VR * (bpp/8) = Video Memory Used In other words, the amount of video memory used is equal to the horizontal resolution (HR) multiplied by the vertical resolution (VR) multiplied by the bytes per pixel (bits per pixel divided by eight). Technically, the video memory used is actually the pitch times the vertical resolution, and the pitch may be slightly greater than (HR * (bpp/8)) to accommodate the hardware requirement that the pitch be a multiple of some value. Note that this is just memory usage for the frame buffer; video memory is also used by other things, such as OpenGL and pixmap caching. Another important relationship is that between the resolution, the pixel clock (also known as the dot clock) and the vertical refresh rate: RR = PCLK / (HFL * VFL) In other words, the refresh rate (RR) is equal to the pixel clock (PCLK) divided by the total number of pixels: the horizontal frame length (HFL) multiplied by the vertical frame length (VFL) (note that these are the frame lengths, and not just the visible resolutions). As described in the XFree86 Video Timings HOWTO, the above formula can be rewritten as: PCLK = RR * HFL * VFL Given a maximum pixel clock, you can adjust the RR, HFL and VFL as desired, as long as the product of the three is consistent. The pixel clock is reported in the log file. Your X log should contain a line like this: (--) NVIDIA(0): ViewSonic VPD150 (DFP-1): 165 MHz maximum pixel clock which indicates the maximum pixel clock for that display device. 18D. HOW MODES ARE VALIDATED In traditional XFree86/X.Org mode validation, the X server takes as a starting point the X server's internal list of VESA standard modes, plus any modes specified with special ModeLines in the X configuration file's Monitor section. These modes are validated against criteria such as the valid HorizSync/VertRefresh frequency ranges for the user's monitor (as specified in the Monitor section of the X configuration file), as well as the maximum pixel clock of the GPU. Once the X server has determined the set of valid modes, it takes the list of user requested modes (i.e., the set of modes named in the "Modes" line in the Display subsection of the Screen section of X configuration file), and finds the "best" validated mode with the requested name. The NVIDIA X driver uses a variation on the above approach to perform mode validation. During X server initialization, the NVIDIA X driver builds a pool of valid modes for each display device. It gathers all possible modes from several sources: o The display device's EDID o The X server's built-in list o Any user-specified ModeLines in the X configuration file o The VESA standard modes For every possible mode, the mode is run through mode validation. The core of mode validation is still performed similarly to traditional XFree86/X.Org mode validation: the mode timings are checked against things such as the valid HorizSync and VertRefresh ranges and the maximum pixelclock. Note that each individual stage of mode validation can be independently controlled through the "ModeValidation" X configuration option. Note that when validating interlaced mode timings, VertRefresh specifies the field rate, rather than the frame rate. For example, the following modeline has a vertical refresh rate of 87 Hz: # 1024x768i @ 87Hz (industry standard) ModeLine "1024x768" 44.9 1024 1032 1208 1264 768 768 776 817 +hsync +vsync Interlace Invalid modes are discarded; valid modes are inserted into the mode pool. See MODE VALIDATION REPORTING for how to get more details on mode validation results for each considered mode. Valid modes are given a unique name that is guaranteed to be unique across the whole mode pool for this display device. This mode name is constructed approximately like this: x_ (e.g., "1600x1200_85") The name may also be prepended with another number to ensure the mode is unique; e.g., "1600x1200_85_0". As validated modes are inserted into the mode pool, duplicate modes are removed, and the mode pool is sorted, such that the "best" modes are at the beginning of the mode pool. The sorting is based roughly on: o Resolution o Source (EDID-provided modes are prioritized higher than VESA-provided modes, which are prioritized higher than modes that were in the X server's built-in list) o Refresh rate Once modes from all mode sources are validated and the mode pool is constructed, all modes with the same resolution are compared; the best mode with that resolution is added to the mode pool a second time, using just the resolution as its unique modename (e.g., "1600x1200"). In this way, when you request a mode using the traditional names (e.g., "1600x1200"), you still get what you got before (the 'best' 1600x1200 mode); the added benefit is that all modes in the mode pool can be addressed by a unique name. When verbose logging is enabled (see "Q. How can I increase the amount of data printed in the X log file?" in Chapter 7), the mode pool for each display device is printed to the X log file. After the mode pool is built for all display devices, the requested modes (as specified in the X configuration file), are looked up from the mode pool. Each requested mode that can be matched against a mode in the mode pool is then advertised to the X server and is available to the user through the X server's mode switching hotkeys (ctrl-alt-plus/minus) and the XRandR and XF86VidMode X extensions. Additionally, all modes in the mode pool of the primary display device are implicitly made available to the X server. See the IncludeImplicitMetaModes X configuration option for details. 18E. THE NVIDIA-AUTO-SELECT MODE You can request a special mode by name in the X config file, named "nvidia-auto-select". When the X driver builds the mode pool for a display device, it selects one of the modes as the "nvidia-auto-select" mode; a new entry is made in the mode pool, and "nvidia-auto-select" is used as the unique name for the mode. The "nvidia-auto-select" mode is intended to be a reasonable mode for the display device in question. For example, the "nvidia-auto-select" mode is normally the native resolution for flat panels, as reported by the flat panel's EDID, or one of the detailed timings from the EDID. The "nvidia-auto-select" mode is guaranteed to always be present, and to always be defined as something considered valid by the X driver for this display device. Note that the "nvidia-auto-select" mode is not necessarily the largest possible resolution, nor is it necessarily the mode with the highest refresh rate. Rather, the "nvidia-auto-select" mode is selected such that it is a reasonable default. The selection process is roughly: o If the EDID for the display device reported a preferred mode timing, and that mode timing is considered a valid mode, then that mode is used as the "nvidia-auto-select" mode. You can check if the EDID reported a preferred timing by starting X with logverbosity greater than or equal to 5 (see "Q. How can I increase the amount of data printed in the X log file?" in Chapter 7), and looking at the EDID printout; if the EDID contains a line: Prefer first detailed timing : Yes Then the first mode listed under the "Detailed Timings" in the EDID will be used. o If the EDID did not provide a preferred timing, the best detailed timing from the EDID is used as the "nvidia-auto-select" mode. o If the EDID did not provide any detailed timings (or there was no EDID at all), the best valid mode not larger than 1024x768 is used as the "nvidia-auto-select" mode. The 1024x768 limit is imposed here to restrict use of modes that may have been validated, but may be too large to be considered a reasonable default, such as 2048x1536. o If all else fails, the X driver will use a built-in 800 x 600 60Hz mode as the "nvidia-auto-select" mode. If no modes are requested in the X configuration file, or none of the requested modes can be found in the mode pool, then the X driver falls back to the "nvidia-auto-select" mode, so that X can always start. Appropriate warning messages will be printed to the X log file in these fallback scenarios. You can add the "nvidia-auto-select" mode to your X configuration file by running the command nvidia-xconfig --mode nvidia-auto-select and restarting your X server. The X driver can generally do a much better job of selecting the "nvidia-auto-select" mode if the display device's EDID is available. This is one reason why it is recommended to only use the "UseEDID" X configuration option sparingly. Note that, rather than globally disable all uses of the EDID with the "UseEDID" option, you can individually disable each particular use of the EDID using the "UseEDIDFreqs", "UseEDIDDpi", and/or the "NoEDIDModes" argument in the "ModeValidation" X configuration option. 18F. MODE VALIDATION REPORTING When log verbosity is set to 6 or higher (see "Q. How can I increase the amount of data printed in the X log file?" in Chapter 7), the X log will record every mode that is considered for each display device's mode pool, and report whether the mode passed or failed. For modes that were considered invalid, the log will report why the mode was considered invalid. 18G. ENSURING IDENTICAL MODE TIMINGS Some functionality, such as Active Stereo with TwinView, requires control over exactly which mode timings are used. For explicit control over which mode timings are used on each display device, you can specify the ModeLine you want to use (using one of the ModeLine generators available), and using a unique name. For example, if you wanted to use 1024x768 at 120 Hz on each monitor in TwinView with active stereo, you might add something like this to the monitor section of your X configuration file: # 1024x768 @ 120.00 Hz (GTF) hsync: 98.76 kHz; pclk: 139.05 MHz Modeline "1024x768_120" 139.05 1024 1104 1216 1408 768 769 772 823 -HSync +Vsync Then, in the Screen section of your X config file, specify a MetaMode like this: Option "MetaModes" "1024x768_120, 1024x768_120" ______________________________________________________________________________ Chapter 19. Configuring Flipping and UBB ______________________________________________________________________________ The NVIDIA Accelerated Linux Graphics Driver supports Unified Back Buffer (UBB) and OpenGL Flipping. These features can provide performance gains in certain situations. o Unified Back Buffer (UBB): UBB is available only on NVIDIA RTX/Quadro GPUs (Quadro NVS GPUs excluded) and is enabled by default when there is sufficient video memory available. This can be disabled with the UBB X config option. When UBB is enabled, all windows share the same back, stencil and depth buffers. When there are many windows, the back, stencil and depth usage will never exceed the size of that used by a full screen window. However, even for a single small window, the back, stencil, and depth video memory usage is that of a full screen window. In that case video memory may be used less efficiently than in the non-UBB case. o Flipping: When OpenGL flipping is enabled, OpenGL can perform buffer swaps by changing which buffer is scanned out rather than copying the back buffer contents to the front buffer; this is generally a higher performance mechanism and allows tearless swapping during the vertical retrace (when __GL_SYNC_TO_VBLANK is set). ______________________________________________________________________________ Chapter 20. Using the "/proc" File System Interface ______________________________________________________________________________ The NVIDIA '/proc' file system interface can be used to obtain run-time information about the NVIDIA kernel driver and about the NVIDIA GPUs present in the system. It can also be used to control some aspects of the NVIDIA driver's operation. This information is contained in several files in '/proc/driver/nvidia': '/proc/driver/nvidia/version' Lists the installed driver revision and the version of the GNU C compiler used to build the Linux kernel module. '/proc/driver/nvidia/warnings' The NVIDIA graphics driver tries to detect potential problems with the host system's kernel and warns about them using the kernel's "printk()" mechanism, typically logged by the system to '/var/log/messages'. Important NVIDIA warning messages are also logged to dedicated text files in this '/proc' directory. '/proc/driver/nvidia/gpus/PCI-BUS-ID/information' Provide information about each of the installed NVIDIA GPUs (model name, IRQ, BIOS version, etc.). Note that some of the information is only available when the GPUs have been initialized. See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine PCI-BUS-ID. '/proc/driver/nvidia/gpus/PCI-BUS-ID/power' Provide information about the runtime D3 (RTD3) Power Management status of each of the installed NVIDIA GPUs. Note that some of the information is only available when the GPUs have been initialized. Please see Chapter 22 for the detailed description of this '/proc' interface. See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine the PCI-BUS-ID. '/proc/driver/nvidia/suspend' Notify the NVIDIA kernel drivers of system power management events, such as suspend and hibernate. Please see Chapter 21 for a detailed description of this '/proc' interface. ______________________________________________________________________________ Chapter 21. Configuring Power Management Support ______________________________________________________________________________ 21A. BACKGROUND The NVIDIA Linux driver includes support for the suspend (suspend-to-RAM) and hibernate (suspend-to-disk) system power management operations, such as ACPI S3 and S4 on the x86_64 platform. When the system suspends or hibernates, the NVIDIA kernel drivers prepare in-use GPUs for the sleep cycle, saving state required to return these GPUs to normal operation when the system is later resumed. The NVIDIA Linux driver also supports S0ix-based s2idle system suspend (suspend-to-idle), if both the platform and the NVIDIA GPU support it. The GPU state saved by the NVIDIA kernel drivers includes allocations made in video memory. However, these allocations are collectively large, and typically cannot be evicted. Since the amount of system memory available to drivers at suspend time is often insufficient to accommodate large portions of video memory, the NVIDIA kernel drivers are designed to act conservatively, and normally only save essential video memory allocations. The resulting loss of video memory contents is partially compensated for by the user-space NVIDIA drivers, and by some applications, but can lead to failures such as rendering corruption and application crashes upon exit from power management cycles. To better support power management with these types of applications, the NVIDIA Linux driver provides a custom power management interface intended for integration with system management tools like 'systemd'. This interface is still considered experimental. It is not used by default, but can be taken advantage of by configuring the system as described in this chapter. 21B. OVERVIEW The NVIDIA Linux driver supports the suspend and hibernate power management operations via two different mechanisms. In this section, each is summarized briefly with its capabilities and requirements: Kernel driver callback When this mechanism is used, the NVIDIA kernel driver receives callbacks from the Linux kernel to suspend, hibernate, and to resume each GPU for which a Linux PCI driver was registered. This is the default mechanism: it is enabled and used without explicit configuration. While this mechanism has no special requirements, yields good results with many workloads, and has been supported by the NVIDIA kernel driver in similar form for years, it suffers from a few limitations. Notably, it can only preserve a relatively small amount of video memory reliably, and it cannot support power management when advanced CUDA features are being used. /proc/driver/nvidia/suspend Instead of callbacks from the Linux kernel, this mechanism, when used, relies on a system management tool, such as 'systemd', to issue suspend, hibernate, and resume commands to the NVIDIA kernel driver via the "/proc/driver/nvidia/suspend" interface. It is still considered experimental, and requires explicit configuration to use. If configured correctly, this mechanism is designed to remove the limitations of the kernel driver callback mechanism. It supports power management with advanced CUDA features (such as UVM), and it is capable of saving and restoring all video memory allocations. 21C. PRESERVE ALL VIDEO MEMORY ALLOCATIONS To save potentially large portions of video memory, the NVIDIA driver supports the following two methods: Save allocations in an unnamed temporary file The NVIDIA driver uses an unnamed temporary file to save potentially large portions of video memory. By default, this file is created in '/tmp' during system suspend. This location can be changed with the "NVreg_TemporaryFilePath" nvidia.ko kernel module parameter, e.g. "NVreg_TemporaryFilePath=/run". The destination file system needs to support unnamed temporary files, and it needs to be large enough to accommodate all the utilized video memory copies for the duration of the power management cycle. When determining a suitable size for the video memory backing store, it is recommended to start with the overall amount of video memory supported by the GPUs installed in the system. For example: "nvidia-smi -q -d MEMORY |grep 'FB Memory Usage' -A1". Each "Total" line returned by this command reflects one GPU's video memory capacity, in MiB. The sum of these numbers, plus 5% of margin, is a conservative starting point for the size of the video memory backing store. Please note that file systems such as '/tmp' and '/run' are often of the type "tmpfs", and potentially relatively small. Most commonly, the size of the type of the file system used is controlled by 'systemd'. For more information, see https://www.freedesktop.org/wiki/Software/systemd/APIFileSystems. To achieve the best performance, file system types other than "tmpfs" are recommended at this time. Additionally, to unlock the full functionality of the interface, the NVIDIA Linux kernel module "nvidia.ko" needs to be loaded with the "NVreg_PreserveVideoMemoryAllocations=1" module parameter. This changes the default video memory save/restore strategy to save and restore all video memory allocations. Also, the "/proc/driver/nvidia/suspend" power management mechanism (with a system management tool, such as 'systemd') is required for using this interface. S0ix-based power management If both the platform and the NVIDIA GPU support S0ix-based power management, then the NVIDIA Linux driver will put the GPU video memory in self refresh mode during 's2idle' system suspend. S0ix-based suspend will consume more power than legacy S3 system suspend, but it will enter and exit suspend/resume more quickly. Also, the suspend/resume latency will be constant irrespective of GPU video memory usage. To check the platform S0ix state support and required configuration, follow the steps mentioned in https://web.archive.org/web/20230614200816/https://01.org/blogs/qwang59/2018/how-achieve-s0ix-states-linux To check if the NVIDIA GPU has support for S0ix-based power management, install the NVIDIA driver and run the following command: "grep 'Video Memory Self Refresh' /proc/driver/nvidia/gpus/::.0/power" For example: "grep 'Video Memory Self Refresh' /proc/driver/nvidia/gpus/0000\:01\:00.0/power" If both the platform and the GPU support S0ix-based power management, then the S0ix support can be enabled in the NVIDIA Linux driver by setting the "nvidia.ko" kernel module parameter "NVreg_EnableS0ixPowerManagement" to "1". With "NVreg_EnableS0ixPowerManagement" set to "1" and system suspend state set to 's2idle', the NVIDIA Linux driver will calculate the video memory usage at system suspend time. o During the S0ix suspend, if video memory usage is less than a certain threshold, then the driver will copy video memory contents to system memory and power off the video memory along with the GPU. This will help in saving power. o During the S0ix suspend, if video memory usage is above a certain threshold, then the video memory will be kept in self-refresh mode while the rest of the GPU is powered down. By default, this threshold is 256 MB and it can be changed with the "NVreg_S0ixPowerManagementVideoMemoryThreshold" module parameter of "nvidia.ko". All the module parameters can be set on the command line when loading the NVIDIA Linux kernel module "nvidia.ko", or via the distribution's kernel module configuration files (such as those under /etc/modprobe.d). 21D. 'systemd' CONFIGURATION This section is specific to the '/proc/driver/nvidia/suspend' interface. This is required if using the "NVreg_PreserveVideoMemoryAllocations=1" kernel module parameter or advanced CUDA features (such as UVM). The NVIDIA Linux kernel driver requires no configuration if the default power management mechanism is used. In order to take advantage of the '/proc' interface, a system management tool like 'systemd' needs to be configured to access it at appropriate times in the power management sequence. Specifically, the interface needs to be used to suspend or hibernate the NVIDIA kernel drivers just before writing to the Linux kernel's '/sys/power/state' interface to request entry into the desired sleep state. The interface also needs to be used to resume the NVIDIA kernel drivers immediately after the return from a sleep state, as well as immediately after any unsuccessful attempts to suspend or hibernate. The following example configuration documents integration with the 'systemd' system and service manager, which is commonly used in modern GNU/Linux distributions to manage system start-up and various aspects of its operation. For systems not using 'systemd', the configuration files provided serve as a reference. The 'systemd' configuration uses the following files: /usr/lib/systemd/system/nvidia-suspend.service A 'systemd' service description file used to instruct the system manager to write "suspend" to the '/proc/driver/nvidia/suspend' interface immediately before accessing '/sys/power/state' to suspend the system. /usr/lib/systemd/system/nvidia-suspend-then-hibernate.service A 'systemd' service description file used to instruct the system manager to write "suspend" to the '/proc/driver/nvidia/suspend' interface immediately before accessing '/sys/power/state' to suspend the system. The 'suspend-then-hibernate' suspend method requires systemd version 248 or higher. /usr/lib/systemd/system/nvidia-hibernate.service A 'systemd' service description file used to instruct the system manager to write "hibernate" to the '/proc/driver/nvidia/suspend' interface immediately before accessing '/sys/power/state' to hibernate the system. /usr/lib/systemd/system/nvidia-resume.service A 'systemd' service description file used to instruct the system manager to write "resume" to the '/proc/driver/nvidia/suspend' interface immediately after returning from a system sleep state. /lib/systemd/system-sleep/nvidia A 'systemd-sleep' script file used to instruct the system manager to write "resume" to the '/proc/driver/nvidia/suspend' interface immediately after an unsuccessful attempt to suspend or hibernate the system via the '/proc/driver/nvidia/suspend' interface. For the 'suspend-then-hibernate' method of system sleep, this script is responsible for resuming and then hibernating the GPU if the system wakes due to a low battery warning. This feature requires systemd version 248 or newer. /usr/bin/nvidia-sleep.sh A shell script used by the 'systemd' service description files and the 'systemd-sleep' file to interact with the '/proc/driver/nvidia/suspend' interface. The script also manages VT switching for the X server, which is currently needed by the NVIDIA X driver to support power management operations. These files are installed and enabled by nvidia-installer automatically if systemd is detected. Installation of systemd units can be disabled by specifying the '--no-systemd' installer option. 21E. EXERCISING POWER MANAGEMENT WITH 'systemd' This section is specific to the '/proc/driver/nvidia/suspend' interface, when configured as described above. When the default power management mechanism is used instead, or when the '/proc' interface is used without 'systemd', then the use of "systemctl" is not required. To suspend (suspend-to-RAM) or to hibernate (suspend-to-disk), respectively, use the following commands: o "sudo systemctl suspend" o "sudo systemctl hibernate" For the full list of sleep operations supported by 'systemd', please see the systemd-suspend.service(8) man page. 21F. KNOWN ISSUES AND WORKAROUNDS o On some systems, where the default suspend mode is '"s2idle"', the system may not resume properly due to a known timing issue in the kernel. The suspend mode can be verified by reading the contents of the file '/sys/power/mem_sleep'. The following upstream kernel changes have been proposed to fix the issue: https://lore.kernel.org/linux-pci/20190927090202.1468-1-drake@endlessm.com/ https://lore.kernel.org/linux-pci/20190821124519.71594-1-mika.westerberg@linux.intel.com/ In the interim, the default suspend mode on the affected systems should be set to '"deep"' using the kernel command line parameter '"mem_sleep_default"' - 'mem_sleep_default=deep' ______________________________________________________________________________ Chapter 22. PCI-Express Runtime D3 (RTD3) Power Management ______________________________________________________________________________ 22A. INTRODUCTION NVIDIA GPUs have many power-saving mechanisms. Some of them will reduce clocks and voltages to different parts of the chip, and in some cases turn off clocks or power to parts of the chip entirely, without affecting functionality or while continuing to function, just at a slower speed. However, the lowest power states for NVIDIA GPUs require turning power off to the entire chip, often through ACPI calls. Obviously, this impacts functionality. Nothing can run on the GPU while it is powered off. Care has to be taken to only enter this state when there are no workloads running on the GPU and any attempts to start work or any memory mapped I/O (MMIO) access must be preceded with a sequence to first turn the GPU back on and restore any necessary state. The NVIDIA GPU may have one, two or four PCI functions: o Function 0: VGA controller / 3D controller o Function 1: Audio device o Function 2: USB xHCI Host controller o Function 3: USB Type-C UCSI controller Out of the four PCI functions, the NVIDIA driver directly manages the VGA controller / 3D Controller PCI function. Other PCI functions are managed by the device drivers provided with the Linux kernel. The NVIDIA driver is capable of handling entry into and exit from these low power states, for the PCI function 0. The remaining PCI functions are also powered down along with function 0 when entering these low power states. As a result, the device drivers for the other three functions also need to be taken into account to: prevent entering the lowest-power state when the device is in use, trigger exiting the lowest-power state when the device is needed, save and restore any hardware state around power-off events. The NVIDIA Linux driver includes support for dynamically managing power to the NVIDIA GPU. It depends on the runtime power management framework within the Linux kernel to arbitrate power needs of various PCI functions. In order to have maximum power saving from this feature, two conditions must be met: 1. Runtime power management for all the PCI functions of the GPU should be enabled. 2. The device drivers for all the PCI functions should support runtime power management. If these conditions are satisfied and if all the PCI functions are idle, then The NVIDIA GPU will go to the lowest power state resulting into maximum power savings. 22B. SUPPORTED CONFIGURATIONS This feature is available only when the following conditions are satisfied: o This feature is supported on notebooks and desktops. o This feature requires system hardware as well as ACPI support (ACPI "_PR0" and "_PR3" methods are needed to control PCIe power). The necessary hardware and ACPI support was first added in Intel Coffeelake chipset series. Hence, this feature is supported from Intel Coffeelake chipset series. o This feature requires a Turing or newer GPU. o This feature is supported with Linux kernel versions 4.18 and newer. With older kernel versions, it may not work as intended. o This feature is supported when Linux kernel defines CONFIG_PM (CONFIG_PM=y). Typically, if the system supports S3 (suspend-to-RAM), then CONFIG_PM would be defined. 22C. SYSTEM SETTINGS 1. Enable runtime power management for each PCI function. For Ampere or later notebooks with supported configurations, runtime D3 power management is enabled by default for the GPU's PCI function 0 (VGA controller / 3D controller). For pre-Ampere notebooks and all desktop SKUs, runtime D3 power management can be enabled for each PCI function using the following command. 'echo auto > /sys/bus/pci/devices/::./power/control' For example: 'echo auto > /sys/bus/pci/devices/0000\:01\:00.0/power/control' 22D. DRIVER SETTINGS This feature is enabled by default on supported Ampere or newer notebook computers. In all other configurations, it is disabled by default. It can be enabled or disabled via the 'NVreg_DynamicPowerManagement' nvidia.ko kernel module parameter. 'Option "NVreg_DynamicPowerManagement=0x00"' This setting will disable runtime D3 power management features. With this setting, the NVIDIA driver will only use the GPU's built-in power management so it always is powered on. Actual power usage will vary with the GPU's workload. 'Option "NVreg_DynamicPowerManagement=0x01"' This setting is called coarse-grained power control. With this setting, the NVIDIA GPU driver will allow the GPU to go into its lowest power state when no applications are running that use the nvidia driver stack. Whenever an application requiring NVIDIA GPU access is started, the GPU is put into an active state. When the application exits, the GPU is put into a low power state. 'Option "NVreg_DynamicPowerManagement=0x02"' This setting is called fine-grained power control. With this setting, the NVIDIA GPU driver will allow the GPU to go into its lowest power state when no applications are running that use the nvidia driver stack. Whenever an application requiring NVIDIA GPU access is started, the GPU is put into an active state. When the application exits, the GPU is put into a low power state. Additionally, the NVIDIA driver will actively monitor GPU usage while applications using the GPU are running. When the applications have not used the GPU for a short period, the driver will allow the GPU to be powered down. As soon as the application starts using the GPU, the GPU is reactivated. Furthermore, the NVIDIA GPU driver controls power to the NVIDIA GPU and its video memory separately. While turning off the NVIDIA GPU, the video memory will be kept in a low power self-refresh mode unless the following conditions are met: o The NVIDIA GPU does not support "Video Memory Self Refresh" mode but supports "Video Memory Off" mode. See "Procfs Interface For Runtime D3" in Chapter 22 for details. o The system is configured in PRIME render offload mode. See Chapter 35 for details. o Less than a certain threshold of video memory is in use. The default threshold value is 200 MB. See "Video Memory Utilization Threshold" in Chapter 22 for details. o Sufficient system memory is available for saving the video memory contents. If these conditions are met, the NVIDIA GPU driver will completely turn off the video memory, in addition to the rest of the GPU. Keeping video memory in a self-refresh mode uses more power than turning off video memory, but allows the GPU to be powered off and reactivated more quickly. It is important to note that the NVIDIA GPU will remain in an active state if it is driving a display. In this case, the NVIDIA GPU will go to a low power state only when the X configuration option HardDPMS is enabled and the display is turned off by some means - either automatically due to an OS setting or manually using commands like 'xset'. Similarly, the NVIDIA GPU will remain in an active state if a CUDA application is running. 'Option "NVreg_DynamicPowerManagement=0x03"' This is the default setting. For Ampere or later notebooks with supported configurations, this value translates to fine-grained power control. For pre-Ampere notebooks, this value disables runtime D3 power management features. For desktop computers, irrespective of the GPU(s) used, this value disables runtime D3 power management features. Option "NVreg_DynamicPowerManagement" can be set on the command line while loading the NVIDIA Linux kernel module. For example, 'modprobe nvidia "NVreg_DynamicPowerManagement=0x02"' 22E. VIDEO MEMORY UTILIZATION THRESHOLD The NVIDIA GPU driver uses 200 MB as the default video memory utilization threshold to decide whether the video memory can be turned off or kept in a self-refresh mode. This threshold value can be decreased or increased (up to 1024 MB) using an option 'NVreg_DynamicPowerManagementVideoMemoryThreshold'. This option can be set on the command line while loading the NVIDIA Linux kernel module. For example, 'modprobe nvidia "NVreg_DynamicPowerManagementVideoMemoryThreshold=100"' The video memory utilization threshold value should be a positive integer. It is expressed in Megabytes (1048576 bytes). In the example above, the threshold value will be set to 100 MB. The maximum threshold value can be 1024 MB. Any value greater than 1024 MB will be clamped to 1024 MB. The use of a higher threshold value will increase the latency during RTD3 entry and exit transitions, so use this option only if the latency increase is not affecting the usability of the system. This threshold can be set to 0 in order to prevent the video memory from being turned off. 22F. PROCFS INTERFACE FOR RUNTIME D3 The following entries in the file '/proc/driver/nvidia/gpus/PCI-BUS-ID/power' provide more details regarding the runtime D3 feature. See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine the PCI-BUS-ID. o "Runtime D3 status" entry gives the current status of this feature. o "Video Memory" entry gives the power status of the video memory. o "Video Memory Self Refresh" entry reports whether the NVIDIA GPU hardware supports video memory self refresh mode. o "Video Memory Off" entry reports whether the NVIDIA GPU hardware supports video memory off mode. 22G. KNOWN ISSUES AND WORKAROUNDS The first two workarounds in the below section are not required on Linux kernel 5.5 or newer. 1. The USB xHCI Host controller and USB Type-C UCSI controller drivers present in most Linux distributions do not fully support runtime power management. Full support for runtime power management in these drivers is available in kernel version 5.5. For kernel versions before 5.5, these two PCI functions have to be disabled for this feature to work properly. This can be done using the following command. 'echo 1 > /sys/bus/pci/devices/::.2/remove' 'echo 1 > /sys/bus/pci/devices/::.3/remove' For example: 'echo 1 > /sys/bus/pci/devices/0000\:01\:00.2/remove' 'echo 1 > /sys/bus/pci/devices/0000\:01\:00.3/remove' 2. There is a known issue with the audio driver due to which the audio PCI function remains in an active state. This affects Linux kernel versions 4.19 (starting with git commit id "37a3a98ef601f89100e3bb657fb0e190b857028c") through 5.4, and is fixed in Linux kernel version 5.5. 'echo 1 > /sys/bus/pci/devices/::.1/remove' For example: 'echo 1 > /sys/bus/pci/devices/0000\:01\:00.1/remove' This workaround will result in audio loss when the audio function is being used to play audio over DP/HDMI connection. To recover from audio loss, rescanning the PCI tree will bring back the audio PCI function and audio operation can be recovered. However, after rescanning the PCI tree, all the disabled PCI functions will again become active. To ensure that this feature works again, the workarounds mentioned in this section have to be done again. 3. When the NVIDIA GPU is driving a console, runtime power management is enabled for the VGA Controller PCI function and nvidia driver is uninstalled, the console will become blank. The workaround for this issue is to disable runtime power management for PCI function 0 before uninstalling the NVIDIA driver using the following command: 'echo on > /sys/bus/pci/devices/::./power/control' For example: 'echo on > /sys/bus/pci/devices/0000\:01\:00.0/power/control' 4. On desktops, if a NVIDIA GPU is in RTD3 state and a display is connected to it, it will not be detected. This is because the parent PCI/e port switches off the power to the NVIDIA GPU. The workaround for this issue is to first put the NVIDIA GPU into an active state and then connect the display to it. The NVIDIA GPU can be brought into an active state by running the following command: 'echo on > /sys/bus/pci/devices/::./power/control' For example: 'echo on > /sys/bus/pci/devices/0000\:01\:00.0/power/control' After connecting the display, runtime power management can be enabled again using the following command: 'echo auto > /sys/bus/pci/devices/::./power/control' For example: 'echo auto > /sys/bus/pci/devices/0000\:01\:00.0/power/control' 5. On systems where nvidia persistence daemon is running with UVM persistence mode enabled, PCI-Express Runtime D3 (RTD3) Power Management will be disabled. This is because in uvm persistence mode some UVM driver components are kept initialized even when the GPU is not in active use. This prevents the GPU from entering RTD3 low power state. 22H. AUTOMATED SETUP This section describes automated ways to perform the manual steps mentioned above so that this feature works seamlessly after boot. 1. Create a file named '80-nvidia-pm.rules' in '/lib/udev/rules.d/' directory. Add the content given below to '80-nvidia-pm.rules' file. This would enable runtime power management for the VGA Controller / 3D Controller PCI function. It would also remove Audio device PCI function, USB xHCI Host Controller function as well as USB Type-C UCSI Controller PCI function. Note that the first three rules given below are not required from Linux kernel 5.5 and newer. # Remove NVIDIA USB xHCI Host Controller devices, if present ACTION=="add", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x0c0330", ATTR{remove}="1" # Remove NVIDIA USB Type-C UCSI devices, if present ACTION=="add", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x0c8000", ATTR{remove}="1" # Remove NVIDIA Audio devices, if present ACTION=="add", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x040300", ATTR{remove}="1" # Enable runtime PM for NVIDIA VGA/3D controller devices on driver bind ACTION=="bind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030000", TEST=="power/control", ATTR{power/control}="auto" ACTION=="bind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030200", TEST=="power/control", ATTR{power/control}="auto" # Disable runtime PM for NVIDIA VGA/3D controller devices on driver unbind ACTION=="unbind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030000", TEST=="power/control", ATTR{power/control}="on" ACTION=="unbind", SUBSYSTEM=="pci", ATTR{vendor}=="0x10de", ATTR{class}=="0x030200", TEST=="power/control", ATTR{power/control}="on" 2. The driver option 'NVreg_DynamicPowerManagement' can be set via the distribution's kernel module configuration files (such as those under '/etc/modprobe.d'). For example, the following line can be added to '/etc/modprobe.d/nvidia.conf' file to seamlessly enable this feature. 'options nvidia "NVreg_DynamicPowerManagement=0x02"' 3. Reboot the system. 22I. REPORTING ISSUES For better error reporting, nvidia-bug-report.sh collects a dump of ACPI tables using 'acpidump' utility. Depending on your Linux distribution, this utility may be found in a package called 'acpica-tools' or 'acpica' or similar. ______________________________________________________________________________ Chapter 23. Dynamic Boost on Linux ______________________________________________________________________________ 23A. INTRODUCTION The 'nvidia-powerd' daemon provides support for the NVIDIA Dynamic Boost feature on Linux platforms. Dynamic Boost is a system-wide power controller which manages GPU and CPU power, according to the workload on the system. By shifting power between the GPU and the CPU, Dynamic Boost can deliver more power to the component that would benefit most from it, without impacting the system's total thermal and electrical budgets. This optimizes overall system performance per watt. Dynamic Boost will be active only when the notebook system is powered by AC and there is enough load on the GPU. It will not be engaged when the system is running on battery. Dynamic Boost is intended to improve performance on balanced as well as heavily GPU-bound or CPU-bound applications. Dynamic Boost requests the CPUFreq Governor to set the CPU frequency by updating the '/sys/devices/system/cpu/cpu*/cpufreq/scaling_max_freq' sysfs entries. 23B. HARDWARE REQUIREMENTS Dynamic Boost is supported only on platforms that meet all of the following requirements: 1. Notebook form factor. 2. Ampere or newer GPUs. 3. Intel CometLake or newer Intel chipsets, or AMD Renoir, Cezanne, Rembrandt, or newer AMD chipsets. 4. SBIOS support for Dynamic Boost Run the following command to check if the SBIOS supports Dynamic Boost: `cat /proc/driver/nvidia/gpus/domain:bus:device.function/power` See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine the PCI-BUS-ID. Alternatively, use the following command: `nvidia-settings -q DynamicBoostSupport` 23C. SOFTWARE REQUIREMENTS The system must fulfill all of the following requirements: 1. Use the 'systemd' init daemon. 2. Support the 'D-Bus' message bus system. 3. Use the 'cpufreq' infrastructure. 23D. CONFIGURATION STEPS Note: The following commands must be run with root permissions 1. Copy the dbus configuration file 'nvidia-dbus.conf' from '/usr/share/doc/NVIDIA_GLX-1.0/' to '/etc/dbus-1/system.d'. If the '/etc/dbus-1/system.d' directory does not exist, create it before copying the file and reboot the system so that dbus can scan the newly created directory in the next boot. 2. To enable the nvidia-powerd service, causing it to start automatically on boot: `systemctl enable nvidia-powerd.service` 3. To start the nvidia-powerd service: `systemctl start nvidia-powerd.service` Steps to Stop and disable the nvidia-powerd service: 1. To stop the service: `systemctl stop nvidia-powerd.service` 2. To disable the service such that it does not automatically start on boot: `systemctl disable nvidia-powerd.service` ______________________________________________________________________________ Chapter 24. Using the X Composite Extension ______________________________________________________________________________ X.Org X servers support an X protocol extension called Composite. This extension allows windows to be drawn into pixmaps instead of directly onto the screen. In conjunction with the Damage and Render extensions, this allows a program called a composite manager to blend windows together to draw the screen. Performance will be degraded significantly if the "RenderAccel" option is disabled in xorg.conf. See Appendix B for more details. When the NVIDIA X driver is used and the Composite extension is enabled, NVIDIA's OpenGL implementation interacts properly with the Damage and Composite X extensions. This means that OpenGL rendering is drawn into offscreen pixmaps and the X server is notified of the Damage event when OpenGL renders to the pixmap. This allows OpenGL applications to behave properly in a composited X desktop. The Composite X extension is enabled by default. It can be disabled with 'nvidia-xconfig --no-composite'. See the nvidia-xconfig(1) man page for details. The Composite extension causes problems with some driver components: o Xv adaptors will ignore the sync-to-vblank option when drawing into a redirected window. o Workstation overlays are incompatible with Composite. Workstation overlays will be automatically disabled when Composite is detected. o The NVIDIA Linux driver supports quad-buffered stereo together with Composite. However, many composite managers will only texture from and display the left eye content, effectively disabling the stereo effect. Workarounds for this problem are either to use a different composite manager that supports stereo or disable the Composite extension. o The Composite extension is incompatible with Xinerama in X.Org X servers prior to version 1.10. Composite will be automatically disabled when Xinerama is enabled on those servers. o Prior to X.Org X server version 1.15, the Damage extension does not properly report rendering events on all physical X screens in Xinerama configurations. This prevents most composite mangers from rendering correctly. This NVIDIA Linux driver supports OpenGL rendering to 32-bit ARGB windows. 32-bit visuals are only available on screens with depths 24 or 30. If you are an application developer, you can use these new visuals in conjunction with a composite manager to create translucent OpenGL applications: int attrib[] = { GLX_RENDER_TYPE, GLX_RGBA_BIT, GLX_DRAWABLE_TYPE, GLX_WINDOW_BIT, GLX_RED_SIZE, 1, GLX_GREEN_SIZE, 1, GLX_BLUE_SIZE, 1, GLX_ALPHA_SIZE, 1, GLX_DOUBLEBUFFER, True, GLX_DEPTH_SIZE, 1, None }; GLXFBConfig *fbconfigs, fbconfig; int numfbconfigs, render_event_base, render_error_base; XVisualInfo *visinfo; XRenderPictFormat *pictFormat; /* Make sure we have the RENDER extension */ if(!XRenderQueryExtension(dpy, &render_event_base, &render_error_base)) { fprintf(stderr, "No RENDER extension found\n"); exit(EXIT_FAILURE); } /* Get the list of FBConfigs that match our criteria */ fbconfigs = glXChooseFBConfig(dpy, scrnum, attrib, &numfbconfigs); if (!fbconfigs) { /* None matched */ exit(EXIT_FAILURE); } /* Find an FBConfig with a visual that has a RENDER picture format that * has alpha */ for (i = 0; i < numfbconfigs; i++) { visinfo = glXGetVisualFromFBConfig(dpy, fbconfigs[i]); if (!visinfo) continue; pictFormat = XRenderFindVisualFormat(dpy, visinfo->visual); if (!pictFormat) continue; if(pictFormat->direct.alphaMask > 0) { fbconfig = fbconfigs[i]; break; } XFree(visinfo); } if (i == numfbconfigs) { /* None of the FBConfigs have alpha. Use a normal (opaque) * FBConfig instead */ fbconfig = fbconfigs[0]; visinfo = glXGetVisualFromFBConfig(dpy, fbconfig); pictFormat = XRenderFindVisualFormat(dpy, visinfo->visual); } XFree(fbconfigs); When rendering to a 32-bit window, keep in mind that composite managers expect "premultiplied alpha" colors. This means that if your color has components (r,g,b) and alpha value a, then you must render (a*r, a*g, a*b, a) into the target window. More information about Composite can be found at http://freedesktop.org/Software/CompositeExt ______________________________________________________________________________ Chapter 25. Using the nvidia-settings Utility ______________________________________________________________________________ A graphical configuration utility, 'nvidia-settings', is included with the NVIDIA Linux graphics driver. After installing the driver and starting X, you can run this configuration utility by running: % nvidia-settings in a terminal window. nvidia-settings requires version 2.4 or later of the GTK+ 2 library. Some architectures of Linux support the GTK+ 3 library and would require version 3.0 or later if available. Detailed information about the configuration options available are documented in the help window in the utility. For more information, see the nvidia-settings man page. The source code to nvidia-settings is released as GPL and is available here: https://download.nvidia.com/XFree86/nvidia-settings/ ______________________________________________________________________________ Chapter 26. NVIDIA Spectre V2 Mitigation ______________________________________________________________________________ The NVIDIA Linux driver supports the retpoline Spectre V2 mitigation technique as identified by: https://software.intel.com/security-software-guidance/api-app/sites/default/files/Retpoline-A-Branch-Target-Injection-Mitigation.pdf For Linux systems that define CONFIG_RETPOLINE or CONFIG_MITIGATION_RETPOLINE, the NVIDIA driver yields to the kernel's implementation of Spectre V2 mitigation. When CONFIG_RETPOLINE or CONFIG_MITIGATION_RETPOLINE is not defined, the NVIDIA Linux driver implements the retpoline thunk without the Spectre V2 mitigation. Linux kernels that implement retpoline based Spectre V2 mitigation provide a kernel boot flag to enable or disable the feature. When enabled, users may notice a performance degradation. In order to recover performance, the user can disable the feature using the 'spectre_v2' boot flag (unsafe). spectre_v2=off Please note that CONFIG_RETPOLINE has been renamed to CONFIG_MITIGATION_RETPOLINE by commit "aefb2f2e619b" ("x86/bugs: Rename CONFIG_RETPOLINE => CONFIG_MITIGATION_RETPOLINE") in Linux kernel v6.9. ______________________________________________________________________________ Chapter 27. Using the nvidia-smi Utility ______________________________________________________________________________ A monitoring and management command line utility, 'nvidia-smi', is included with the NVIDIA Linux graphics driver. After installing the driver, you can run this utility by running: % nvidia-smi in a terminal window. Detailed help information is available via the --help command line option and via the nvidia-smi man page. To include nvidia-smi information in other applications see Chapter 28 ______________________________________________________________________________ Chapter 28. The NVIDIA Management Library ______________________________________________________________________________ A C-based API for monitoring and managing various states of the NVIDIA GPU devices. NVIDIA Management Library (NVML) provides a direct access to the queries and commands exposed via nvidia-smi. NVML is included with the NVIDIA Linux graphics driver. To write applications against this library see the NVML developer page: http://developer.nvidia.com/nvidia-management-library-NVML To include NVML functionality in scripting languages see: http://search.cpan.org/~nvbinding/nvidia-ml-pl/lib/nvidia/ml.pm and http://pypi.python.org/pypi/nvidia-ml-py/ ______________________________________________________________________________ Chapter 29. Using the nvidia-debugdump Utility ______________________________________________________________________________ A utility for collecting internal GPU state, 'nvidia-debugdump', is included with the NVIDIA Linux graphics driver. After installing the driver, you can run this utility by running: % nvidia-debugdump in a terminal window. Detailed help information is available via the --help command line option, or when no parameters are supplied. In most cases, this utility is invoked by the 'nvidia-bug-report.sh' (/usr/bin/nvidia-bug-report.sh) script. In rare cases, typically when directed by an NVIDIA technical support representative, nvidia-debugdump may also be invoked as a stand-alone diagnostics program. The "dump" output of nvidia-debugdump is a binary blob that requires internal NVIDIA engineering tools in order to be interpreted. ______________________________________________________________________________ Chapter 30. Using the nvidia-persistenced Utility ______________________________________________________________________________ 30A. BACKGROUND A Linux daemon utility, 'nvidia-persistenced', addresses an undesirable side effect of the NVIDIA kernel driver behavior in certain computing environments. Whenever the NVIDIA device resources are no longer in use, the NVIDIA kernel driver will tear down the device state. Normally, this is the intended behavior of the device driver, but for some applications, the latencies incurred by repetitive device initialization can significantly impact performance. To avoid this behavior, 'nvidia-persistenced' provides a configuration option called "persistence mode" that can be set by NVIDIA management software, such as 'nvidia-smi'. When persistence mode is enabled, the daemon holds the NVIDIA character device files open, preventing the NVIDIA kernel driver from tearing down device state when no other process is using the device. This utility does not actually use any device resources itself - it will simply sleep while maintaining a reference to the NVIDIA device state. 30B. USAGE 'nvidia-persistenced' is included with the NVIDIA Linux GPU driver. After installing the driver, this utility may be installed to run on system startup or manually with the command: # nvidia-persistenced in a terminal window. Note that the daemon may require root privileges to create its runtime data directory, /var/run/nvidia-persistenced/, or it may otherwise need to be run as a user that has access to that directory. Detailed help and usage information is available primarily via the 'nvidia-persistenced' man page, as well as the '--help' command line option. The source code to nvidia-persistenced is released under the MIT license and is available at: https://download.nvidia.com/XFree86/nvidia-persistenced/. 30C. TROUBLESHOOTING If you have difficulty getting 'nvidia-persistenced' to work as expected, the best way to gather information as to what is happening is to run the daemon with the '--verbose' option. 'nvidia-persistenced' detaches from its parent process very early on, and as such only invalid command line argument errors will be printed in the terminal window. All other output, including verbose informational messages, are sent to the syslog interface instead. Consult your distribution's documentation for accessing syslog output. 30D. NOTES FOR PACKAGE MAINTAINERS The daemon utility 'nvidia-persistenced' is installed by the NVIDIA Linux GPU driver installer, but it is not installed to run on system startup. Due to the wide variety of init systems used by the various Linux distributions that the NVIDIA Linux GPU driver supports, we request that package maintainers for those distributions provide the packaging necessary to integrate well with their platform. NVIDIA provides sample init scripts for some common init systems in /usr/share/doc/NVIDIA_GLX-1.0/sample/nvidia-persistenced-init.tar.bz2 to aid in installation of the utility. 'nvidia-persistenced' is intended to be run as a daemon from system initialization, and is generally designed as a tool for compute-only platforms where the NVIDIA device is not used to display a graphical user interface. As such, depending on how your package is typically used, it may not be necessary to install the daemon to run on system initialization. If 'nvidia-persistenced' is packaged to run on system initialization, the package installation, init script or system management utility that runs the daemon should provide the following: A non-root user to run as It is strongly recommended, though not required, that the daemon be run as a non-root user for security purposes. The daemon may either be started with root privileges and the '--user' option, or it may be run directly as the non-root user. Runtime access to /var/run/nvidia-persistenced/ The daemon must be able to create its socket and PID file in this directory. If the daemon is run as root, it will create this directory itself and remove it when it shuts down cleanly. If the daemon is run as a non-root user, this directory must already exist, and the daemon will not attempt to remove it when it shuts down cleanly. If the daemon is started as root, but provided a non-root user to run as via the '--user' option, the daemon will create this directory itself, 'chown' it to the provided user, and 'setuid' to the provided user to drop root privileges. The daemon may be unable to remove this directory when it shuts down cleanly, depending on the privileges of the provided user. ______________________________________________________________________________ Chapter 31. Configuring SLI Mosaic ______________________________________________________________________________ The NVIDIA Linux driver contains support for NVIDIA SLI Mosaic. This technology enables you to extend a single X screen transparently across all of the available display outputs on each GPU. See below for the exact set of configurations which can be used with SLI Mosaic Mode. If any SLI mode is enabled, applications may override which rendering mode is in use by creating an OpenGL context with the GLX_CONTEXT_MULTIGPU_ATTRIB_NV attribute. In addition, applications may gain explicit control over individual GPU rendering in SLI configurations through the GL_NV_gpu_multicast extension by creating a context with the GLX_CONTEXT_MULTIGPU_ATTRIB_MULTICAST_NV attribute. Multicast rendering in SLI Mosaic configurations requires use of the GLX_CONTEXT_MULTIGPU_ATTRIB_MULTI_DISPLAY_MULTICAST_NV attribute, which is only allowed on Quadro GPUs. 31A. ENABLING SLI SLI is enabled by setting the "SLI" option in the X configuration file; see Appendix B for details about the SLI option. The nvidia-xconfig utility can be used to set the SLI option, rather than modifying the X configuration file by hand. For example: % nvidia-xconfig --sli=mosaic 31B. ENABLING SLI MOSAIC MODE The simplest way to configure SLI Mosaic Mode using a grid of monitors is to use 'nvidia-settings' (see Chapter 25). The steps to perform this configuration are as follows: 1. Connect each of the monitors you would like to use to any connector from any GPU used for SLI Mosaic Mode. If you are going to use fewer monitors than there are connectors, connect one monitor to each GPU before adding a second monitor to any GPUs. 2. Install the NVIDIA display driver set. 3. Configure an X screen to use the "nvidia" driver on at least one of the GPUs (see Chapter 6 for more information). 4. Start X. 5. Run 'nvidia-settings' and open the "X Server Display Configuration" page using the navigation pane on the left side of the application. 6. Below the layout window, enable the "Use SLI Mosaic Mode" check box if not already enabled. Then click the "Configure SLI Mosaic Displays" button. You may need to click the "Advanced..." button near the bottom of the page. 7. In the "Configure SLI Mosaic Layout" dialog, select the monitor grid configuration you'd like to use from the "Display Configuration" dropdown. 8. Choose the resolution and refresh rate at which you would like to drive each individual monitor. 9. Set any overlap you would like between the displays. 10. Click the "Apply to Layout" button. The configuration described in the dialog will be loaded into the "Display Configuration" page and can be adjusted, if needed. 11. Click the "Save to X Configuration File" button. NOTE: If you don't have permissions to write to your system's X configuration file, you will be prompted to choose a location to save the file. After doing so, you MUST copy the X configuration file into a location the X server will consider upon startup (usually '/etc/X11/xorg.conf'). 12. Exit nvidia-settings and restart your X server. Alternatively, nvidia-xconfig can be used to configure SLI Mosaic Mode via a command like 'nvidia-xconfig --sli=Mosaic --metamodes=METAMODES' where the METAMODES string specifies the desired grid configuration. For example: nvidia-xconfig --sli=Mosaic --metamodes="GPU-0.DFP-0: 1920x1024+0+0, GPU-0.DFP-1: 1920x1024+1920+0, GPU-1.DFP-0: 1920x1024+0+1024, GPU-1.DFP-1: 1920x1024+1920+1024" will configure four DFPs in a 2x2 configuration, each running at 1920x1024, with the two DFPs on GPU-0 driving the top two monitors of the 2x2 configuration, and the two DFPs on GPU-1 driving the bottom two monitors of the 2x2 configuration. See the MetaModes X configuration description in details in Chapter 12. See Appendix C for further details on GPU and Display Device Names. 31C. HARDWARE REQUIREMENTS SLI functionality requires: o Identical PCI Express graphics cards o A supported motherboard o For the displays to be synchronized across GPUs, the GPUs must be connected using an SLI video bridge or nvlink bridge. o SLI Mosaic Mode requires NVIDIA Quadro GPUs. 31D. OTHER NOTES AND REQUIREMENTS The following other requirements apply to SLI: o Mobile GPUs are NOT supported o GPUs with ECC enabled may not be used in an SLI configuration o If X is configured to use multiple screens and screen 0 has SLI enabled, the other screens configured to use the nvidia driver will be disabled. Note that if SLI is enabled, the GPUs used by that configuration will be unavailable for single GPU rendering. FREQUENTLY ASKED SLI QUESTIONS Q. Why does SLI fail to initialize? A. There are several reasons why SLI may fail to initialize. Most of these should be clear from the warning message in the X log file; e.g.: o "Unsupported bus type" o "The video link was not detected" o "GPUs do not match" o "Unsupported GPU video BIOS" o "Insufficient PCIe link width" The warning message "'Unsupported PCI topology'" is likely due to problems with your Linux kernel. The NVIDIA driver must have access to the PCI Bridge (often called the Root Bridge) that each NVIDIA GPU is connected to in order to configure SLI correctly. There are many kernels that do not properly recognize this bridge and, as a result, do not allow the NVIDIA driver to access this bridge. See the below "How can I determine if my kernel correctly detects my PCI Bridge?" FAQ for details. Below are some specific troubleshooting steps to help deal with SLI initialization failures. o Make sure that ACPI is enabled in your kernel. NVIDIA's experience has been that ACPI is needed for the kernel to correctly recognize the Root Bridge. Note that in some cases, the kernel's version of ACPI may still have problems and require an update to a newer kernel. o Run 'lspci' to check that multiple NVIDIA GPUs can be identified by the operating system; e.g: % /sbin/lspci | grep -i nvidia If 'lspci' does not report all the GPUs that are in your system, then this is a problem with your Linux kernel, and it is recommended that you use a different kernel. Please note: the 'lspci' utility may be installed in a location other than '/sbin' on your system. If the above command fails with the error: "'/sbin/lspci: No such file or directory'", please try: % lspci | grep -i nvidia , instead. You may also need to install your distribution's "pciutils" package. o Make sure you have the most recent SBIOS available for your motherboard. o The PCI Express slots on the motherboard must provide a minimum link width. Please make sure that the PCI Express slot(s) on your motherboard meet the following requirements and that you have connected the graphics board to the correct PCI Express slot(s): o A pair of boards requires one of the following supported link width combinations: o x16 + x16 o x16 + x8 o x16 + x4 o x8 + x8 Q. How can I determine if my kernel correctly detects my PCI Bridge? A. As discussed above, the NVIDIA driver must have access to the PCI Bridge that each NVIDIA GPU is connected to in order to configure SLI correctly. The following steps will identify whether the kernel correctly recognizes the PCI Bridge: o Identify both NVIDIA GPUs: % /sbin/lspci | grep -i vga 0a:00.0 VGA compatible controller: nVidia Corporation [...] 81:00.0 VGA compatible controller: nVidia Corporation [...] o Verify that each GPU is connected to a bus connected to the Root Bridge (note that the GPUs in the above example are on buses 0a and 81): % /sbin/lspci -t good: -+-[0000:80]-+-00.0 | +-01.0 | \-0e.0-[0000:81]----00.0 ... \-[0000:00]-+-00.0 +-01.0 +-01.1 +-0e.0-[0000:0a]----00.0 bad: -+-[0000:81]---00.0 ... \-[0000:00]-+-00.0 +-01.0 +-01.1 +-0e.0-[0000:0a]----00.0 Note that in the first example, bus 81 is connected to Root Bridge 80, but that in the second example there is no Root Bridge 80 and bus 81 is incorrectly connected at the base of the device tree. In the bad case, the only solution is to upgrade your kernel to one that properly detects your PCI bus layout. ______________________________________________________________________________ Chapter 32. Configuring Frame Lock and Genlock ______________________________________________________________________________ NOTE: Frame Lock and Genlock features are supported only on specific hardware, as noted below. Visual computing applications that involve multiple displays, or even multiple windows within a display, can require special signal processing and application controls in order to function properly. For example, in order to produce quality video recording of animated graphics, the graphics display must be synchronized with the video camera. As another example, applications presented on multiple displays must be synchronized in order to complete the illusion of a larger, virtual canvas. This synchronization is enabled through the Frame Lock and Genlock capabilities of the NVIDIA driver. This section describes the setup and use of Frame Lock and Genlock. 32A. DEFINITION OF TERMS GENLOCK: Genlock refers to the process of synchronizing the pixel scanning of one or more displays to an external synchronization source. Genlock requires the external signal to be either TTL or composite, such as used for NTSC, PAL, or HDTV. It should be noted that Genlock is guaranteed only to be frame-synchronized, and not necessarily pixel-synchronized. FRAME LOCK: Frame Lock involves the use of hardware to synchronize the frames on each display in a connected system. When graphics and video are displayed across multiple monitors, Frame Locked systems help maintain image continuity to create a virtual canvas. Frame Lock is especially critical for stereo viewing, where the left and right fields must be in sync across all displays. In short, to enable Genlock means to sync to an external signal. To enable Frame Lock means to sync 2 or more display devices to a signal generated internally by the hardware, and to use both means to sync 2 or more display devices to an external signal. SWAP SYNC: Swap sync refers to the synchronization of buffer swaps of multiple application windows. By means of swap sync, applications running on multiple systems can synchronize the application buffer swaps between all the systems. In order to work across multiple systems, swap sync requires that the systems are Frame Locked. QUADRO SYNC DEVICE: A Quadro Sync Device refers to a device capable of Frame Lock/Genlock. See "Supported Hardware" below. 32B. SUPPORTED HARDWARE Frame Lock and Genlock are supported for the following hardware: o Quadro Sync II, used in conjunction with an NVIDIA RTX A6000, NVIDIA A40, Quadro RTX 8000, Quadro RTX 6000, Quadro RTX 5000, Quadro RTX 4000, Quadro GV100, Quadro GP100, Quadro P6000, Quadro P5000, or Quadro P4000 o Quadro Sync, used in conjunction with a Quadro M6000 24GB, Quadro M6000, Quadro M5000, Quadro M4000, Quadro K6000, Quadro K5200, Quadro K5000, or Quadro K4200 32C. HARDWARE SETUP Before you begin, you should check that your hardware has been properly installed. The following steps must be performed while the system is off. 1. On a Quadro Sync card with four Sync connectors, connect a ribbon cable to any of the four connectors, if none are already connected. 2. Install the Quadro Sync card in any available slot. Note that the slot itself is only used for physical mounting, so even a known "bad" slot is acceptable. The slot must be close enough to the graphics card that the ribbon cable can reach. 3. On a Quadro Sync card with four Sync connectors, external power is required. Connect a 6-pin PCIe power cable or a SATA power cable to the card. 4. Connect the other end of the ribbon cable to the Quadro Sync connector on the graphics card. On supported Quadro Kepler cards, the Quadro Sync connector is identical in appearance to the SLI connector. The ribbon cable from the Quadro Sync card should be connected to the connector labeled "SDI | SYNC". If the ribbon cable is connected to the SLI connector, the GPU will not be able to synchronize with the Quadro Sync card. You may now boot the system and begin the software setup of Genlock and/or Frame Lock. These instructions assume that you have already successfully installed the NVIDIA Accelerated Linux Driver Set. If you have not done so, see Chapter 4. 32D. CONFIGURATION WITH NVIDIA-SETTINGS GUI Frame Lock and Genlock are configured through the nvidia-settings utility. See the 'nvidia-settings(1)' man page, and the nvidia-settings online help (click the "Help" button in the lower right corner of the interface for per-page help information). From the nvidia-settings Frame Lock panel, you may control the addition of Quadro Sync (and display) devices to the Frame Lock/Genlock group, monitor the status of that group, and enable/disable Frame Lock and Genlock. After the system has booted and X Windows has been started, run nvidia-settings as % nvidia-settings You may wish to start this utility before continuing, as we refer to it frequently in the subsequent discussion. The setup of Genlock and Frame Lock are described separately. We then describe the use of Genlock and Frame Lock together. 32E. GENLOCK SETUP After the system has been booted, connect the external signal to the house sync connector (the BNC connector) on either the graphics card or the Quadro Sync card. There is a status LED next to the connector. A solid red or unlit LED indicates that the hardware cannot detect the timing signal. A green LED indicates that the hardware is detecting a timing signal. An occasional red flash is okay. On a Quadro Sync card with four Sync connectors, a blinking green LED indicates that the server is locked to the house sync. The Quadro Sync device (graphics card or Quadro Sync card) will need to be configured correctly for the signal to be detected. In the Frame Lock panel of the nvidia-settings interface, add the X Server that contains the display and Quadro Sync devices that you would like to sync to this external source by clicking the "Add Devices..." button. An X Server is typically specified in the format "system:m", e.g.: mycomputer.domain.com:0 or localhost:0 After adding an X Server, rows will appear in the "Quadro Sync Devices" section on the Frame Lock panel that displays relevant status information about the Quadro Sync devices, GPUs attached to those Quadro Sync devices and the display devices driven by those GPUs. In particular, the Quadro Sync rows will display the server name and Quadro Sync device number along with "Receiving" LED, "Rate", "House" LED, "Port 0"/"Port 1" Images, and "Delay" information. The GPU rows will display the GPU product name information along with the GPU ID for the server. The Display Device rows will show the display device name and device type along with server/client check boxes, refresh rate, "Timing" LED and "Stereo" LED. Once the Quadro Sync and display devices have been added to the Frame Lock/Genlock group, a Server display device will need to be selected. This is done by selecting the "Server" check box of the desired display device. If you are using a Quadro Sync card, you must also click the "Use House Sync if Present" check box. To enable synchronization of this Quadro Sync device to the external source, click the "Enable Frame Lock" button. The display device(s) may take a moment to stabilize. If it does not stabilize, you may have selected a synchronization signal that the system cannot support. You should disable synchronization by clicking the "Disable Frame Lock" button and check the external sync signal. Modifications to Genlock settings (e.g., "Use House Sync if Present", "Add Devices...") must be done while synchronization is disabled. 32F. FRAME LOCK SETUP Frame Lock is supported across an arbitrary number of Quadro Sync systems, although mixing different generations of Quadro Sync products in the same Frame Lock group is not supported. Additionally, each system to be included in the Frame Lock group must be configured with identical mode timings. See Chapter 18 for information on mode timings. Connect the systems through their RJ45 ports using standard CAT5 patch cables. These ports are located on the Frame Lock card. DO NOT CONNECT A FRAME LOCK PORT TO AN ETHERNET CARD OR HUB. DOING SO MAY PERMANENTLY DAMAGE THE HARDWARE. The connections should be made in a daisy-chain fashion: each card has two RJ45 ports, call them 1 and 2. Connect port 1 of system A to port 2 of system B, connect port 1 of system B to port 2 of system C, etc. Note that you will always have two empty ports in your Frame Lock group. The ports self-configure as inputs or outputs once Frame Lock is enabled. Each port has a yellow and a green LED that reflect this state. A flashing yellow LED indicates an output and a flashing green LED indicates an input. On a Quadro Sync card with four Sync connectors, a solid green LED indicates that the port has been configured as an input, but no sync pulse is detected, and a solid yellow LED means the card is configured as an output, but no sync is being transmitted. In the Frame Lock panel of the nvidia-settings interface, add the X server that contains the display devices that you would like to include in the Frame Lock group by clicking the "Add Devices..." button (see the description for adding display devices in the previous section on GENLOCK SETUP. Like the Genlock status indicators, the "Port 0" and "Port 1" columns in the table on the Frame Lock panel contain indicators whose states mirror the states of the physical LEDs on the RJ45 ports. Thus, you may monitor the status of these ports from the software interface. Any X Server can be added to the Frame Lock group, provided that 1. The system supporting the X Server is configured to support Frame Lock and is connected via RJ45 cable to the other systems in the Frame Lock group. 2. The system driving nvidia-settings can communicate with the X server that is to be included for Frame Lock. This means that either the server must be listening over TCP and the system's firewall is permissive enough to allow remote X11 display connections, or that you've configured an alternative mechanism such as ssh(1) forwarding between the machines. For the case of listening over TCP, verify that the "-nolisten tcp" commandline option was not used when starting the X server. You can find the X server commandline with a command such as % ps ax | grep X If "-nolisten tcp" is on the X server commandline, consult your Linux distribution documentation for details on how to properly remove this option. For example, distributions configured to use the GDM login manager may need to set "DisallowTCP=false" in the GDM configuration file (e.g., /etc/gdm/custom.conf, /etc/X11/gdm/gdm.conf, or /etc/gdb/gdb.conf; the exact configuration file name and path varies by the distribution). Or, distributions configured to use the KDM login manager may have the line ServerArgsLocal=-nolisten tcp in their kdm file (e.g., /etc/kde3/kdm/kdmrc). This line can be commented out by prepending with "#". Starting with version 1.17, the X.org X server no longer allows listening over TCP by default when built with its default build configuration options. On newer X servers that were not built with --enable-listen-tcp at build configuration time, in addition to ensuring that "-nolisten tcp" is not set on the X server commandline, you will also need to ensure that "-listen tcp" is explicitly set. 3. The system driving nvidia-settings can locate and has display privileges on the X server that is to be included for Frame Lock. A system can gain display privileges on a remote system by executing % xhost + on the remote system. See the xhost(1) man page for details. Typically, Frame Lock is controlled through one of the systems that will be included in the Frame Lock group. While this is not a requirement, note that nvidia-settings will only display the Frame Lock panel when running on an X server that supports Frame Lock. To enable synchronization on these display devices, click the "Enable Frame Lock" button. The screens may take a moment to stabilize. If they do not stabilize, you may have selected mode timings that one or more of the systems cannot support. In this case you should disable synchronization by clicking the "Disable Frame Lock" button and refer to Chapter 18 for information on mode timings. Modifications to Frame Lock settings (e.g. "Add/Remove Devices...") must be done while synchronization is disabled. nvidia-settings will not automatically enable Frame Lock via the nvidia-settings.rc file. To enable Frame Lock when starting the X server, a line such as the following can be added to the '~/.xinitrc' file: # nvidia-settings -a [gpu:0]/FrameLockEnable=1 32G. FRAME LOCK + GENLOCK The use of Frame Lock and Genlock together is a simple extension of the above instructions for using them separately. You should first follow the instructions for Frame Lock Setup, and then to one of the systems that will be included in the Frame Lock group, attach an external sync source. In order to sync the Frame Lock group to this single external source, you must select a display device driven by the GPU connected to the Quadro Sync card that is connected to the external source to be the signal server for the group. This is done by selecting the check box labeled "Server" of the tree on the Frame Lock panel in nvidia-settings. If you are using a Quadro Sync based Frame Lock group, you must also select the "Use House Sync if Present" check box. Enable synchronization by clicking the "Enable Frame Lock" button. As with other Frame Lock/Genlock controls, you must select the signal server while synchronization is disabled. 32H. GPU STATUS LEDS ON THE QUADRO SYNC CARD In addition to the graphical indicators in the control panel described in the Genlock Setup section above, the Quadro Sync card has two status LEDs for each of the four ports: A sync status LED indicates the sync status for each port. An unlit LED indicates that no GPU is connected to the port; a steady amber LED indicates that a GPU is connected, but not synced to any sync source; and a steady green LED indicates that a GPU is connected and in sync with an internal or external sync source. A flashing LED indicates that a connected GPU is in the process of locking to a sync source; flashing green indicates that the sync source's timings are within a reasonable range, and flashing amber indicates that the timings are out of range, and the GPU may be unable to lock to the sync source. A stereo status LED indicates the stereo sync status for each port. The LED will be lit steady amber when the card first powers on. An unlit LED indicates that stereo is not active, or that no GPU is connected; a blinking green LED indicates that stereo is active, but not locked to the stereo master; and a steady green LED indicates that stereo is active and locked to the stereo master. 32I. CONFIGURATION WITH NVIDIA-SETTINGS COMMAND LINE Frame Lock may also be configured through the nvidia-settings command line. This method of configuring Frame Lock may be useful in a scripted environment to automate the setup process. (Note that the examples listed below depend on the actual hardware configuration and as such may not work as-is.) To properly configure Frame Lock, the following steps should be completed: 1. Make sure Frame Lock Sync is disabled on all GPUs. 2. Make sure all display devices that are to be Frame Locked have the same refresh rate. 3. Configure which (display/GPU) device should be the master. 4. Configure house sync (if applicable). 5. Configure the slave display devices. 6. Enable Frame Lock sync on the master GPU. 7. Enable Frame Lock sync on the slave GPUs. 8. Toggle the test signal on the master GPU (for testing the hardware connectivity.) For a full list of the nvidia-settings Frame Lock attributes, please see the 'nvidia-settings(1)' man page. Examples: 1. 1 System, 1 Frame Lock board, 1 GPU, and 1 display device syncing to the house signal: # - Make sure Frame Lock sync is disabled nvidia-settings -a [gpu:0]/FrameLockEnable=0 nvidia-settings -q [gpu:0]/FrameLockEnable # - Enable use of house sync signal nvidia-settings -a [framelock:0]/FrameLockUseHouseSync=1 # - Configure the house sync signal video mode nvidia-settings -a [framelock:0]/FrameLockVideoMode=0 # - Query the enabled displays on the gpu(s) nvidia-settings -V all -q gpus # - Check the refresh rate is as desired nvidia-settings -q [dpy:DVI-I-0]/RefreshRate # - Query the valid Frame Lock configurations for the display device nvidia-settings -q [dpy:DVI-I-0]/FrameLockDisplayConfig # - Set DVI-I-0 as a slave (this display will be synchronized to the # input signal) # # NOTE: FrameLockDisplayConfig takes one of three values: # 0 (disabled), 1 (client), 2 (server). nvidia-settings -a [dpy:DVI-I-0]/FrameLockDisplayConfig=0 # - Enable Frame Lock nvidia-settings -a [gpu:0]/FrameLockEnable=1 # - Toggle the test signal nvidia-settings -a [gpu:0]/FrameLockTestSignal=1 nvidia-settings -a [gpu:0]/FrameLockTestSignal=0 2. 2 Systems, each with 2 GPUs, 1 Frame Lock board and 1 display device per GPU syncing from the first system's first display device: # - Make sure Frame Lock sync is disabled on all gpus nvidia-settings -a myserver:0[gpu]/FrameLockEnable=0 nvidia-settings -a myslave1:0[gpu]/FrameLockEnable=0 # - Disable the house sync signal on the master device nvidia-settings -a myserver:0[framelock:0]/FrameLockUseHouseSync=0 # - Query the enabled displays on the GPUs nvidia-settings -c myserver:0 -q gpus nvidia-settings -c myslave1:0 -q gpus # - Check the refresh rate is the same for all displays nvidia-settings -q myserver:0[dpy]/RefreshRate nvidia-settings -q myslave1:0[dpy]/RefreshRate # - Query the valid Frame Lock configurations for the display devices nvidia-settings -q myserver:0[dpy]/FrameLockDisplayConfig nvidia-settings -q myslave1:0[dpy]/FrameLockDisplayConfig # - Set the server display device nvidia-settings -a myserver:0[dpy:DVI-I-0]/FrameLockDisplayConfig=2 # - Set the slave display devices nvidia-settings -a myserver:0[dpy:DVI-I-1]/FrameLockDisplayConfig=1 nvidia-settings -a myslave1:0[dpy]/FrameLockDisplayConfig=1 # - Enable Frame Lock on server nvidia-settings -a myserver:0[gpu:0]/FrameLockEnable=1 # - Enable Frame Lock on slave devices nvidia-settings -a myserver:0[gpu:1]/FrameLockEnable=1 nvidia-settings -a myslave1:0[gpu]/FrameLockEnable=1 # - Toggle the test signal (on the master GPU) nvidia-settings -a myserver:0[gpu:0]/FrameLockTestSignal=1 nvidia-settings -a myserver:0[gpu:0]/FrameLockTestSignal=0 3. 1 System, 4 GPUs, 2 Frame Lock boards and 2 display devices per GPU syncing from the first GPU's display device: # - Make sure Frame Lock sync is disabled nvidia-settings -a [gpu]/FrameLockEnable=0 # - Disable the house sync signal on the master device nvidia-settings -a [framelock:0]/FrameLockUseHouseSync=0 # - Query the enabled displays on the GPUs nvidia-settings -V all -q gpus # - Check the refresh rate is the same for all displays nvidia-settings -q [dpy]/RefreshRate # - Query the valid Frame Lock configurations for the display devices nvidia-settings -q [dpy]/FrameLockDisplayConfig # - Set the master display device nvidia-settings -a [gpu:0.dpy:DVI-I-0]/FrameLockDisplayConfig=2 # - Set the slave display devices nvidia-settings -a [gpu:0.dpy:DVI-I-1]/FrameLockDisplayConfig=1 nvidia-settings -a [gpu:1.dpy]/FrameLockDisplayConfig=1 nvidia-settings -a [gpu:2.dpy]/FrameLockDisplayConfig=1 nvidia-settings -a [gpu:3.dpy]/FrameLockDisplayConfig=1 # - Enable Frame Lock on master GPU nvidia-settings -a [gpu:0]/FrameLockEnable=1 # - Enable Frame Lock on slave devices nvidia-settings -a [gpu:1]/FrameLockEnable=1 nvidia-settings -a [gpu:2]/FrameLockEnable=1 nvidia-settings -a [gpu:3]/FrameLockEnable=1 # - Toggle the test signal nvidia-settings -a [gpu:0]/FrameLockTestSignal=1 nvidia-settings -a [gpu:0]/FrameLockTestSignal=0 32J. LEVERAGING FRAME LOCK/GENLOCK IN OPENGL With the GLX_NV_swap_group extension, OpenGL applications can be implemented to join a group of applications within a system for local swap sync, and bind the group to a barrier for swap sync across a Frame Lock group. A universal frame counter is also provided to promote synchronization across applications. 32K. FRAME LOCK RESTRICTIONS: The following restrictions must be met for enabling Frame Lock: 1. All display devices set as client in a Frame Lock group must have the same mode timings as the server (master) display device. If a House Sync signal is used (instead of internal timings), all client display devices must be set to have the same refresh rate as the incoming house sync signal. 2. All X Screens (driving the selected client/server display devices) must have the same stereo setting. See the Stereo X configuration option for instructions on how to set the stereo X option. 3. In configurations with more than one display device per GPU, we recommend enabling Frame Lock on all display devices on those GPUs. 4. Virtual terminal switching or mode switching will disable Frame Lock on the display device. Note that the glXQueryFrameCountNV entry point (provided by the GLX_NV_swap_group extension) will only provide incrementing numbers while Frame Lock is enabled. Therefore, applications that use glXQueryFrameCountNV to control animation will appear to stop animating while Frame Lock is disabled. 32L. SUPPORTED FRAME LOCK CONFIGURATIONS: The following configurations are currently supported: 1. Basic Frame Lock: Single GPU, Single X Screen, Single Display Device with or without OpenGL applications that make use of Quad-Buffered Stereo and/or the GLX_NV_swap_group extension. 2. Frame Lock + TwinView: Single GPU, Single X Screen, Multiple Display Devices with or without OpenGL applications that make use of Quad-Buffered Stereo and/or the GLX_NV_swap_group extension. 3. Frame Lock + Xinerama: 1 or more GPU(s), Multiple X Screens, Multiple Display Devices with or without OpenGL applications that make use of Quad-Buffered Stereo and/or the GLX_NV_swap_group extension. 4. Frame Lock + TwinView + Xinerama: 1 or more GPU(s), Multiple X Screens, Multiple Display Devices with or without OpenGL applications that make use of Quad-Buffered Stereo and/or the GLX_NV_swap_group extension. ______________________________________________________________________________ Chapter 33. Configuring Depth 30 Displays ______________________________________________________________________________ This driver release supports X screens with screen depths of 30 bits per pixel (10 bits per color component). This provides about 1 billion possible colors, allowing for higher color precision and smoother gradients. When displaying a depth 30 image, the color data may be dithered to lower bit depths, depending on the capabilities of the display device and how it is connected to the GPU. Some devices connected via analog VGA, DisplayPort, or HDMI can display the full 10 bit range of colors. Devices connected via DVI, as well as laptop internal panels connected via LVDS, will be dithered to 8 or 6 bits per pixel. To work reliably, depth 30 requires pixman 0.11.6 or higher. In addition to the above software requirements, many X applications and toolkits do not understand depth 30 visuals as of this writing. Some programs may work correctly, some may work but display incorrect colors, and some may simply fail to run. In particular, many OpenGL applications request 8 bits of alpha when searching for FBConfigs. Since depth 30 visuals have only 2 bits of alpha, no suitable FBConfigs will be found and such applications will fail to start. ______________________________________________________________________________ Chapter 34. Offloading Graphics Display with RandR 1.4 ______________________________________________________________________________ Version 1.4 of the X Resize, Rotate, and Reflect Extension (RandR 1.4 for short) adds a way for drivers to work together so that one graphics device can display images rendered by another. 34A. SYSTEM REQUIREMENTS o For displaying NVIDIA GPU desktop contents on a screen connected to another graphics device, X.Org X server version 1.13 or higher. o For displaying another graphics device's desktop contents on a screen connected to an NVIDIA GPU, X.Org X server version 1.20.7 or higher. X server version 1.20.6 is also supported using Appendix B. o A Linux kernel, version 3.13 or higher, with CONFIG_DRM enabled. o Version 1.4.0 of the xrandr command-line utility. 34B. USING THE NVIDIA DRIVER AS AN RANDR 1.4 OUTPUT SOURCE OR OUTPUT SINK PROVIDER To use the NVIDIA driver as an RandR 1.4 output source provider, also known as "PRIME", the X server needs to be configured to use the NVIDIA driver for its primary screen and to use the "modesetting" driver for the other graphics device. This can be achieved by placing the following in "/etc/X11/xorg.conf": Section "ServerLayout" Identifier "layout" Screen 0 "nvidia" Inactive "intel" EndSection Section "Device" Identifier "nvidia" Driver "nvidia" BusID "" EndSection Section "Screen" Identifier "nvidia" Device "nvidia" Option "AllowEmptyInitialConfiguration" EndSection Section "Device" Identifier "intel" Driver "modesetting" EndSection Section "Screen" Identifier "intel" Device "intel" EndSection To use the NVIDIA driver as an RandR 1.4 output sink provider, also known as "Reverse PRIME", the X server needs to be configured to use the "modesetting" driver for its primary screen and to use the NVIDIA driver for the other graphics device. This can be achieved by placing the following in "/etc/X11/xorg.conf": Section "ServerLayout" Identifier "layout" Screen 0 "intel" Inactive "nvidia" Option "AllowNVIDIAGPUScreens" EndSection Section "Device" Identifier "intel" Driver "modesetting" BusID "" EndSection Section "Screen" Identifier "intel" Device "intel" EndSection Section "Device" Identifier "nvidia" Driver "nvidia" EndSection Section "Screen" Identifier "nvidia" Device "nvidia" EndSection When using the NVIDIA driver as a "Reverse PRIME" RandR 1.4 output sink provider combined with an application being run via PRIME Render Offload, an optimization known as "Reverse PRIME Bypass" may be used, bypassing the bandwidth overhead of both PRIME Render Offload and PRIME Display Offload. In order for Reverse PRIME Bypass to be used, a PRIME Render Offload application must be unredirected, fullscreen, and visible only on a single NVIDIA-driven PRIME Display Offload output. Usage of Reverse PRIME Bypass is printed to the X log when verbose logging is enabled in the X server. If an NVIDIA Reverse PRIME output is the sole display in the system then special conditions apply. That configuration is supported if the NVIDIA driver version is 495.46 or later and the X.Org X server version is newer than 21.1.3. Failing these conditions the NVIDIA Reverse PRIME output will not be synchronized to the native refresh rate of the NVIDIA graphics card in which case X.Org will revert the display to a default rate of 1 frame per second. Note that at the time of writing the latest X.Org X server is 21.1.3 so there is no official X.Org release yet where this configuration is supported. For maintainers of Linux distributions and others who are willing to compile the X.Org X server locally, please cherry-pick this Git commit to support the configuration: https://gitlab.freedesktop.org/xorg/xserver/-/commit/69774044716039fa70655b3bc6dd6a4ff4535cfd.The commit already lives in the branch where the next X.Org X server release after 21.1.3 will come from. See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on determining the appropriate BusID string for your graphics card. The nvidia-xconfig(1) utility can be used to update the X configuration file for using the NVIDIA driver as an output source provider. $ nvidia-xconfig --prime See the nvidia-xconfig(1) man page for details. The X server does not automatically enable displays attached using the output sink in this configuration. To do that, use the "xrandr" command line tool. For NVIDIA as an output source: $ xrandr --setprovideroutputsource modesetting NVIDIA-0 $ xrandr --auto For NVIDIA as an output sink: $ xrandr --setprovideroutputsource NVIDIA-G0 modesetting $ xrandr --auto This pair of commands can be added to your X session startup scripts, for example by putting them in "$HOME/.xinitrc" before running "startx". Use the $ xrandr --listproviders command to query the capabilities of the graphics devices. If the system requirements are met and the X server is configured correctly, there should be a provider named "NVIDIA-0" or "NVIDIA-G0" with the "Source Output" or "Sink Output" capability, respectively, and one named "modesetting" with the "Sink Output" and/or "Source Output" capabilities. If either provider is missing or doesn't have the expected capability, check your system configuration. 34C. SYNCHRONIZED RANDR 1.4 OUTPUTS When running against X.Org X server with video driver ABI 23 or higher, synchronization is supported with compatible drivers. At the time of writing, synchronization is compatible with the "modesetting" driver with Intel devices on Linux version 4.5 or newer. If all requirements are met, synchronization will be used automatically. X.Org X server version 1.19 or newer is required to support synchronization. Without synchronization, displays are prone to "tearing". See Caveats for details. If synchronization is being used but is not desired, it can be disabled with: $ xrandr --output --set "PRIME Synchronization" 0 and re-enabled with: $ xrandr --output --set "PRIME Synchronization" 1 See Vblank syncing for information on how OpenGL applications can synchronize with sink-provided outputs. 34D. CAVEATS o Support for PRIME Synchronization relies on DRM KMS support. See Chapter 36 for more information. o Some Intel i915 DRM driver versions, such as that included with Linux 4.5, have a bug where drmModeMoveCursor() and drmModePageFlip() interfere with each other, resulting in only one occurring per frame. If choppy performance is observed in configurations using PRIME Synchronization and i915, it is suggested to add "Option "SWCursor"" to Intel's device section in xorg.conf. The bug appears to be fixed as of Linux 4.6. o When running against X.Org X server version 1.18.x or lower, there is no synchronization between the images rendered by the NVIDIA GPU and the output device. This means that the output device can start reading the next frame of video while it is still being updated, producing a graphical artifact known as "tearing". Tearing is expected due to limitations in the design of the X.Org X server prior to video driver ABI 23. o NVIDIA's implementation of PRIME requires support for DRM render nodes, a feature first merged in Linux 3.12. However, the feature was not enabled by default until Linux 3.17. To enable it on earlier supported kernels, specify the "drm.rnodes=1" kernel boot parameter. o PRIME Synchronization is compatible with xf86-video-amdgpu as an output sink. xf86-video-amdgpu implements a separate interface for PRIME Synchronization that the RandR layer of the X server does not recognize. As a result, X will print "randr: falling back to unsynchronized pixmap sharing", despite the fact that PRIME is synchronized. Additionally, the "PRIME Synchronization" output property will not function to disable PRIME Synchronization when set to 0. o The NVIDIA driver only exposes the "Output Sink" capability by default on X server version 1.20.7 or later, but can be used without issue on X server version 1.20.6 with "Option "AllowPRIMEDisplayOffloadSink"". See Appendix B for more information. o The NVIDIA driver requires DRM KMS support to operate as an output sink when the output source driver is either NVIDIA or AMDGPU. See Chapter 36 for more information. ______________________________________________________________________________ Chapter 35. PRIME Render Offload ______________________________________________________________________________ PRIME render offload is the ability to have an X screen rendered by one GPU, but choose certain applications within that X screen to be rendered on a different GPU. This is particularly useful in combination with dynamic power management to leave an NVIDIA GPU powered off except when it is needed to render select performance-sensitive applications. The GPU rendering the majority of the X screen is known as the "sink", and the GPU to which certain application rendering is "offloaded" is known as the "source". The render offload source produces content that is presented on the render offload sink. The NVIDIA driver can function as a PRIME render offload source, to offload rendering of GLX+OpenGL or Vulkan, presenting to an X screen driven by the xf86-video-modesetting X driver. 35A. X SERVER REQUIREMENTS NVIDIA's PRIME render offload support requires X.Org xserver version 1.20.7 or newer. 35B. CONFIGURE THE X SERVER On systems with both an integrated GPU and an NVIDIA discrete GPU, the X.Org X server version 1.20.7 and newer will automatically use NVIDIA's PRIME render offload support if the system BIOS is configured to boot on the iGPU and no other explicit configuration files are present. Note that some Linux distributions (such as Ubuntu) may configure the X server differently. Please refer to your distribution's documentation for details. If GPU screen creation was successful, the log file '/var/log/Xorg.0.log' should contain lines with "NVIDIA(G0)", and querying the RandR providers with "xrandr --listproviders" should display a provider named "NVIDIA-G0" (for "NVIDIA GPU screen 0"). For example: Providers: number : 2 Provider 0: id: 0x221 cap: 0x9, Source Output, Sink Offload crtcs: 3 outputs: 6 associated providers: 0 name:modesetting Provider 1: id: 0x1f8 cap: 0x0 crtcs: 0 outputs: 0 associated providers: 0 name:NVIDIA-G0 35C. CONFIGURE GRAPHICS APPLICATIONS TO RENDER USING THE GPU SCREEN To configure a graphics application to be offloaded to the NVIDIA GPU screen, set the environment variable "__NV_PRIME_RENDER_OFFLOAD" to "1". If the graphics application uses Vulkan or EGL, that should be all that is needed. If the graphics application uses GLX, then also set the environment variable "__GLX_VENDOR_LIBRARY_NAME" to "nvidia", so that GLVND loads the NVIDIA GLX driver. Examples: __NV_PRIME_RENDER_OFFLOAD=1 vkcube __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo | grep vendor 35D. FINER-GRAINED CONTROL OF VULKAN The "__NV_PRIME_RENDER_OFFLOAD" environment variable causes the special Vulkan layer "VK_LAYER_NV_optimus" to be loaded. Vulkan applications use the Vulkan API to enumerate the GPUs in the system and select which GPU to use; most Vulkan applications will use the first GPU reported by Vulkan. Newer Vulkan loaders can sort the GPU enumeration to put the preferred GPUs first, based on platform defined criteria. The "VK_LAYER_NV_optimus" layer causes the GPUs to be sorted such that the NVIDIA GPUs are enumerated first if the loader does not do its own sorting. For finer-grained control, the "VK_LAYER_NV_optimus" layer looks at the "__VK_LAYER_NV_optimus" environment variable. The value "NVIDIA_only" causes "VK_LAYER_NV_optimus" to always sort the GPUs so NVIDIA GPUs are enumerated first for the Vulkan application, overriding any Vulkan loader sorting. The value "non_NVIDIA_only" causes "VK_LAYER_NV_optimus" to always sort the GPUs so non-NVIDIA GPUs are enumerated first for to the Vulkan application, overriding any Vulkan loader sorting. Note that the "VK_LAYER_NV_optimus" layer affects the ordering of discrete and integrated GPUs but won't touch CPU devices. Examples: __NV_PRIME_RENDER_OFFLOAD=1 __VK_LAYER_NV_optimus=NVIDIA_only vkcube __NV_PRIME_RENDER_OFFLOAD=1 __VK_LAYER_NV_optimus=non_NVIDIA_only vkcube 35E. FINER-GRAINED CONTROL OF OPENGL For OpenGL with either GLX or EGL, the environment variable "__NV_PRIME_RENDER_OFFLOAD_PROVIDER" provides finer-grained control. While "__NV_PRIME_RENDER_OFFLOAD=1" tells GLX or EGL to use the first NVIDIA GPU screen, "__NV_PRIME_RENDER_OFFLOAD_PROVIDER" can use an RandR provider name to pick a specific NVIDIA GPU screen, using the NVIDIA GPU screen names reported by "`xrandr --listproviders`". Examples: __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears __NV_PRIME_RENDER_OFFLOAD_PROVIDER=NVIDIA-G0 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears __NV_PRIME_RENDER_OFFLOAD=1 eglinfo __NV_PRIME_RENDER_OFFLOAD_PROVIDER=NVIDIA-G0 eglinfo 35F. TROUBLESHOOTING After starting the X server, verify that the xf86-video-modesetting X driver is using "glamoregl". The log file '/var/log/Xorg.0.log' should contain something like this: [1272173.618] (II) Loading sub module "glamoregl" [1272173.618] (II) LoadModule: "glamoregl" [1272173.618] (II) Loading /usr/lib/xorg/modules/libglamoregl.so [1272173.622] (II) Module glamoregl: vendor="X.Org Foundation" [1272173.622] compiled for 1.20.4, module version = 1.0.1 [1272173.622] ABI class: X.Org ANSI C Emulation, version 0.4 [1272173.638] (II) modeset(0): glamor X acceleration enabled on Mesa DRI Intel(R) HD Graphics 630 (Kaby Lake GT2) [1272173.638] (II) modeset(0): glamor initialized If glamoregl could not be loaded, the X log may report something like: [1271802.673] (II) Loading sub module "glamoregl" [1271802.673] (II) LoadModule: "glamoregl" [1271802.673] (WW) Warning, couldn't open module glamoregl [1271802.673] (EE) modeset: Failed to load module "glamoregl" (module does not exist, 0) [1271802.673] (EE) modeset(0): Failed to load glamor module. in which case, consult your distribution's documentation for how to (re-)install the package containing glamoregl. If the server didn't create a GPU screen automatically, ensure that the nvidia_drm kernel module is loaded. This should normally happen by default, but you can confirm by running "lsmod | grep nvidia_drm" to see if the kernel module is loaded. Run "modprobe nvidia_drm" to load it. If automatic configuration does not work, it may be necessary to explicitly configure the iGPU and dGPU devices in xorg.conf: Section "ServerLayout" Identifier "layout" Screen 0 "iGPU" EndSection Section "Device" Identifier "iGPU" Driver "modesetting" EndSection Section "Screen" Identifier "iGPU" Device "iGPU" EndSection Section "Device" Identifier "dGPU" Driver "nvidia" EndSection ______________________________________________________________________________ Chapter 36. Direct Rendering Manager Kernel Modesetting (DRM KMS) ______________________________________________________________________________ The NVIDIA GPU driver package provides a kernel module, 'nvidia-drm.ko', which registers a DRM driver with the DRM subsystem of the Linux kernel. The capabilities advertised by this DRM driver depend on the Linux kernel version and configuration: o PRIME: This is needed to support graphics display offload in RandR 1.4. Linux kernel version 3.13 or higher is required, with CONFIG_DRM enabled. o Atomic Modeset: This is used for display of non-X11 based desktop environments, such as Wayland and Mir. Linux kernel version 4.1 or higher is required, with CONFIG_DRM and CONFIG_DRM_KMS_HELPER enabled. NVIDIA's DRM KMS support is still considered experimental. It is disabled by default, but can be enabled on suitable kernels with the 'modeset' kernel module parameter. E.g., modprobe -r nvidia_drm ; modprobe nvidia_drm modeset=1 Applications can present through NVIDIA's DRM KMS implementation using any of the following: o The DRM KMS "dumb buffer" mechanism to create and map CPU-accessible buffers: DRM_IOCTL_MODE_CREATE_DUMB, DRM_IOCTL_MODE_MAP_DUMB, and DRM_IOCTL_MODE_DESTROY_DUMB. o Using the EGL_EXT_device_drm, EGL_EXT_output_drm, and EGL_EXT_stream_consumer_egloutput EGL extensions to associate EGLStream producers with specific DRM KMS planes. 36A. KNOWN ISSUES o The NVIDIA DRM KMS implementation is currently incompatible with SLI Mosaic. The X server will fail to initialize SLI Mosaic if DRM KMS is enabled. o The NVIDIA DRM KMS implementation does not yet register an overlay plane: only primary and cursor planes are currently provided. o Buffer allocation and submission to DRM KMS using gbm is not currently supported. ______________________________________________________________________________ Chapter 37. Configuring External and Removable GPUs ______________________________________________________________________________ This driver release supports the use of external GPUs, or eGPUs, such as those commonly connected via Thunderbolt. However, system stability when an eGPU is unplugged while in use (also known as "hot-unplug") is not guaranteed. This includes situations where the eGPU is being used to display the X11 desktop. External GPUs are often used in short-running compute scenarios, which better tolerate the eGPU being hot-unplugged. In such cases, a different GPU may be used to display the X11 desktop. To prevent system instability from hot-unplugging an eGPU while being used to display the X11 desktop, the NVIDIA X driver does not configure X screens on external GPUs by default. This behavior may also apply to GPUs attached to internal PCIe slots with hot-unplug support, such as in some enterprise systems. To override this behavior in xorg.conf, see Appendix B. Then, external GPUs may be configured with X as one would any other secondary GPU, by specifying the BusID in the Device section in xorg.conf. See the xorg.conf man page for more information on the BusID option, and "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine the BusID. ______________________________________________________________________________ Chapter 38. NGX ______________________________________________________________________________ NGX is a framework allowing for easy and consistent integration of Deep Learning features into applications. There are three components shipped with the driver that comprise NGX: o NGX Core: The interface used by applications to communicate with NGX features. o NGX Updater: A program optionally launched by NGX Core which downloads updated versions of NGX features. o NGX Core for Proton: A copy of NGX Core built as a DLL for use with Proton. 38A. CONFIGURING THE NGX UPDATER: NGX Configuration file search order: 1. Path supplied via "__NGX_CONF_FILE" environment variable. Note: if the application's effective UID/GID does not match its actual UID/GID (if the SETUID sticky bit is set), the location specified here will not be searched. 2. "${XDG_CONFIG_HOME}/nvidia-ngx-conf.json". Note: if the application's effective UID/GID does not match its actual UID/GID (if the SETUID sticky bit is set), the location specified here will not be searched. 3. "/usr/share/nvidia/nvidia-ngx-conf.json" JSON Config Options: Key Description Example --------------------- --------------------- --------------------- "file_format_version" The version of the "1.0.0" config file in the following format "major.minor.patch". Values should be of type "string". "ngx_models_path" The absolute path to "/usr/share/nvidia/ngx" the NGX models directory. The NGX Updater will save files here to later be read by applications which make use of NGX functionality. Values should be of type "string". "allow_ngx_updater" A boolean value which "false" enables the NGX Updater functionality on a system, defaulting to false if non-existent. Values should be of type "boolean". For systems that have multiple user accounts who may run applications which make use of NGX, it may be preferable to choose a ngx_models_path which is in a shared directory. If a shared directory is chosen it is important that the NGX Updater has permissions to write into it, this can be attained either through use of setuid/setgid sticky bit, or by altering the directories permissions accordingly. 38B. CONFIGURING THE NGX SIGNED LOAD MECHANISM: The NGX Core component of the driver performs a signature check when loading NGX features. This check can be disabled by setting the "__NV_SIGNED_LOAD_CHECK" environment variable to a value of "none". Values other than "none" will be ignored. By default this check is enabled to prevent loading of potentially malicious non-NVIDIA provided NGX features. This check should only be disabled when there is confidence that only authentic NGX features are available on the current system. 38C. CONFIGURING THE NGX UPDATER FOR WINE/PROTON: The NGX Updater for Wine/Proton can be enabled by setting the "PROTON_ENABLE_NGX_UPDATER" environment variable to a value of "1". Values other than "1" will be ignored. ______________________________________________________________________________ Chapter 39. OpenGL, Vulkan and VDPAU on Xwayland ______________________________________________________________________________ 39A. OVERVIEW Xwayland is an X11 server implemented as a Wayland client, allowing one to run X11 applications on a Wayland desktop in a relatively seamless fashion. The NVIDIA driver is able to facilitate accelerated 3D rendering for such applications, but there are some particular considerations compared to a typical X11 configuration which are outlined in this chapter. Note that some NVIDIA driver features may not be available, see Appendix L for details. 39B. REQUIREMENTS The following are necessary to enable accelerated rendering on Xwayland with the NVIDIA driver: o DRM KMS must be enabled. See Chapter 36 for details. o The installed copy of Xwayland should be a build from the master branch of https://gitlab.freedesktop.org/xorg/xserver at least as recent as commit c468d34c. Note that if this requirement is not satisfied, the NVIDIA GPU can still be used for rendering, however it will fall back to a suboptimal path for presentation resulting in degraded performance. o libxcb version 1.13 or later must be present. o egl-wayland version 1.1.7 or later must be present (if installed separately from the the NVIDIA driver). o If using the GNOME desktop environment, kms-modifiers must be enabled through gsettings. This can be done with the following command `gsettings set org.gnome.mutter experimental-features [\"kms-modifiers\"]` 39C. REPORTING ISSUES The Wayland desktop relies on the close interaction of several components to function correctly. Because of this, it can be tricky to determine exactly which one is responsible for a suspected bug. o If the issue only reproduces on NVIDIA hardware with the proprietary driver, see Chapter 45 for links to the appropriate channels to report the bug to us. o If the issue only reproduces with a particular compositor, it can be reported on that project's bug tracker, for example: o https://gitlab.gnome.org/GNOME/mutter/-/issues/ for GNOME o https://bugs.kde.org/ for KDE o If the issue is specific to X11 applications running on Xwayland, it can be reported on the X.org xserver issue tracker https://gitlab.freedesktop.org/xorg/xserver/-/issues. ______________________________________________________________________________ Chapter 40. GBM and GBM-based Wayland Compositors ______________________________________________________________________________ 40A. OVERVIEW GBM is a memory allocation API developed as a component of the Mesa project. Most Wayland compositors use the GBM API to initialize an EGLDisplay object directly on a GPU, and to allocate the EGLSurface representing the desktop. The NVIDIA driver includes a GBM backend enabling the use of such software on NVIDIA GPUs. A GBM EGL external platform library is also included to enable the use of GBM objects in EGL. 40B. REQUIREMENTS The following are necessary to enable use of the NVIDIA driver's GBM backend and its EGL and related Wayland client support: o DRM KMS must be enabled. See Chapter 36 for details. o libgbm.so.1 from Mesa version 21.2 (https://gitlab.freedesktop.org/mesa/mesa). o egl-wayland version 1.1.8 or later must be present (if installed separately from the the NVIDIA driver). To support applications that use the wl_drm Wayland protocol to select a GPU to render on, such as Xwayland, version 1.1.9 or later must be present. 40C. XWAYLAND AND GBM Xwayland uses the GLAMOR backend to accelerate X application rendering using EGL. While not required for basic functionality, GLAMOR acceleration is required for good performance when running accelerated Vulkan or OpenGL X11 applications. The GLAMOR integration code in Xwayland has paths to use either EGLStreams or GBM to allocate surfaces for use with accelerated rendering. To use the GBM path with the NVIDIA driver's GBM backend and GBM EGL external platform library, the installed copy of Xwayland should be a build from the master branch of https://gitlab.freedesktop.org/xorg/xserver at least as recent as commit 5daf42b4. Xwayland versions 21.1.3 and above satisfy this requirement. See Chapter 39 for more details on Xwayland support in the NVIDIA driver. 40D. LOGGING/TRACING The NVIDIA GBM backend supports logging information about its behavior and the reasons for various function call failures to stderr. Such information may be useful for developers debugging applications or end users diagnosing system behavior issues. Set the environment variable '__NV_GBM_TRACE_ENABLED' to a non-zero integer value to enable this logging. ______________________________________________________________________________ Chapter 41. Addressing Capabilities ______________________________________________________________________________ Many PCIe devices have limitations in what memory addresses they can access for DMA purposes (based on the number of lines dedicated to memory addressing). This can cause problems if the host system has memory mapped to addresses beyond what the PCIe device can support. If a PCIe device is allocated memory at an address beyond what the device can support, the address may be truncated and the device will access the incorrect memory location. Note that since certain system resources, such as ACPI tables and PCI I/O regions, are mapped to address ranges below the 4 GB boundary, the RAM installed in x86/x86-64 systems cannot necessarily be mapped contiguously. Similarly, system firmware is free to map the available RAM at its or its users' discretion. As a result, it is common for systems to have RAM mapped outside of the address range [0, RAM_SIZE], where RAM_SIZE is the amount of RAM installed in the system. For example, it is common for a system with 512 GB of RAM installed to have physical addresses up to ~513 GB. In this scenario, a GPU with an addressing capability of 512 GB would force the driver to fall back to the 4 GB DMA zone for this GPU. The NVIDIA Linux driver attempts to identify the scenario where the host system has more memory than a given GPU can address. If this scenario is detected, the NVIDIA driver will drop back to allocations from the 4 GB DMA zone to avoid address truncation. This means that the driver will use the __GFP_DMA32 flag and limit itself to memory addresses below the 4 GB boundary. This is done on a per-GPU basis, so limiting one GPU will not limit other GPUs in the system. The addressing capabilities of an NVIDIA GPU can be queried at runtime via the procfs interface: % cat /proc/driver/nvidia/gpus/PCI-BUS-ID/information ... DMA Size: 40 bits DMA Mask: 0xffffffffff ... See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine the PCI-BUS-ID. The memory mapping of RAM on a given system can be seen in the BIOS-e820 table printed out by the kernel and available via `dmesg`. Note that the 'usable' ranges are actual RAM: [ 0.000000] BIOS-provided physical RAM map: [ 0.000000] BIOS-e820: 0000000000000000 - 000000000009f000 (usable) [ 0.000000] BIOS-e820: 000000000009f000 - 00000000000a0000 (reserved) [ 0.000000] BIOS-e820: 0000000000100000 - 000000003fe5a800 (usable) [ 0.000000] BIOS-e820: 000000003fe5a800 - 0000000040000000 (reserved) 41A. SOLUTIONS There are multiple potential ways to solve a discrepancy between your system configuration and a GPU's addressing capabilities. 1. Select a GPU with addressing capabilities that match your target configuration. The best way to achieve optimal system and GPU performance is to make sure that the capabilities of the two are in alignment. This is especially important with multiple GPUs in the system, as the GPUs may have different addressing capabilities. In this multiple GPU scenario, other solutions could needlessly impact the GPU that has larger addressing capabilities. 2. Configure the system's IOMMU to the GPU's addressing capabilities. This is a solution targeted at developers and system builders. The use of IOMMU may be an option, depending on system configuration and IOMMU capabilities. Please contact NVIDIA to discuss solutions for specific configurations. 3. Limit the amount of memory seen by the Operating System to match your GPU's addressing capabilities with kernel configuration. This is best used in the scenario where RAM is mapped to addresses that slightly exceeds a GPU's capabilities and other solutions are either not achievable or more intrusive. A good example is the 512 GB RAM scenario outlined above with a GPU capable of addressing 512 GB. The kernel parameter can be used to ignore the RAM mapped above 512 GB. This can be achieved in Linux by use of the "mem" kernel parameter. See the kernel-parameters.txt documentation for more details on this parameter. This solution does affect the entire system and will limit how much memory the OS and other devices can use. In scenarios where there is a large discrepancy between the system configuration and GPU capabilities, this is not a desirable solution. 4. Remove RAM from the system to align with the GPU's addressing capabilities. This is the most heavy-handed, but may ultimately be the most reliable solution. ______________________________________________________________________________ Chapter 42. GPUDirect RDMA Peer Memory Client ______________________________________________________________________________ 42A. BACKGROUND GPUDirect RDMA (Remote Direct Memory Access) is a technology that enables a direct path for data exchange between the GPU and a third-party peer device using standard features of PCI Express. The NVIDIA GPU driver package provides a kernel module, 'nvidia-peermem.ko', which provides Mellanox InfiniBand based HCAs (Host Channel Adapters) direct peer-to-peer read and write access to the NVIDIA GPU's video memory. It allows GPUDirect RDMA-based applications to use GPU computing power with the RDMA interconnect without needing to copy data to host memory. This capability is supported with Mellanox ConnectX-3 VPI or newer adapters. It works with both InfiniBand and RoCE (RDMA over Converged Ethernet) technologies. Mellanox OFED (Open Fabrics Enterprise Distribution) or MOFED, introduces an API between the InfiniBand Core and peer memory clients such as NVIDIA GPUs, called PeerDirect, see https://community.mellanox.com/s/article/howto-implement-peerdirect-client-using-mlnx-ofed. The 'nvidia-peermem.ko' module registers the NVIDIA GPU with the InfiniBand subsystem by using peer-to-peer APIs provided by the NVIDIA GPU driver. This module, originally maintained by Mellanox on GitHub, is now included with the NVIDIA Linux GPU driver. The original GitHub project at https://github.com/Mellanox/nv_peer_memory should be considered deprecated and only critical bugs will be addressed for existing installations. 42B. USAGE The kernel must have the required support for RDMA peer memory either through additional patches to the kernel or via Mellanox OFED package (MOFED) as a prerequisite for loading and using 'nvidia-peermem.ko'. It is possible that the 'nv_peer_mem' module from the GitHub project may be installed and loaded on the system. Installation of 'nvidia-peermem.ko' will not affect the functionality of the existing 'nv_peer_mem' module. But, to load and use 'nvidia-peermem.ko', users must disable the 'nv_peer_mem' service. Additionally, it is encouraged to uninstall the 'nv_peer_mem' package to avoid any conflict with 'nvidia-peermem.ko' since only one module can be loaded at any time. Stop the 'nv_peer_mem' service: # service nv_peer_mem stop Check if 'nv_peer_mem.ko' is still loaded after stopping the service: # lsmod | grep nv_peer_mem If 'nv_peer_mem.ko' is still loaded, unload it with: # rmmod nv_peer_mem Uninstall 'nv_peer_mem' package: For DEB based OS: # dpkg -P nvidia-peer-memory # dpkg -P nvidia-peer-memory-dkms For RPM based OS: # rpm -e nvidia_peer_memory After ensuring kernel support and installing the GPU driver, 'nvidia-peermem.ko' can be loaded with the following command with root privileges in a terminal window: # modprobe nvidia-peermem Note: If the NVIDIA GPU driver is installed before MOFED, the GPU driver must be uninstalled and installed again to make sure 'nvidia-peermem.ko' is compiled with the RDMA APIs that are provided by MOFED. 42C. MODULE PARAMETERS "peerdirect_support": this parameter takes the following integer values: o 0, which is the default and is appropriate for a kernel that has the PeerDirect APIs roughly corresponding to MOFED 5.1. o 1, which is required in combination with the legacy PeerDirect APIs, as currently shipping in MOFED 5.0 and older releases, notably in MOFED LTS. As a reference, in the legacy PeerDirect APIs, the 'peer_memory_client' structure declared in 'peer_mem.h' has the two extra function pointers shown below: void* (*get_context_private_data)(u64 peer_id); void (*put_context_private_data)(void *context); Note that MOFED LTS as well as MOFED 5.0 and previous releases ship with legacy PeerDirect APIs. So for example, when using MOFED LTS, GPUDirect RDMA support for the Mellanox HCAs will not work correctly unless "peerdirect_support" is set to one. Instead for MOFED 5.1 or newer, the default value of zero is appropriate, so no special actions are needed. 42D. KNOWN ISSUES o Currently, there is no service to automatically load 'nvidia-peermem.ko'. Users need to load the module manually. o When loading 'nvidia-peermem.ko' on a kernel with legacy PeerDirect APIs, the module parameter "peerdirect_support" has to be set to one. o The "PeerDirect" APIs shipping in MOFED releases 5.1 and later are affected by a lock inversion bug which may lead to a kernel-side deadlock. This is tracked by the NVIDIA-internal reference number "2696789". PeerDirect APIs in newer MOFED releases belonging to some branches, like "5.3-1.0.0.1.43", offer an opt-in feature to mitigate that problem. Starting from this release the 'nvidia-peermem.ko' kernel module explicitly enables it, unless "peerdirect_support" is set to one. ______________________________________________________________________________ Chapter 43. GSP Firmware ______________________________________________________________________________ Some GPUs include a GPU System Processor (GSP) which can be used to offload GPU initialization and management tasks. This processor is driven by firmware files distributed with the driver. The GSP firmware is used by default on GPUs which support it. Offloading tasks which were traditionally performed by the driver on the CPU can improve performance due to lower latency access to GPU hardware internals. Firmware files 'gsp_*.bin' are installed in '/lib/firmware/nvidia/570.123.06/'. Each GSP firmware file is named after a GPU architecture (for example, 'gsp_tu10x.bin' is named after Turing) and supports GPUs from one or more architectures. 43A. QUERY MODE The nvidia-smi utility can be used to query the current use of GSP firmware. It will display a valid version if GSP firmware is enabled, or "N/A" if disabled: $ nvidia-smi -q ... GSP Firmware Version : 570.123.06 ... This information is also present in the per-GPU information file in the '/proc' file system. $ cat /proc/driver/nvidia/gpus/PCI-BUS-ID/information ... GSP Firmware: 570.123.06 See "Q. What is the format of a PCI Bus ID?" in Chapter 7 for information on how to determine the PCI-BUS-ID. 43B. DISABLING GSP MODE The driver can be forced to disable use of GSP firmware by setting the kernel module parameter NVreg_EnableGpuFirmware=0. 43C. ENABLING GSP MODE The GSP firmware will be used by default for all Turing and later GPUs. The driver can be explicitly configured to use the GSP firmware by setting the kernel module parameter NVreg_EnableGpuFirmware=1. ______________________________________________________________________________ Chapter 44. Open Linux Kernel Modules ______________________________________________________________________________ 44A. INTRODUCTION The NVIDIA Linux GPU Driver contains several kernel modules: nvidia.ko, nvidia-modeset.ko, nvidia-uvm.ko, nvidia-drm.ko, and nvidia-peermem.ko. Two "flavors" of these kernel modules are provided: o Open, i.e. source-published, kernel modules that are dual licensed MIT/GPLv2. With every driver release, the source code to the open kernel modules is published on https://github.com/NVIDIA/open-gpu-kernel-modules and a tarball is provided on https://download.nvidia.com/XFree86/. o Proprietary. This is the flavor that NVIDIA has historically shipped. We recommend the use of open kernel modules on all GPUs that support it. Though the kernel modules in the two flavors are different, they are based on the same underlying source code. The two flavors are mutually exclusive: they cannot be used within the kernel at the same time, and they should not be installed on the filesystem at the same time. The user space components of the NVIDIA Linux GPU driver are identical and behave in the same way, regardless of which flavor of kernel module is used. 44B. SUPPORTED FEATURES The proprietary flavor supports the GPU architectures Maxwell, Pascal, Volta, Turing, and later GPUs until Blackwell. Blackwell and later are only supported by the open kernel modules. The open flavor of kernel modules supports Turing and later GPUs. The open kernel modules cannot support GPUs before Turing, because the open kernel modules depend on the GPU System Processor (GSP) first introduced in Turing. The following features will only work with the open kernel modules flavor of the driver: o NVIDIA Confidential Computing o Magnum IO GPUDirect Storage (GDS) o Heterogeneous Memory Management (HMM) o CPU affinity for GPU fault handlers o DMABUF support for CUDA allocations 44C. KNOWN ISSUES The following are some known limitations of the open kernel modules versus the proprietary kernel modules with GSP firmware mode disabled: o GPU initialization is slower. One possible mitigation is to use nvidia-persistenced to initialize the GPU(s) in advance, before running applications that use the GPU. o Enter and exit latencies for power-saving modes like S3, S4 and Run Time D3 (RTD3) can be longer due to additional GSP state being restored. o GPU power consumption can be marginally impacted in some scenarios. o Run Time D3 (RTD3) is only supported on Ampere and above GPUs. 44D. INSTALLATION Because the two flavors of kernel modules are mutually exclusive, one or the other must be chosen at install time. By default, installation will choose which flavor of kernel modules to install, based on the GPUs detected in the system. If a pre-Turing GPU is detected, installation will default to the proprietary flavor of kernel modules. Otherwise, installation will default to the open flavor of kernel modules. This default can be overridden with the "--kernel-module-type" .run file option, or its short form "-M". Use "-M=proprietary" to force installation of the proprietary flavor of kernel modules, or "-M=open" to force installation of the open flavor of kernel modules. E.g., sh ./NVIDIA-Linux-[...].run -m=kernel-open As a convenience, the open kernel modules distributed in the .run file are pre-compiled. Advanced users, who want to instead build the open kernel modules from source, should do the following: o Uninstall any existing driver with `nvidia-uninstall`. o Install from the .run file with "--no-kernel-modules" option, to install everything except the kernel modules. o Fetch, build, and install the open kernel module source from https://github.com/NVIDIA/open-gpu-kernel-modules. See https://github.com/NVIDIA/open-gpu-kernel-modules/blob/main/README.md for details on building. Note that you must use the same version of the .run file and the open kernel module source from https://github.com/NVIDIA/open-gpu-kernel-modules You can determine which flavor of kernel modules is installed using either `modinfo` or looking at /proc/driver/nvidia/version. E.g., the proprietary flavor will report: # modinfo nvidia | grep license license: NVIDIA # cat /proc/driver/nvidia/version NVRM version: NVIDIA UNIX x86_64 Kernel Module [...] The open flavor will report: # modinfo nvidia | grep license license: Dual MIT/GPL # cat /proc/driver/nvidia/version NVRM version: NVIDIA UNIX Open Kernel Module for x86_64 [...] ______________________________________________________________________________ Chapter 45. NVIDIA Contact Info and Additional Resources ______________________________________________________________________________ 45A. GENERAL SUPPORT There is an NVIDIA Linux Driver web forum. You can access it by going to http://forums.developer.nvidia.com and following the "Linux" link in the "GPU Unix Graphics" section. This is the preferable tool for seeking help; users can post questions, answer other users' questions, and search the archives of previous postings. If all else fails, you can contact NVIDIA for support at: linux-bugs@nvidia.com. But please, only send email to this address after you have explored the Chapter 7 and Chapter 8 chapters of this document, and asked for help on the forums.developer.nvidia.com web forum. When emailing linux-bugs@nvidia.com, please include the 'nvidia-bug-report.log.gz' file generated by the 'nvidia-bug-report.sh' script (which is installed as part of driver installation), along with a detailed description of your problem. 45B. SECURITY ISSUES To report a potential security vulnerability in any NVIDIA product, please use either: o this web form: https://www.nvidia.com/object/submit-security-vulnerability.html, or o send email to: psirt@nvidia.com If reporting a potential vulnerability via email, please encrypt it using NVIDIA's public PGP key (see https://www.nvidia.com/en-us/security/pgp-key/) and include the following information: o Product/Driver name and version/branch that contains the vulnerability o Type of vulnerability (code execution, denial of service, buffer overflow, etc.) o Instructions to reproduce the vulnerability o Proof-of-concept or exploit code o Potential impact of the vulnerability, including how an attacker could exploit the vulnerability 45C. ADDITIONAL RESOURCES Linux OpenGL ABI http://www.opengl.org/registry/ABI/ The X.Org Project https://www.x.org/ XFree86 Video Timings HOWTO http://www.tldp.org/HOWTO/XFree86-Video-Timings-HOWTO/index.html The X.Org Foundation http://www.x.org/ OpenGL http://www.opengl.org/ ______________________________________________________________________________ Chapter 46. Acknowledgements ______________________________________________________________________________ loki_update 'nvidia-installer' was inspired by the 'loki_update' tool: http://icculus.org/loki_setup/ makeself The self-extracting archive (also known as the '.run' file) is generated using 'makeself.sh': http://www.megastep.org/makeself/ The version of makeself.sh used to create the .run is bundled within the .run file, and can retrieved by extracting the .run file's contents, e.g.: $ sh NVIDIA-Linux-x86_64-570.123.06.run --extract-only $ ls -l NVIDIA-Linux-x86_64-570.123.06/makeself.sh LLVM Portions of the NVIDIA OpenCL implementation contain components licensed from third parties under the following terms: Clang & LLVM: Copyright (c) 2003-2008 University of Illinois at Urbana-Champaign. All rights reserved. Portions of LLVM's System library: Copyright (C) 2004 eXtensible Systems, Inc. Developed by: LLVM Team University of Illinois at Urbana-Champaign http://llvm.org Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal with the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: o Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimers. o Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimers in the documentation and/or other materials provided with the distribution. o Neither the names of the LLVM Team, University of Illinois at Urbana-Champaign, nor the names of its contributors may be used to endorse or promote products derived from this Software without specific prior written permission. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS WITH THE SOFTWARE. xz-embedded The self-installing .run package is compressed using xz, and includes a decompressor built from the xz-embedded project, available at http://tukaani.org/xz/embedded.html. jansson nvidia-settings uses jansson for parsing configuration files, available at http://www.digip.org/jansson/. This library carries the following copyright notice: Copyright (c) 2009-2012 Petri Lehtinen Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. lz4 The NVIDIA GPU driver uses the lz4 compression algorithm as implemented by the lz4 library, available at https://code.google.com/p/lz4/. This library carries the following copyright notice: LZ4 - Fast LZ compression algorithm Copyright (C) 2011-2013, Yann Collet. BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: o Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. o Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. You can contact the author at : o LZ4 source repository : http://code.google.com/p/lz4/ o LZ4 public forum : https://groups.google.com/forum/#!forum/lz4c libtirpc The libnvidia-ml (NVML) and nvidia-persistenced components rely on libtirpc for communication, available at https://sourceforge.net/projects/libtirpc/. This library carries the following copyright notice: Copyright (c) Copyright (c) Bull S.A. 2005 All Rights Reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: o Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. o Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. o The name of the author may not be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. OpenSSL Portions of the NVIDIA NGX implementation contain components licensed from OpenSSL under the following terms: Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 1. Definitions. "License" shall mean the terms and conditions for use, reproduction, and distribution as defined by Sections 1 through 9 of this document. "Licensor" shall mean the copyright owner or entity authorized by the copyright owner that is granting the License. "Legal Entity" shall mean the union of the acting entity and all other entities that control, are controlled by, or are under common control with that entity. For the purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the outstanding shares, or (iii) beneficial ownership of such entity. "You" (or "Your") shall mean an individual or Legal Entity exercising permissions granted by this License. "Source" form shall mean the preferred form for making modifications, including but not limited to software source code, documentation source, and configuration files. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation, and conversions to other media types. "Work" shall mean the work of authorship, whether in Source or Object form, made available under the License, as indicated by a copyright notice that is included in or attached to the work (an example is provided in the Appendix below). "Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works shall not include works that remain separable from, or merely link (or bind by name) to the interfaces of, the Work and Derivative Works thereof. "Contribution" shall mean any work of authorship, including the original version of the Work and any modifications or additions to that Work or Derivative Works thereof, that is intentionally submitted to Licensor for inclusion in the Work by the copyright owner or by an individual or Legal Entity authorized to submit on behalf of the copyright owner. For the purposes of this definition, "submitted" means any form of electronic, verbal, or written communication sent to the Licensor or its representatives, including but not limited to communication on electronic mailing lists, source code control systems, and issue tracking systems that are managed by, or on behalf of, the Licensor for the purpose of discussing and improving the Work, but excluding communication that is conspicuously marked or otherwise designated in writing by the copyright owner as "Not a Contribution." "Contributor" shall mean Licensor and any individual or Legal Entity on behalf of whom a Contribution has been received by Licensor and subsequently incorporated within the Work. 2. Grant of Copyright License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form. 3. Grant of Patent License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Work, where such license applies only to those patent claims licensable by such Contributor that are necessarily infringed by their Contribution(s) alone or by combination of their Contribution(s) with the Work to which such Contribution(s) was submitted. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Work or a Contribution incorporated within the Work constitutes direct or contributory patent infringement, then any patent licenses granted to You under this License for that Work shall terminate as of the date such litigation is filed. 4. Redistribution. You may reproduce and distribute copies of the Work or Derivative Works thereof in any medium, with or without modifications, and in Source or Object form, provided that You meet the following conditions: (a) You must give any other recipients of the Work or Derivative Works a copy of this License; and (b) You must cause any modified files to carry prominent notices stating that You changed the files; and (c) You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works; and (d) If the Work includes a "NOTICE" text file as part of its distribution, then any Derivative Works that You distribute must include a readable copy of the attribution notices contained within such NOTICE file, excluding those notices that do not pertain to any part of the Derivative Works, in at least one of the following places: within a NOTICE text file distributed as part of the Derivative Works; within the Source form or documentation, if provided along with the Derivative Works; or, within a display generated by the Derivative Works, if and wherever such third-party notices normally appear. The contents of the NOTICE file are for informational purposes only and do not modify the License. You may add Your own attribution notices within Derivative Works that You distribute, alongside or as an addendum to the NOTICE text from the Work, provided that such additional attribution notices cannot be construed as modifying the License. You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise complies with the conditions stated in this License. 5. Submission of Contributions. Unless You explicitly state otherwise, any Contribution intentionally submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of this License, without any additional terms or conditions. Notwithstanding the above, nothing herein shall supersede or modify the terms of any separate license agreement you may have executed with Licensor regarding such Contributions. 6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks, or product names of the Licensor, except as required for reasonable and customary use in describing the origin of the Work and reproducing the content of the NOTICE file. 7. Disclaimer of Warranty. Unless required by applicable law or agreed to in writing, Licensor provides the Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of using or redistributing the Work and assume any risks associated with Your exercise of permissions under this License. 8. Limitation of Liability. In no event and under no legal theory, whether in tort (including negligence), contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts) or agreed to in writing, shall any Contributor be liable to You for damages, including any direct, indirect, special, incidental, or consequential damages of any character arising as a result of this License or out of the use or inability to use the Work (including but not limited to damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if such Contributor has been advised of the possibility of such damages. 9. Accepting Warranty or Additional Liability. While redistributing the Work or Derivative Works thereof, You may choose to offer, and charge a fee for, acceptance of support, warranty, indemnity, or other liability obligations and/or rights consistent with this License. However, in accepting such obligations, You may act only on Your own behalf and on Your sole responsibility, not on behalf of any other Contributor, and only if You agree to indemnify, defend, and hold each Contributor harmless for any liability incurred by, or claims asserted against, such Contributor by reason of your accepting any such warranty or additional liability. END OF TERMS AND CONDITIONS libbele Portions of the NVIDIA NGX implementation contain components licensed from libbele under the following terms: Copyright (c) 2015-2016 David du Colombier Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. libelf Portions of the NVIDIA NGX implementation contain components licensed from libelf under the following terms: Copyright (c) 2016 David du Colombier Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. libcurl The NVIDIA NGX Updater implementation contains components licensed from libcurl under the following terms: COPYRIGHT AND PERMISSION NOTICE Copyright (c) 1996 - 2023, Daniel Stenberg, daniel@haxx.se, and many contributors, see the THANKS file. All rights reserved. Permission to use, copy, modify, and distribute this software for any purpose with or without fee is hereby granted, provided that the above copyright notice and this permission notice appear in all copies. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY RIGHTS. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Except as contained in this notice, the name of a copyright holder shall not be used in advertising or otherwise to promote the sale, use or other dealings in this Software without prior written authorization of the copyright holder. PKCS 11 This NVIDIA driver makes use of PKCS 11 header files from the OASIS Open organization. Copyright © OASIS Open 2023. All Rights Reserved. Distributed under the terms of the OASIS IPR Policy, [http://www.oasis-open.org/policies-guidelines/ipr], AS-IS, WITHOUT ANY IMPLIED OR EXPRESS WARRANTY; there is no warranty of MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE or NONINFRINGEMENT of the rights of others. Mbed TLS The proprietary flavor of nvidia.ko (see Chapter 44) contains code from Mbed TLS https://github.com/ARMmbed/mbedtls under the Apache License 2.0. ===== APACHE LICENSE BEGIN ===== Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 1. Definitions. "License" shall mean the terms and conditions for use, reproduction, and distribution as defined by Sections 1 through 9 of this document. "Licensor" shall mean the copyright owner or entity authorized by the copyright owner that is granting the License. "Legal Entity" shall mean the union of the acting entity and all other entities that control, are controlled by, or are under common control with that entity. For the purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the outstanding shares, or (iii) beneficial ownership of such entity. "You" (or "Your") shall mean an individual or Legal Entity exercising permissions granted by this License. "Source" form shall mean the preferred form for making modifications, including but not limited to software source code, documentation source, and configuration files. "Object" form shall mean any form resulting from mechanical transformation or translation of a Source form, including but not limited to compiled object code, generated documentation, and conversions to other media types. "Work" shall mean the work of authorship, whether in Source or Object form, made available under the License, as indicated by a copyright notice that is included in or attached to the work (an example is provided in the Appendix below). "Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works shall not include works that remain separable from, or merely link (or bind by name) to the interfaces of, the Work and Derivative Works thereof. "Contribution" shall mean any work of authorship, including the original version of the Work and any modifications or additions to that Work or Derivative Works thereof, that is intentionally submitted to Licensor for inclusion in the Work by the copyright owner or by an individual or Legal Entity authorized to submit on behalf of the copyright owner. For the purposes of this definition, "submitted" means any form of electronic, verbal, or written communication sent to the Licensor or its representatives, including but not limited to communication on electronic mailing lists, source code control systems, and issue tracking systems that are managed by, or on behalf of, the Licensor for the purpose of discussing and improving the Work, but excluding communication that is conspicuously marked or otherwise designated in writing by the copyright owner as "Not a Contribution." "Contributor" shall mean Licensor and any individual or Legal Entity on behalf of whom a Contribution has been received by Licensor and subsequently incorporated within the Work. 2. Grant of Copyright License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute the Work and such Derivative Works in Source or Object form. 3. Grant of Patent License. Subject to the terms and conditions of this License, each Contributor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Work, where such license applies only to those patent claims licensable by such Contributor that are necessarily infringed by their Contribution(s) alone or by combination of their Contribution(s) with the Work to which such Contribution(s) was submitted. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Work or a Contribution incorporated within the Work constitutes direct or contributory patent infringement, then any patent licenses granted to You under this License for that Work shall terminate as of the date such litigation is filed. 4. Redistribution. You may reproduce and distribute copies of the Work or Derivative Works thereof in any medium, with or without modifications, and in Source or Object form, provided that You meet the following conditions: (a) You must give any other recipients of the Work or Derivative Works a copy of this License; and (b) You must cause any modified files to carry prominent notices stating that You changed the files; and (c) You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work, excluding those notices that do not pertain to any part of the Derivative Works; and (d) If the Work includes a "NOTICE" text file as part of its distribution, then any Derivative Works that You distribute must include a readable copy of the attribution notices contained within such NOTICE file, excluding those notices that do not pertain to any part of the Derivative Works, in at least one of the following places: within a NOTICE text file distributed as part of the Derivative Works; within the Source form or documentation, if provided along with the Derivative Works; or, within a display generated by the Derivative Works, if and wherever such third-party notices normally appear. The contents of the NOTICE file are for informational purposes only and do not modify the License. You may add Your own attribution notices within Derivative Works that You distribute, alongside or as an addendum to the NOTICE text from the Work, provided that such additional attribution notices cannot be construed as modifying the License. You may add Your own copyright statement to Your modifications and may provide additional or different license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise complies with the conditions stated in this License. 5. Submission of Contributions. Unless You explicitly state otherwise, any Contribution intentionally submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of this License, without any additional terms or conditions. Notwithstanding the above, nothing herein shall supersede or modify the terms of any separate license agreement you may have executed with Licensor regarding such Contributions. 6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks, or product names of the Licensor, except as required for reasonable and customary use in describing the origin of the Work and reproducing the content of the NOTICE file. 7. Disclaimer of Warranty. Unless required by applicable law or agreed to in writing, Licensor provides the Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of using or redistributing the Work and assume any risks associated with Your exercise of permissions under this License. 8. Limitation of Liability. In no event and under no legal theory, whether in tort (including negligence), contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts) or agreed to in writing, shall any Contributor be liable to You for damages, including any direct, indirect, special, incidental, or consequential damages of any character arising as a result of this License or out of the use or inability to use the Work (including but not limited to damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if such Contributor has been advised of the possibility of such damages. 9. Accepting Warranty or Additional Liability. While redistributing the Work or Derivative Works thereof, You may choose to offer, and charge a fee for, acceptance of support, warranty, indemnity, or other liability obligations and/or rights consistent with this License. However, in accepting such obligations, You may act only on Your own behalf and on Your sole responsibility, not on behalf of any other Contributor, and only if You agree to indemnify, defend, and hold each Contributor harmless for any liability incurred by, or claims asserted against, such Contributor by reason of your accepting any such warranty or additional liability. END OF TERMS AND CONDITIONS ===== APACHE LICENSE END ===== boost Portions of the NVIDIA Linux GPU driver contain components licensed from boost under the following terms: Boost Software License - Version 1.0 - August 17th, 2003 Permission is hereby granted, free of charge, to any person or organization obtaining a copy of the software and accompanying documentation covered by this license (the "Software") to use, reproduce, display, distribute, execute, and transmit the Software, and to prepare derivative works of the Software, and to permit third-parties to whom the Software is furnished to do so, all subject to the following: The copyright notices in the Software and this entire statement, including the above license grant, this restriction and the following disclaimer, must be included in all copies of the Software, in whole or in part, and all derivative works of the Software, unless such copies or derivative works are solely in the form of machine-executable object code generated by a source language processor. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. gsl-lite Portions of the NVIDIA CUDA Debugger implementation, libcudadebugger.so, contain components licensed from gsl-lite under the following terms: The MIT License (MIT) Copyright (c) 2015 Martin Moene Copyright (c) 2015 Microsoft Corporation. All rights reserved. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Protocol Buffer Portions of the NVIDIA CUDA Debugger implementation, libcudadebugger.so, contain components licensed from Protocol Buffer under the following terms: Copyright 2008 Google Inc. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: o Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. o Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. o Neither the name of Google Inc. nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Code generated by the Protocol Buffer compiler is owned by the owner of the input file used when generating it. This code is not standalone and requires a support library to be linked with it. This support library is itself covered by the above license. ThreadPool Portions of the NVIDIA CUDA Debugger implementation, libcudadebugger.so, contain components licensed from ThreadPool under the following terms: Copyright (c) 2012 Jakob Progsch, Vaclav Zeman This software is provided 'as-is', without any express or implied warranty. In no event will the authors be held liable for any damages arising from the use of this software. Permission is granted to anyone to use this software for any purpose, including commercial applications, and to alter it and redistribute it freely, subject to the following restrictions: 1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required. 2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software. 3. This notice may not be removed or altered from any source distribution. zlib Portions of the NVIDIA CUDA Debugger implementation, libcudadebugger.so, contain components licensed from zlib under the following terms: Copyright (C) 1995-2017 Jean-loup Gailly and Mark Adler This software is provided 'as-is', without any express or implied warranty. In no event will the authors be held liable for any damages arising from the use of this software. Permission is granted to anyone to use this software for any purpose, including commercial applications, and to alter it and redistribute it freely, subject to the following restrictions: 1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required. 2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software. 3. This notice may not be removed or altered from any source distribution. expected Portions of the NVIDIA CUDA Debugger implementation, libcudadebugger.so, contain components licensed from the software 'expected' under the following terms: Creative Commons Legal Code CC0 1.0 Universal CREATIVE COMMONS CORPORATION IS NOT A LAW FIRM AND DOES NOT PROVIDE LEGAL SERVICES. DISTRIBUTION OF THIS DOCUMENT DOES NOT CREATE AN ATTORNEY-CLIENT RELATIONSHIP. CREATIVE COMMONS PROVIDES THIS INFORMATION ON AN "AS-IS" BASIS. CREATIVE COMMONS MAKES NO WARRANTIES REGARDING THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS PROVIDED HEREUNDER, AND DISCLAIMS LIABILITY FOR DAMAGES RESULTING FROM THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS PROVIDED HEREUNDER. Statement of Purpose The laws of most jurisdictions throughout the world automatically confer exclusive Copyright and Related Rights (defined below) upon the creator and subsequent owner(s) (each and all, an "owner") of an original work of authorship and/or a database (each, a "Work"). Certain owners wish to permanently relinquish those rights to a Work for the purpose of contributing to a commons of creative, cultural and scientific works ("Commons") that the public can reliably and without fear of later claims of infringement build upon, modify, incorporate in other works, reuse and redistribute as freely as possible in any form whatsoever and for any purposes, including without limitation commercial purposes. These owners may contribute to the Commons to promote the ideal of a free culture and the further production of creative, cultural and scientific works, or to gain reputation or greater distribution for their Work in part through the use and efforts of others. For these and/or other purposes and motivations, and without any expectation of additional consideration or compensation, the person associating CC0 with a Work (the "Affirmer"), to the extent that he or she is an owner of Copyright and Related Rights in the Work, voluntarily elects to apply CC0 to the Work and publicly distribute the Work under its terms, with knowledge of his or her Copyright and Related Rights in the Work and the meaning and intended legal effect of CC0 on those rights. 1. Copyright and Related Rights. A Work made available under CC0 may be protected by copyright and related or neighboring rights ("Copyright and Related Rights"). Copyright and Related Rights include, but are not limited to, the following: 1. the right to reproduce, adapt, distribute, perform, display, communicate, and translate a Work; 2. moral rights retained by the original author(s) and/or performer(s); 3. publicity and privacy rights pertaining to a person's image or likeness depicted in a Work; 4. rights protecting against unfair competition in regards to a Work, subject to the limitations in paragraph 4(a), below; 5. rights protecting the extraction, dissemination, use and reuse of data in a Work; 6. database rights (such as those arising under Directive 96/9/EC of the European Parliament and of the Council of 11 March 1996 on the legal protection of databases, and under any national implementation thereof, including any amended or successor version of such directive); and 7. other similar, equivalent or corresponding rights throughout the world based on applicable law or treaty, and any national implementations thereof. 2. Waiver. To the greatest extent permitted by, but not in contravention of, applicable law, Affirmer hereby overtly, fully, permanently, irrevocably and unconditionally waives, abandons, and surrenders all of Affirmer's Copyright and Related Rights and associated claims and causes of action, whether now known or unknown (including existing as well as future claims and causes of action), in the Work (i) in all territories worldwide, (ii) for the maximum duration provided by applicable law or treaty (including future time extensions), (iii) in any current or future medium and for any number of copies, and (iv) for any purpose whatsoever, including without limitation commercial, advertising or promotional purposes (the "Waiver"). Affirmer makes the Waiver for the benefit of each member of the public at large and to the detriment of Affirmer's heirs and successors, fully intending that such Waiver shall not be subject to revocation, rescission, cancellation, termination, or any other legal or equitable action to disrupt the quiet enjoyment of the Work by the public as contemplated by Affirmer's express Statement of Purpose. 3. Public License Fallback. Should any part of the Waiver for any reason be judged legally invalid or ineffective under applicable law, then the Waiver shall be preserved to the maximum extent permitted taking into account Affirmer's express Statement of Purpose. In addition, to the extent the Waiver is so judged Affirmer hereby grants to each affected person a royalty-free, non transferable, non sublicensable, non exclusive, irrevocable and unconditional license to exercise Affirmer's Copyright and Related Rights in the Work (i) in all territories worldwide, (ii) for the maximum duration provided by applicable law or treaty (including future time extensions), (iii) in any current or future medium and for any number of copies, and (iv) for any purpose whatsoever, including without limitation commercial, advertising or promotional purposes (the "License"). The License shall be deemed effective as of the date CC0 was applied by Affirmer to the Work. Should any part of the License for any reason be judged legally invalid or ineffective under applicable law, such partial invalidity or ineffectiveness shall not invalidate the remainder of the License, and in such case Affirmer hereby affirms that he or she will not (i) exercise any of his or her remaining Copyright and Related Rights in the Work or (ii) assert any associated claims and causes of action with respect to the Work, in either case contrary to Affirmer's express Statement of Purpose. 4. Limitations and Disclaimers. 1. No trademark or patent rights held by Affirmer are waived, abandoned, surrendered, licensed or otherwise affected by this document. 2. Affirmer offers the Work as-is and makes no representations or warranties of any kind concerning the Work, express, implied, statutory or otherwise, including without limitation warranties of title, merchantability, fitness for a particular purpose, non infringement, or the absence of latent or other defects, accuracy, or the present or absence of errors, whether or not discoverable, all to the greatest extent permissible under applicable law. 3. Affirmer disclaims responsibility for clearing rights of other persons that may apply to the Work or any use thereof, including without limitation any person's Copyright and Related Rights in the Work. Further, Affirmer disclaims responsibility for obtaining any necessary consents, permissions or other rights required for any use of the Work. 4. Affirmer understands and acknowledges that Creative Commons is not a party to this document and has no duty or obligation with respect to this CC0 or use of the Work. Linux kernel headers The NVIDIA Linux GPU driver makes use of various Linux kernel header files licensed under GPL-2.0 WITH Linux-syscall-note: SPDX-Exception-Identifier: Linux-syscall-note SPDX-URL: https://spdx.org/licenses/Linux-syscall-note.html SPDX-Licenses: GPL-2.0, GPL-2.0+, GPL-1.0+, LGPL-2.0, LGPL-2.0+, LGPL-2.1, LGPL-2.1+, GPL-2.0-only, GPL-2.0-or-later Usage-Guide: This exception is used together with one of the above SPDX-Licenses to mark user space API (uapi) header files so they can be included into non GPL compliant user space application code. To use this exception add it with the keyword WITH to one of the identifiers in the SPDX-Licenses tag: SPDX-License-Identifier: WITH Linux-syscall-note License-Text: NOTE! This copyright does *not* cover user programs that use kernel services by normal system calls - this is merely considered normal use of the kernel, and does *not* fall under the heading of "derived work". Also note that the GPL below is copyrighted by the Free Software Foundation, but the instance of code that it refers to (the Linux kernel) is copyrighted by me and others who actually wrote it. Also note that the only valid version of the GPL as far as the kernel is concerned is _this_ particular version of the license (ie v2, not v2.2 or v3.x or whatever), unless explicitly otherwise stated. Linus Torvalds X.Org This NVIDIA Linux driver contains code from the X.Org project. Source code from the X.Org project is available from http://cgit.freedesktop.org/xorg/xserver JSMN This NVIDIA Linux driver uses a JSON parser based on 'jsmn': http://zserge.bitbucket.org/jsmn.html This library carries the following copyright notice: Copyright (c) 2010 Serge A. Zaitsev Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. SHA-256 Portions of the driver use the SHA-256 algorithm derived from sha2.c: https://github.com/ouah/sha2/blob/master/sha2.c This library carries the following copyright notice: Copyright (C) 2005, 2007 Olivier Gay All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the project nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. libselinux This NVIDIA Linux driver contains code from libselinux, which is released in the public domain. SoftFloat Portions of the driver use the SoftFloat floating point emulation library: http://www.jhauser.us/arithmetic/SoftFloat.html SoftFloat has the following license terms: John R. Hauser 2017 August 10 The following applies to the whole of SoftFloat Release 3d as well as to each source file individually. Copyright 2011, 2012, 2013, 2014, 2015, 2016, 2017 The Regents of the University of California. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions, and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions, and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the University nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. musl libc This NVIDIA Linux driver uses code from 'musl libc': http://musl.libc.org This code carries the following copyright notice: musl as a whole is licensed under the following standard MIT license: ---------------------------------------------------------------------- Copyright © 2005-2014 Rich Felker, et al. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ---------------------------------------------------------------------- Authors/contributors include: A. Wilcox Alex Dowad Alexander Monakov Andrew Kelley Anthony G. Basile Arvid Picciani Bartosz Brachaczek Bobby Bingham Boris Brezillon Brent Cook Chris Spiegel Clément Vasseur Daniel Micay Daniel Sabogal Daurnimator David Edelsohn Denys Vlasenko Dmitry Ivanov Dmitry V. Levin Emil Renner Berthing Felix Fietkau Felix Janda Gianluca Anzolin Hauke Mehrtens He X Hiltjo Posthuma Isaac Dunham Jaydeep Patil Jens Gustedt Jeremy Huntwork Jo-Philipp Wich Joakim Sindholt John Spencer Josiah Worcester Julien Ramseier Justin Cormack Khem Raj Kylie McClain Leah Neukirchen Luca Barbato Luka Perkov M Farkas-Dyck (Strake) Mahesh Bodapati Masanori Ogino Michael Forney Mikhail Kremnyov Natanael Copa Nicholas J. Kain orc Pascal Cuoq Petr Hosek Petr Skocik Pierre Carrier Reini Urban Rich Felker Richard Pennington Samuel Holland Shiz sin Solar Designer Stefan Kristiansson Szabolcs Nagy Timo Teräs Trutz Behn Valentin Ochs William Haddon William Pitcock Portions of this software are derived from third-party works licensed under terms compatible with the above MIT license: The TRE regular expression implementation (src/regex/reg* and src/regex/tre*) is Copyright © 2001-2008 Ville Laurikari and licensed under a 2-clause BSD license (license text in the source files). The included version has been heavily modified by Rich Felker in 2012, in the interests of size, simplicity, and namespace cleanliness. Much of the math library code (src/math/* and src/complex/*) is Copyright © 1993,2004 Sun Microsystems or Copyright © 2003-2011 David Schultz or Copyright © 2003-2009 Steven G. Kargl or Copyright © 2003-2009 Bruce D. Evans or Copyright © 2008 Stephen L. Moshier and labelled as such in comments in the individual source files. All have been licensed under extremely permissive terms. The ARM memcpy code (src/string/arm/memcpy_el.S) is Copyright © 2008 The Android Open Source Project and is licensed under a two-clause BSD license. It was taken from Bionic libc, used on Android. The implementation of DES for crypt (src/crypt/crypt_des.c) is Copyright © 1994 David Burren. It is licensed under a BSD license. The implementation of blowfish crypt (src/crypt/crypt_blowfish.c) was originally written by Solar Designer and placed into the public domain. The code also comes with a fallback permissive license for use in jurisdictions that may not recognize the public domain. The smoothsort implementation (src/stdlib/qsort.c) is Copyright © 2011 Valentin Ochs and is licensed under an MIT-style license. The x86_64 port was written by Nicholas J. Kain and is licensed under the standard MIT terms. The mips and microblaze ports were originally written by Richard Pennington for use in the ellcc project. The original code was adapted by Rich Felker for build system and code conventions during upstream integration. It is licensed under the standard MIT terms. The mips64 port was contributed by Imagination Technologies and is licensed under the standard MIT terms. The powerpc port was also originally written by Richard Pennington, and later supplemented and integrated by John Spencer. It is licensed under the standard MIT terms. All other files which have no copyright comments are original works produced specifically for use as part of this library, written either by Rich Felker, the main author of the library, or by one or more contibutors listed above. Details on authorship of individual files can be found in the git version control history of the project. The omission of copyright and license comments in each file is in the interest of source tree size. In addition, permission is hereby granted for all public header files (include/* and arch/*/bits/*) and crt files intended to be linked into applications (crt/*, ldso/dlstart.c, and arch/*/crt_arch.h) to omit the copyright notice and permission notice otherwise required by the license, and to use these files without any requirement of attribution. These files include substantial contributions from: Bobby Bingham John Spencer Nicholas J. Kain Rich Felker Richard Pennington Stefan Kristiansson Szabolcs Nagy all of whom have explicitly granted such permission. This file previously contained text expressing a belief that most of the files covered by the above exception were sufficiently trivial not to be subject to copyright, resulting in confusion over whether it negated the permissions granted in the license. In the spirit of permissive licensing, and of not having licensing issues being an obstacle to adoption, that text has been removed. Zstandard This NVIDIA Linux driver uses the 'Zstandard' library for compression: https://github.com/facebook/zstd BSD License For Zstandard software Copyright (c) 2016-present, Facebook, Inc. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: o Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. o Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. o Neither the name Facebook nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. libspdm This NVIDIA Linux driver uses code from 'libspdm': https://github.com/DMTF/libspdm This code carries the following license: The Distributed Management Task Force (DMTF) grants rights under copyright in this software on the terms of the BSD 3-Clause License as set forth below; no other rights are granted by DMTF. This software might be subject to other rights (such as patent rights) of other parties. Copyrights. Copyright (c) 2021-2022, Contributing Member(s) of Distributed Management Task Force, Inc.. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the Distributed Management Task Force (DMTF) nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Patents. This software may be subject to third party patent rights, including provisional patent rights ("patent rights"). DMTF makes no representations to users of the standard as to the existence of such rights, and is not responsible to recognize, disclose, or identify any or all such third party patent right, owners or claimants, nor for any incomplete or inaccurate identification or disclosure of such rights, owners or claimants. DMTF shall have no liability to any party, in any manner or circumstance, under any legal theory whatsoever, for failure to recognize, disclose, or identify any such third party patent rights, or for such party's reliance on the software or incorporation thereof in its product, protocols or testing procedures. DMTF shall have no liability to any party using such software, whether such use is foreseeable or not, nor to any patent owner or claimant, and shall have no liability or responsibility for costs or losses incurred if software is withdrawn or modified after publication, and shall be indemnified and held harmless by any party using the software from any and all claims of infringement by a patent owner for such use. DMTF Members that contributed to this software source code might have made patent licensing commitments in connection with their participation in the DMTF. For details, see http://dmtf.org/sites/default/files/patent-10-18-01.pdf and http://www.dmtf.org/about/policies/disclosures. Mesa This NVIDIA Linux driver uses code from the 'Mesa' project: https://gitlab.freedesktop.org/mesa/mesa The code used carries the following copyright notices: Copyright (c) 2011 Intel Corporation Copyright (c) 2021 NVIDIA Corporation Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice (including the next paragraph) shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Authors: o Benjamin Franzke o James Jones --- Copyright (c) 2008-2011 Kristian Høgsberg Copyright (c) 2010-2011 Intel Corporation Permission to use, copy, modify, distribute, and sell this software and its documentation for any purpose is hereby granted without fee, provided that\n the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation, and that the name of the copyright holders not be used in advertising or publicity pertaining to distribution of the software without specific, written prior permission. The copyright holders make no representations about the suitability of this software for any purpose. It is provided "as is" without express or implied warranty. THE COPYRIGHT HOLDERS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. Wayland protocols This NVIDIA Linux driver uses code from the 'Wayland protocols' project: https://gitlab.freedesktop.org/wayland/wayland-protocols The project carries the following copyright notice: Copyright (c) 2008-2013 Kristian Høgsberg Copyright (c) 2010-2013 Intel Corporation Copyright (c) 2013 Rafael Antognolli Copyright (c) 2013 Jasper St. Pierre Copyright (c) 2014 Jonas Ådahl Copyright (c) 2014 Jason Ekstrand Copyright (c) 2014-2015 Collabora, Ltd. Copyright (c) 2015 Red Hat Inc. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice (including the next paragraph) shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. --- The above is the version of the MIT "Expat" License used by X.org: http://cgit.freedesktop.org/xorg/xserver/tree/COPYING virglrenderer This NVIDIA Linux driver contains code from the virglrenderer project Source code for the virglrenderer project is available from http://gitlab.freedesktop.org/virgl/virglrenderer The project carries the following copyright notice: virglrenderer is under MIT license and derived from mesa in many parts. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. fdlibm This NVIDIA Linux driver uses code from the fdlibm project: https://www.netlib.org/fdlibm/ The code used carries the following copyright notices: Copyright (C) 2004 by Sun Microsystems, Inc. All rights reserved. Permission to use, copy, modify, and distribute this software is freely granted, provided that this notice is preserved. Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. Developed at SunSoft, a Sun Microsystems, Inc. business. Permission to use, copy, modify, and distribute this software is freely granted, provided that this notice is preserved. libpciaccess This NVIDIA Linux driver contains code from the libpciaccess project Source code for the libpciaccess project is available from https://gitlab.freedesktop.org/xorg/lib/libpciaccess The project carries the following copyright notices: (C) Copyright IBM Corporation 2006, 2007 (C) Copyright Eric Anholt 2006 (C) Copyright Mark Kettenis 2011 (C) Copyright Robert Millan 2012 Copyright (c) 2007, 2008, 2009, 2011, 2012, 2013 Oracle and/or its affiliates. Copyright 2009, 2012 Red Hat, Inc. All Rights Reserved. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation on the rights to use, copy, modify, merge, publish, distribute, sub license, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice (including the next paragraph) shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Copyright (c) 2008 Juan Romero Pardines Copyright (c) 2008, 2011 Mark Kettenis Copyright (c) 2009 Michael Lorenz Copyright (c) 2009, 2012 Samuel Thibault Permission to use, copy, modify, and distribute this software for any purpose with or without fee is hereby granted, provided that the above copyright notice and this permission notice appear in all copies. THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. Copyright (C) 2000 The XFree86 Project, Inc. All Rights Reserved. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE XFREE86 PROJECT BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. Except as contained in this notice, the name of the XFree86 Project shall not be used in advertising or otherwise to promote the sale, use or other dealings in this Software without prior written authorization from the XFree86 Project. Copyright (c) 2007 Paulo R. Zanoni, Tiago Vignatti Copyright (c) 2009 Tiago Vignatti Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. hwloc This NVIDIA Linux driver uses the Portable Hardware Locality (hwloc) library from https://www.open-mpi.org/software/hwloc This library carries the following copyright notice: Copyright © 2004-2006 The Trustees of Indiana University and Indiana University Research and Technology Corporation. All rights reserved. Copyright © 2004-2005 The University of Tennessee and The University of Tennessee Research Foundation. All rights reserved. Copyright © 2004-2005 High Performance Computing Center Stuttgart, University of Stuttgart. All rights reserved. Copyright © 2004-2005 The Regents of the University of California. All rights reserved. Copyright © 2009 CNRS Copyright © 2009-2016 Inria. All rights reserved. Copyright © 2009-2015 Université Bordeaux Copyright © 2009-2015 Cisco Systems, Inc. All rights reserved. Copyright © 2009-2012 Oracle and/or its affiliates. All rights reserved. Copyright © 2010 IBM Copyright © 2010 Jirka Hladky Copyright © 2012 Aleksej Saushev, The NetBSD Foundation Copyright © 2012 Blue Brain Project, EPFL. All rights reserved. Copyright © 2013-2014 University of Wisconsin-La Crosse. All rights reserved. Copyright © 2015 Research Organization for Information Science and Technology (RIST). All rights reserved. Copyright © 2015-2016 Intel, Inc. All rights reserved. See COPYING in top-level directory. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. The name of the author may not be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. ______________________________________________________________________________ Appendix A. Supported NVIDIA GPU Products ______________________________________________________________________________ For the most complete and accurate listing of supported GPUs, please see the Supported Products List, available from the NVIDIA Linux x86_64 Graphics Driver download page. Please go to http://www.nvidia.com/object/unix.html, follow the Archive link under the Linux x86_64 heading, follow the link for the 570.123.06 driver, and then go to the Supported Products List. For an explanation of the VDPAU features column, please refer to the section called "VdpDecoder" of Appendix G. Note that the list of supported GPU products provided below and on the driver download page is provided to indicate which GPUs are supported by a particular driver version. Some designs incorporating supported GPUs may not be compatible with the NVIDIA Linux driver: in particular, notebook and all-in-one desktop designs with switchable (hybrid) or Optimus graphics will not work if means to disable the integrated graphics in hardware are not available. Hardware designs will vary from manufacturer to manufacturer, so please consult with a system's manufacturer to determine whether that particular system is compatible. A1. CURRENT NVIDIA GPUS NVIDIA GPU product Device PCI ID* VDPAU features ----------------------------------------------- -------------- -------------- NVIDIA GeForce 830M 1340 E NVIDIA GeForce 830A 1340 103C 2B2B E NVIDIA GeForce 840M 1341 E NVIDIA GeForce 840A 1341 17AA 3697 E NVIDIA GeForce 840A 1341 17AA 3699 E NVIDIA GeForce 840A 1341 17AA 369C E NVIDIA GeForce 840A 1341 17AA 36AF E NVIDIA GeForce 845M 1344 E NVIDIA GeForce 930M 1346 E NVIDIA GeForce 930A 1346 17AA 30BA E NVIDIA GeForce 930A 1346 17AA 362C E NVIDIA GeForce 930A 1346 17AA 362F E NVIDIA GeForce 930A 1346 17AA 3636 E NVIDIA GeForce 940M 1347 E NVIDIA GeForce 940A 1347 17AA 36B9 E NVIDIA GeForce 940A 1347 17AA 36BA E NVIDIA GeForce 945M 1348 E NVIDIA GeForce 945A 1348 103C 2B5C E NVIDIA GeForce 930M 1349 E NVIDIA GeForce 930A 1349 17AA 3124 E NVIDIA GeForce 930A 1349 17AA 364B E NVIDIA GeForce 930A 1349 17AA 36C3 E NVIDIA GeForce 930A 1349 17AA 36D1 E NVIDIA GeForce 930A 1349 17AA 36D8 E NVIDIA GeForce 940MX 134B E NVIDIA GeForce 940MX 134D E NVIDIA GeForce 930MX 134E E NVIDIA GeForce 920MX 134F E Quadro K620M 137A 17AA 2225 E Quadro M500M 137A 17AA 2232 E Quadro M500M 137A 17AA 505A E Quadro M520 137B E NVIDIA GeForce GTX 750 Ti 1380 E NVIDIA GeForce GTX 750 1381 E NVIDIA GeForce GTX 745 1382 E NVIDIA GeForce 845M 1390 E NVIDIA GeForce GTX 850M 1391 E NVIDIA GeForce GTX 850A 1391 17AA 3697 E NVIDIA GeForce GTX 860M 1392 E NVIDIA GeForce GPU 1392 1028 066A E NVIDIA GeForce GTX 750 Ti 1392 1043 861E E NVIDIA GeForce GTX 750 Ti 1392 1043 86D9 E NVIDIA GeForce 840M 1393 E NVIDIA GeForce 845M 1398 E NVIDIA GeForce 945M 1399 E NVIDIA GeForce GTX 950M 139A E NVIDIA GeForce GTX 950A 139A 17AA 362C E NVIDIA GeForce GTX 950A 139A 17AA 362F E NVIDIA GeForce GTX 950A 139A 17AA 363F E NVIDIA GeForce GTX 950A 139A 17AA 3640 E NVIDIA GeForce GTX 950A 139A 17AA 3647 E NVIDIA GeForce GTX 950A 139A 17AA 36B9 E NVIDIA GeForce GTX 960M 139B E NVIDIA GeForce GTX 750 Ti 139B 1025 107A E NVIDIA GeForce GTX 860M 139B 1028 06A3 E NVIDIA GeForce GTX 960A 139B 103C 2B4C E NVIDIA GeForce GTX 750Ti 139B 17AA 3649 E NVIDIA GeForce GTX 960A 139B 17AA 36BF E NVIDIA GeForce GTX 750 Ti 139B 19DA C248 E NVIDIA GeForce GTX 750Ti 139B 1AFA 8A75 E NVIDIA GeForce 940M 139C E NVIDIA GeForce GTX 750 Ti 139D E Quadro M2000M 13B0 E Quadro M1000M 13B1 E Quadro M600M 13B2 E Quadro K2200M 13B3 E Quadro M620 13B4 E Quadro M1200 13B6 E NVS 810 13B9 E Quadro K2200 13BA E Quadro K620 13BB E Quadro K1200 13BC E NVIDIA GeForce GTX 980 13C0 E NVIDIA GeForce GTX 970 13C2 E NVIDIA GeForce GTX 980M 13D7 E NVIDIA GeForce GTX 970M 13D8 E NVIDIA GeForce GTX 960 13D8 1462 1198 E NVIDIA GeForce GTX 960 13D8 1462 1199 E NVIDIA GeForce GTX 960 13D8 19DA B282 E NVIDIA GeForce GTX 960 13D8 19DA B284 E NVIDIA GeForce GTX 960 13D8 19DA B286 E NVIDIA GeForce GTX 965M 13D9 E NVIDIA GeForce GTX 980 13DA E Quadro M5000 13F0 E Quadro M4000 13F1 E Tesla M60 13F2 E Tesla M6 13F3 E Quadro M5000M 13F8 E Quadro M5000 SE 13F8 10DE 11DD E Quadro M4000M 13F9 E Quadro M3000M 13FA E Quadro M3000 SE 13FA 10DE 11C9 E Quadro M5500 13FB E NVIDIA GeForce GTX 960 1401 E NVIDIA GeForce GTX 950 1402 F NVIDIA GeForce GTX 960 1406 E NVIDIA GeForce GTX 750 1407 E NVIDIA GeForce GTX 965M 1427 E NVIDIA GeForce GTX 950 1427 1458 D003 F Quadro M2000 1430 F Tesla M4 1431 F Quadro M2200 1436 F Quadro GP100 15F0 G Tesla P100-PCIE-12GB 15F7 G Tesla P100-PCIE-16GB 15F8 G Tesla P100-SXM2-16GB 15F9 G NVIDIA GeForce GTX 980M 1617 E NVIDIA GeForce GTX 970M 1618 E NVIDIA GeForce GTX 965M 1619 E NVIDIA GeForce GTX 980 161A E NVIDIA GeForce GTX 965M 1667 E NVIDIA GeForce MX130 174D E NVIDIA GeForce MX110 174E E NVIDIA GeForce 940MX 179C E NVIDIA GeForce GTX TITAN X 17C2 E NVIDIA GeForce GTX 980 Ti 17C8 E Quadro M6000 17F0 E Quadro M6000 24GB 17F1 E Tesla M40 17FD E Tesla M40 24GB 17FD 10DE 1173 E NVIDIA TITAN X (Pascal) 1B00 H NVIDIA TITAN Xp 1B02 H NVIDIA TITAN Xp COLLECTORS EDITION 1B02 10DE 123E H NVIDIA TITAN Xp COLLECTORS EDITION 1B02 10DE 123F H NVIDIA GeForce GTX 1080 Ti 1B06 H Quadro P6000 1B30 H Tesla P40 1B38 H NVIDIA GeForce GTX 1080 1B80 H 2 NVIDIA GeForce GTX 1070 1B81 H 2 NVIDIA GeForce GTX 1070 Ti 1B82 H 2 NVIDIA GeForce GTX 1060 6GB 1B83 H 2 NVIDIA GeForce GTX 1060 3GB 1B84 H 2 NVIDIA P104-100 1B87 H 2 NVIDIA GeForce GTX 1080 1BA0 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BA0 1028 0887 H 2 NVIDIA GeForce GTX 1070 1BA1 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1028 08A1 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1028 08A2 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1043 1CCE H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1458 1651 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1458 1653 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1462 11E8 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1462 11E9 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1462 1225 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1462 1226 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1462 1227 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1558 9501 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1558 95E1 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1A58 2000 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BA1 1D05 1032 H 2 NVIDIA GeForce GTX 1070 1BA2 H 2 Quadro P5000 1BB0 H 2 Quadro P4000 1BB1 H 2 Tesla P6 1BB4 H 2 Quadro P5200 1BB5 H 2 Quadro P5200 with Max-Q Design 1BB5 17AA 2268 H 2 Quadro P5200 with Max-Q Design 1BB5 17AA 2269 H 2 Quadro P5000 1BB6 H 2 Quadro P4000 1BB7 H 2 Quadro P4000 with Max-Q Design 1BB7 1462 11E9 H 2 Quadro P4000 with Max-Q Design 1BB7 1558 9501 H 2 Quadro P3000 1BB8 H 2 Quadro P4200 1BB9 H 2 Quadro P4200 with Max-Q Design 1BB9 1558 95E1 H 2 Quadro P4200 with Max-Q Design 1BB9 17AA 2268 H 2 Quadro P4200 with Max-Q Design 1BB9 17AA 2269 H 2 Quadro P3200 1BBB H 2 Quadro P3200 with Max-Q Design 1BBB 17AA 225F H 2 Quadro P3200 with Max-Q Design 1BBB 17AA 2262 H 2 NVIDIA P104-101 1BC7 H 2 NVIDIA GeForce GTX 1080 1BE0 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1025 1221 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1025 123E H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1028 07C0 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1028 0876 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1028 088B H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1043 1031 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1043 1BF0 H 2 NVIDIA GeForce GTX 1080 with Max-Q Design 1BE0 1458 355B H 2 NVIDIA GeForce GTX 1070 1BE1 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BE1 103C 84DB H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BE1 1043 16F0 H 2 NVIDIA GeForce GTX 1070 with Max-Q Design 1BE1 3842 2009 H 2 NVIDIA GeForce GTX 1060 3GB 1C02 H 2 NVIDIA GeForce GTX 1060 6GB 1C03 H 2 NVIDIA GeForce GTX 1060 5GB 1C04 H NVIDIA GeForce GTX 1060 6GB 1C06 H 2 NVIDIA P106-100 1C07 H NVIDIA P106-090 1C09 H NVIDIA GeForce GTX 1060 1C20 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1028 0802 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1028 0803 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1028 0825 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1028 0827 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1028 0885 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1028 0886 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 103C 8467 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 103C 8478 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 103C 8581 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1462 1244 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1558 95E5 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 17AA 39B9 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1A58 2000 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1A58 2001 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C20 1D05 1059 H NVIDIA GeForce GTX 1050 Ti 1C21 H NVIDIA GeForce GTX 1050 1C22 H NVIDIA GeForce GTX 1060 1C23 H Quadro P2000 1C30 H Quadro P2200 1C31 H NVIDIA GeForce GTX 1060 1C60 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C60 103C 8390 H NVIDIA GeForce GTX 1060 with Max-Q Design 1C60 103C 8467 H NVIDIA GeForce GTX 1050 Ti 1C61 H NVIDIA GeForce GTX 1050 1C62 H NVIDIA GeForce GTX 1050 1C81 H NVIDIA GeForce GTX 1050 Ti 1C82 H NVIDIA GeForce GTX 1050 1C83 H NVIDIA GeForce GTX 1050 Ti 1C8C H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 1028 087C H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 103C 8519 H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 103C 856A H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 1462 123C H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 1462 126C H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 17AA 2266 H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 17AA 2267 H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8C 17AA 39FF H NVIDIA GeForce GTX 1050 1C8D H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 103C 84E9 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 103C 84EB H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 103C 856A H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 114F H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 1341 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 1351 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 1481 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 14A1 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 18C1 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1043 1B5E H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1462 126C H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 152D 1217 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C8D 1D72 1707 H NVIDIA GeForce GTX 1050 Ti 1C8F H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8F 1462 123C H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8F 1462 126C H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8F 1462 126D H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8F 1462 1284 H NVIDIA GeForce GTX 1050 Ti with Max-Q Design 1C8F 1462 1297 H NVIDIA GeForce MX150 1C90 H NVIDIA GeForce MX250 1C90 1028 09C1 H NVIDIA GeForce GTX 1050 1C91 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C91 103C 856A H NVIDIA GeForce GTX 1050 with Max-Q Design 1C91 103C 86E3 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C91 152D 1232 H NVIDIA GeForce GTX 1050 1C92 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C92 1043 149F H NVIDIA GeForce GTX 1050 with Max-Q Design 1C92 1043 1B31 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C92 1462 1245 H NVIDIA GeForce GTX 1050 with Max-Q Design 1C92 1462 126C H NVIDIA GeForce MX350 1C94 H NVIDIA GeForce MX350 1C96 H Quadro P1000 1CB1 H Quadro P600 1CB2 H Quadro P400 1CB3 H Quadro P620 1CB6 H Quadro P2000 1CBA H Quadro P2000 with Max-Q Design 1CBA 17AA 2266 H Quadro P2000 with Max-Q Design 1CBA 17AA 2267 H Quadro P1000 1CBB H Quadro P600 1CBC H Quadro P620 1CBD H Quadro P2000 1CFA H Quadro P1000 1CFB H Matrox D-Series D1480 1CFB 102B 2600 H Matrox D-Series D1450 1CFB 102B 2700 H NVIDIA GeForce GT 1030 1D01 H NVIDIA GeForce GT 1010 1D02 H NVIDIA GeForce MX150 1D10 H NVIDIA GeForce MX230 1D11 H NVIDIA GeForce MX150 1D12 H NVIDIA GeForce MX250 1D13 H NVIDIA GeForce MX330 1D16 H Quadro P500 1D33 H Quadro P520 1D34 H NVIDIA GeForce MX250 1D52 H NVIDIA TITAN V 1D81 I Tesla V100-SXM2-16GB 1DB1 I Tesla V100-SXM2-16GB-LS 1DB1 10DE 1307 I Tesla V100-FHHL-16GB 1DB3 I Tesla V100-PCIE-16GB 1DB4 I Tesla V100-PCIE-16GB-LS 1DB4 10DE 1306 I Tesla V100-SXM2-32GB 1DB5 I Tesla V100-SXM2-32GB-LS 1DB5 10DE 1308 I Tesla V100-PCIE-32GB 1DB6 I Tesla V100-DGXS-32GB 1DB7 I Tesla V100-SXM3-32GB 1DB8 I Tesla V100-SXM3-32GB-H 1DB8 10DE 131D I Quadro GV100 1DBA I NVIDIA TITAN V JHH Special Edition 1DBA 10DE 12EB I Tesla PG500-216 1DF0 I Tesla PG503-216 1DF2 I Tesla V100S-PCIE-32GB 1DF6 I NVIDIA TITAN RTX 1E02 J NVIDIA GeForce RTX 2080 Ti 1E04 J NVIDIA GeForce RTX 2080 Ti 1E07 J NVIDIA CMP 50HX 1E09 - Quadro RTX 6000 1E30 J Quadro RTX 8000 1E30 1028 129E J Quadro RTX 8000 1E30 103C 129E J Quadro RTX 8000 1E30 10DE 129E J Quadro RTX 6000 1E36 J Quadro RTX 8000 1E78 10DE 13D8 J Quadro RTX 6000 1E78 10DE 13D9 J NVIDIA GeForce RTX 2080 SUPER 1E81 J NVIDIA GeForce RTX 2080 1E82 J NVIDIA GeForce RTX 2070 SUPER 1E84 J NVIDIA GeForce RTX 2080 1E87 J NVIDIA GeForce RTX 2060 1E89 J NVIDIA GeForce RTX 2080 1E90 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1025 1375 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08A1 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08A2 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08EA J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08EB J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08EC J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08ED J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08EE J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 08EF J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 093B J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1028 093C J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 8572 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 8573 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 8602 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 8606 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 86C6 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 86C7 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 87A6 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 103C 87A7 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1043 131F J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1043 137F J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1043 141F J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1043 1751 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 1660 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 1661 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 1662 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 75A6 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 75A7 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 86A6 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1458 86A7 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1462 1274 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1462 1277 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 152D 1220 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1558 95E1 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1558 97E1 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1A58 2002 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1A58 2005 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1A58 2007 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1A58 3000 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1A58 3001 J NVIDIA GeForce RTX 2080 with Max-Q Design 1E90 1D05 1069 J NVIDIA GeForce RTX 2070 Super 1E91 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 103C 8607 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 103C 8736 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 103C 8738 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 103C 8772 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 103C 878A J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 103C 878B J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1043 1E61 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 1511 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 75B3 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 75B4 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 76B2 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 76B3 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 78A2 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 78A3 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 86B2 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1458 86B3 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1462 12AE J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1462 12B0 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1462 12C6 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 17AA 22C3 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 17AA 22C5 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1A58 2009 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1A58 200A J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 1A58 3002 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1E91 8086 3012 J NVIDIA GeForce RTX 2080 Super 1E93 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1025 1401 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1025 149C J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1028 09D2 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 8607 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 86C7 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 8736 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 8738 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 8772 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 87A6 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 103C 87A7 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 75B1 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 75B2 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 76B0 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 76B1 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 78A0 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 78A1 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 86B0 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1458 86B1 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1462 12AE J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1462 12B0 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1462 12B4 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1462 12C6 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1558 50D3 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1558 70D1 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 17AA 22C3 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 17AA 22C5 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1A58 2009 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1A58 200A J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1A58 3002 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1E93 1D05 1089 J Quadro RTX 5000 1EB0 J Quadro RTX 4000 1EB1 J Quadro RTX 5000 1EB5 J Quadro RTX 5000 with Max-Q Design 1EB5 1025 1375 J Quadro RTX 5000 with Max-Q Design 1EB5 1025 1401 J Quadro RTX 5000 with Max-Q Design 1EB5 1025 149C J Quadro RTX 5000 with Max-Q Design 1EB5 1028 09C3 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8736 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8738 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8772 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8780 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8782 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8783 J Quadro RTX 5000 with Max-Q Design 1EB5 103C 8785 J Quadro RTX 5000 with Max-Q Design 1EB5 1043 1DD1 J Quadro RTX 5000 with Max-Q Design 1EB5 1462 1274 J Quadro RTX 5000 with Max-Q Design 1EB5 1462 12B0 J Quadro RTX 5000 with Max-Q Design 1EB5 1462 12C6 J Quadro RTX 5000 with Max-Q Design 1EB5 17AA 22B8 J Quadro RTX 5000 with Max-Q Design 1EB5 17AA 22BA J Quadro RTX 5000 with Max-Q Design 1EB5 1A58 2005 J Quadro RTX 5000 with Max-Q Design 1EB5 1A58 2007 J Quadro RTX 5000 with Max-Q Design 1EB5 1A58 2008 J Quadro RTX 5000 with Max-Q Design 1EB5 1A58 200A J Quadro RTX 4000 1EB6 J Quadro RTX 4000 with Max-Q Design 1EB6 1028 09C3 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8736 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8738 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8772 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8780 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8782 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8783 J Quadro RTX 4000 with Max-Q Design 1EB6 103C 8785 J Quadro RTX 4000 with Max-Q Design 1EB6 1462 1274 J Quadro RTX 4000 with Max-Q Design 1EB6 1462 1277 J Quadro RTX 4000 with Max-Q Design 1EB6 1462 12B0 J Quadro RTX 4000 with Max-Q Design 1EB6 1462 12C6 J Quadro RTX 4000 with Max-Q Design 1EB6 17AA 22B8 J Quadro RTX 4000 with Max-Q Design 1EB6 17AA 22BA J NVIDIA GeForce RTX 2070 SUPER 1EC2 J NVIDIA GeForce RTX 2070 SUPER 1EC7 J NVIDIA GeForce RTX 2080 1ED0 J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 1025 132D J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 1028 08ED J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 1028 08EE J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 1028 08EF J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 103C 8572 J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 103C 8573 J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 103C 8600 J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 103C 8605 J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 1043 138F J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 1043 15C1 J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 17AA 3FEE J NVIDIA GeForce RTX 2080 with Max-Q Design 1ED0 17AA 3FFE J NVIDIA GeForce RTX 2070 Super 1ED1 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 1025 1432 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 103C 8746 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 103C 878A J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 1043 165F J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 144D C192 J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 17AA 3FCE J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 17AA 3FCF J NVIDIA GeForce RTX 2070 Super with Max-Q Design 1ED1 17AA 3FD0 J NVIDIA GeForce RTX 2080 Super 1ED3 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 1025 1432 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 1028 09D1 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 103C 8746 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 103C 878A J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 1043 1D61 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 1043 1E51 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 1043 1F01 J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 17AA 3FCE J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 17AA 3FCF J NVIDIA GeForce RTX 2080 Super with Max-Q Design 1ED3 17AA 3FD0 J Quadro RTX 5000 1EF5 J NVIDIA GeForce RTX 2070 1F02 J NVIDIA GeForce RTX 2060 1F03 J NVIDIA GeForce RTX 2060 SUPER 1F06 J NVIDIA GeForce RTX 2070 1F07 J NVIDIA GeForce RTX 2060 1F08 J NVIDIA GeForce GTX 1650 1F0A J NVIDIA CMP 40HX 1F0B - NVIDIA GeForce RTX 2070 1F10 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1025 132D J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1025 1342 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08A1 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08A2 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08EA J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08EB J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08EC J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08ED J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08EE J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 08EF J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 093B J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1028 093C J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 103C 8572 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 103C 8573 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 103C 8602 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 103C 8606 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1043 132F J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1043 136F J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1043 1881 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1043 1E6E J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 1658 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 1663 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 1664 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 75A4 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 75A5 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 86A4 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1458 86A5 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1462 1274 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1462 1277 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1558 95E1 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1558 97E1 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1A58 2002 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1A58 2005 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1A58 2007 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1A58 3000 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1A58 3001 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1D05 105E J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1D05 1070 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 1D05 2087 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F10 8086 2087 J NVIDIA GeForce RTX 2060 1F11 J NVIDIA GeForce RTX 2060 1F12 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 1028 098F J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 103C 8741 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 103C 8744 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 103C 878E J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 103C 880E J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 1043 1E11 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 1043 1F11 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 1462 12D9 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 17AA 3801 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 17AA 3802 J NVIDIA GeForce RTX 2060 with Max-Q Design 1F12 17AA 3803 J NVIDIA GeForce RTX 2070 1F14 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1025 1401 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1025 1432 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1025 1442 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1025 1446 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1025 147D J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1028 09E2 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1028 09F3 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 8607 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 86C6 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 86C7 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 8736 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 8738 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 8746 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 8772 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 878A J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 878B J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 87A6 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 103C 87A7 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1043 174F J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 1512 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 75B5 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 75B6 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 76B4 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 76B5 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 78A4 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 78A5 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 86B4 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1458 86B5 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1462 12AE J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1462 12B0 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1462 12C6 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1558 50D3 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1558 70D1 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1A58 200C J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1A58 2011 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F14 1A58 3002 J NVIDIA GeForce RTX 2060 1F15 J Quadro RTX 3000 1F36 J Quadro RTX 3000 with Max-Q Design 1F36 1028 0990 J Quadro RTX 3000 with Max-Q Design 1F36 103C 8736 J Quadro RTX 3000 with Max-Q Design 1F36 103C 8738 J Quadro RTX 3000 with Max-Q Design 1F36 103C 8772 J Quadro RTX 3000 with Max-Q Design 1F36 1043 13CF J Quadro RTX 3000 with Max-Q Design 1F36 1414 0032 J NVIDIA GeForce RTX 2060 SUPER 1F42 J NVIDIA GeForce RTX 2060 SUPER 1F47 J NVIDIA GeForce RTX 2070 1F50 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 1028 08ED J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 1028 08EE J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 1028 08EF J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 103C 8572 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 103C 8573 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 103C 8574 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 103C 8600 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 103C 8605 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 17AA 3FEE J NVIDIA GeForce RTX 2070 with Max-Q Design 1F50 17AA 3FFE J NVIDIA GeForce RTX 2060 1F51 J NVIDIA GeForce RTX 2070 1F54 J NVIDIA GeForce RTX 2070 with Max-Q Design 1F54 103C 878A J NVIDIA GeForce RTX 2070 with Max-Q Design 1F54 17AA 3FCE J NVIDIA GeForce RTX 2070 with Max-Q Design 1F54 17AA 3FCF J NVIDIA GeForce RTX 2070 with Max-Q Design 1F54 17AA 3FD0 J NVIDIA GeForce RTX 2060 1F55 J Quadro RTX 3000 1F76 J Matrox D-Series D2450 1F76 102B 2800 J Matrox D-Series D2480 1F76 102B 2900 J NVIDIA GeForce GTX 1650 1F82 J NVIDIA GeForce GTX 1630 1F83 J NVIDIA GeForce GTX 1650 1F91 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 103C 863E J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 103C 86E7 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 103C 86E8 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1043 12CF J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1043 156F J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1414 0032 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 144D C822 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1462 127E J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1462 1281 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1462 1284 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1462 1285 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1462 129C J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 17AA 229F J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 17AA 3802 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 17AA 3806 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 17AA 3F1A J NVIDIA GeForce GTX 1650 with Max-Q Design 1F91 1A58 1001 J NVIDIA GeForce GTX 1650 Ti 1F95 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1025 1479 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1025 147A J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1025 147B J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1025 147C J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 103C 86E7 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 103C 86E8 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 103C 8815 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1043 1DFF J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1043 1E1F J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 144D C838 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1462 12BD J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1462 12C5 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1462 12D2 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 17AA 22C0 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 17AA 22C1 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 17AA 3837 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 17AA 3F95 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1A58 1003 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1A58 1006 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1A58 1007 J NVIDIA GeForce GTX 1650 Ti with Max-Q Design 1F95 1E83 3E30 J NVIDIA GeForce GTX 1650 1F96 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F96 1462 1297 J NVIDIA GeForce MX450 1F97 J NVIDIA GeForce MX450 1F98 J NVIDIA GeForce GTX 1650 1F99 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1025 1479 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1025 147A J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1025 147B J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1025 147C J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 103C 8815 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1043 13B2 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1043 1402 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1043 1902 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1462 12BD J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1462 12C5 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1462 12D2 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 17AA 22DA J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 17AA 3F93 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F99 1E83 3E30 J NVIDIA GeForce MX450 1F9C J NVIDIA GeForce GTX 1650 1F9D J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1043 128D J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1043 130D J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1043 149C J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1043 185C J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1043 189C J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1462 12F4 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1462 1302 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1462 131B J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1462 1326 J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1462 132A J NVIDIA GeForce GTX 1650 with Max-Q Design 1F9D 1462 132E J NVIDIA GeForce MX550 1F9F J NVIDIA GeForce MX550 1FA0 J NVIDIA T1000 1FB0 1028 12DB J NVIDIA T1000 1FB0 103C 12DB J NVIDIA T1000 1FB0 103C 8A80 J NVIDIA T1000 1FB0 10DE 12DB J NVIDIA DGX Display 1FB0 10DE 1485 J NVIDIA T1000 1FB0 17AA 12DB J NVIDIA T600 1FB1 1028 1488 J NVIDIA T600 1FB1 103C 1488 J NVIDIA T600 1FB1 103C 8A80 J NVIDIA T600 1FB1 10DE 1488 J NVIDIA T600 1FB1 17AA 1488 J NVIDIA T400 1FB2 1028 1489 J NVIDIA T400 1FB2 103C 1489 J NVIDIA T400 1FB2 103C 8A80 J NVIDIA T400 1FB2 10DE 1489 J NVIDIA T400 1FB2 17AA 1489 J NVIDIA T600 Laptop GPU 1FB6 J NVIDIA T550 Laptop GPU 1FB7 J Quadro T2000 1FB8 J Quadro T2000 with Max-Q Design 1FB8 1028 097E J Quadro T2000 with Max-Q Design 1FB8 103C 8736 J Quadro T2000 with Max-Q Design 1FB8 103C 8738 J Quadro T2000 with Max-Q Design 1FB8 103C 8772 J Quadro T2000 with Max-Q Design 1FB8 103C 8780 J Quadro T2000 with Max-Q Design 1FB8 103C 8782 J Quadro T2000 with Max-Q Design 1FB8 103C 8783 J Quadro T2000 with Max-Q Design 1FB8 103C 8785 J Quadro T2000 with Max-Q Design 1FB8 103C 87F0 J Quadro T2000 with Max-Q Design 1FB8 1462 1281 J Quadro T2000 with Max-Q Design 1FB8 1462 12BD J Quadro T2000 with Max-Q Design 1FB8 17AA 22C0 J Quadro T2000 with Max-Q Design 1FB8 17AA 22C1 J Quadro T1000 1FB9 J Quadro T1000 with Max-Q Design 1FB9 1025 1479 J Quadro T1000 with Max-Q Design 1FB9 1025 147A J Quadro T1000 with Max-Q Design 1FB9 1025 147B J Quadro T1000 with Max-Q Design 1FB9 1025 147C J Quadro T1000 with Max-Q Design 1FB9 103C 8736 J Quadro T1000 with Max-Q Design 1FB9 103C 8738 J Quadro T1000 with Max-Q Design 1FB9 103C 8772 J Quadro T1000 with Max-Q Design 1FB9 103C 8780 J Quadro T1000 with Max-Q Design 1FB9 103C 8782 J Quadro T1000 with Max-Q Design 1FB9 103C 8783 J Quadro T1000 with Max-Q Design 1FB9 103C 8785 J Quadro T1000 with Max-Q Design 1FB9 103C 87F0 J Quadro T1000 with Max-Q Design 1FB9 1462 12BD J Quadro T1000 with Max-Q Design 1FB9 17AA 22C0 J Quadro T1000 with Max-Q Design 1FB9 17AA 22C1 J NVIDIA T600 Laptop GPU 1FBA J NVIDIA T500 1FBB J NVIDIA T1200 Laptop GPU 1FBC J NVIDIA GeForce GTX 1650 1FDD J NVIDIA T1000 8GB 1FF0 1028 1612 J NVIDIA T1000 8GB 1FF0 103C 1612 J NVIDIA T1000 8GB 1FF0 103C 8A80 J NVIDIA T1000 8GB 1FF0 10DE 1612 J NVIDIA T1000 8GB 1FF0 17AA 1612 J NVIDIA T400 4GB 1FF2 1028 1613 J NVIDIA T400 4GB 1FF2 103C 1613 J NVIDIA T400E 1FF2 103C 18FF J NVIDIA T400 4GB 1FF2 103C 8A80 J NVIDIA T400 4GB 1FF2 10DE 1613 J NVIDIA T400E 1FF2 10DE 18FF J NVIDIA T400 4GB 1FF2 17AA 1613 J NVIDIA T400E 1FF2 17AA 18FF J Quadro T1000 1FF9 J NVIDIA A100-SXM4-40GB 20B0 J NVIDIA A100-PG509-200 20B0 10DE 1450 J NVIDIA A100-SXM4-80GB 20B2 10DE 1463 J NVIDIA A100-SXM4-80GB 20B2 10DE 147F J NVIDIA A100-SXM4-80GB 20B2 10DE 1622 J NVIDIA A100-SXM4-80GB 20B2 10DE 1623 J NVIDIA PG509-210 20B2 10DE 1625 J NVIDIA A100-SXM-64GB 20B3 10DE 14A7 J NVIDIA A100-SXM-64GB 20B3 10DE 14A8 J NVIDIA A100 80GB PCIe 20B5 10DE 1533 J NVIDIA A100 80GB PCIe 20B5 10DE 1642 J NVIDIA PG506-232 20B6 10DE 1492 J NVIDIA A30 20B7 10DE 1532 J NVIDIA A30 20B7 10DE 1804 J NVIDIA A30 20B7 10DE 1852 J NVIDIA A800-SXM4-40GB 20BD 10DE 17F4 J NVIDIA A100-PCIE-40GB 20F1 10DE 145F J NVIDIA A800-SXM4-80GB 20F3 10DE 179B J NVIDIA A800-SXM4-80GB 20F3 10DE 179C J NVIDIA A800-SXM4-80GB 20F3 10DE 179D J NVIDIA A800-SXM4-80GB 20F3 10DE 179E J NVIDIA A800-SXM4-80GB 20F3 10DE 179F J NVIDIA A800-SXM4-80GB 20F3 10DE 17A0 J NVIDIA A800-SXM4-80GB 20F3 10DE 17A1 J NVIDIA A800-SXM4-80GB 20F3 10DE 17A2 J NVIDIA A800 80GB PCIe 20F5 10DE 1799 J NVIDIA A800 80GB PCIe LC 20F5 10DE 179A J NVIDIA A800 40GB Active 20F6 1028 180A J NVIDIA A800 40GB Active 20F6 103C 180A J NVIDIA A800 40GB Active 20F6 10DE 180A J NVIDIA A800 40GB Active 20F6 17AA 180A J NVIDIA AX800 20FD 10DE 17F8 J NVIDIA GeForce GTX 1660 Ti 2182 J NVIDIA GeForce GTX 1660 2184 J NVIDIA GeForce GTX 1650 SUPER 2187 J NVIDIA GeForce GTX 1650 2188 J NVIDIA CMP 30HX 2189 - NVIDIA GeForce GTX 1660 Ti 2191 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1028 0949 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 85FB J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 85FE J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 86D6 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 8741 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 8744 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 878D J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 87AF J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 103C 87B3 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1043 171F J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1043 17EF J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1043 18D1 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1414 0032 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1462 128A J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1462 128B J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1462 12C6 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1462 12CB J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1462 12CC J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 1462 12D9 J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 17AA 380C J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 17AA 381D J NVIDIA GeForce GTX 1660 Ti with Max-Q Design 2191 17AA 381E J NVIDIA GeForce GTX 1650 Ti 2192 J NVIDIA GeForce GTX 1660 SUPER 21C4 J NVIDIA GeForce GTX 1660 Ti 21D1 J NVIDIA GeForce RTX 3090 Ti 2203 K NVIDIA GeForce RTX 3090 2204 K NVIDIA GeForce RTX 3080 2206 K NVIDIA GeForce RTX 3070 Ti 2207 K NVIDIA GeForce RTX 3080 Ti 2208 K NVIDIA GeForce RTX 3080 220A K NVIDIA CMP 90HX 220D K NVIDIA GeForce RTX 3080 2216 K NVIDIA RTX A6000 2230 1028 1459 K NVIDIA RTX A6000 2230 103C 1459 K NVIDIA RTX A6000 2230 10DE 1459 K NVIDIA RTX A6000 2230 17AA 1459 K NVIDIA RTX A5000 2231 1028 147E K NVIDIA RTX A5000 2231 103C 147E K NVIDIA RTX A5000 2231 10DE 147E K NVIDIA RTX A5000 2231 17AA 147E K NVIDIA RTX A4500 2232 1028 163C K NVIDIA RTX A4500 2232 103C 163C K NVIDIA RTX A4500 2232 10DE 163C K NVIDIA RTX A4500 2232 17AA 163C K NVIDIA RTX A5500 2233 1028 165A K NVIDIA RTX A5500 2233 103C 165A K NVIDIA RTX A5500 2233 10DE 165A K NVIDIA RTX A5500 2233 17AA 165A K NVIDIA A40 2235 10DE 145A K NVIDIA A10 2236 10DE 1482 K NVIDIA A10G 2237 10DE 152F K NVIDIA A10M 2238 10DE 1677 K NVIDIA H100 NVL 2321 10DE 1839 K NVIDIA H800 PCIe 2322 10DE 17A4 K NVIDIA H800 2324 10DE 17A6 K NVIDIA H800 2324 10DE 17A8 K NVIDIA H20 2329 10DE 198B K NVIDIA H20 2329 10DE 198C K NVIDIA H20-3e 232C 10DE 2063 K NVIDIA H100 80GB HBM3 2330 10DE 16C0 K NVIDIA H100 80GB HBM3 2330 10DE 16C1 K NVIDIA H100 PCIe 2331 10DE 1626 K NVIDIA H200 2335 10DE 18BE K NVIDIA H200 2335 10DE 18BF K NVIDIA H100 2339 10DE 17FC K NVIDIA H800 NVL 233A 10DE 183A K NVIDIA H200 NVL 233B 10DE 1996 K NVIDIA GH200 120GB 2342 10DE 16EB K NVIDIA GH200 120GB 2342 10DE 1805 K NVIDIA GH200 480GB 2342 10DE 1809 K NVIDIA GH200 144G HBM3e 2348 10DE 18D2 K NVIDIA GeForce RTX 3060 Ti 2414 K NVIDIA GeForce RTX 3080 Ti Laptop GPU 2420 K NVIDIA RTX A5500 Laptop GPU 2438 K NVIDIA GeForce RTX 3080 Ti Laptop GPU 2460 K NVIDIA GeForce RTX 3070 Ti 2482 K NVIDIA GeForce RTX 3070 2484 K NVIDIA GeForce RTX 3060 Ti 2486 K NVIDIA GeForce RTX 3060 2487 K NVIDIA GeForce RTX 3070 2488 K NVIDIA GeForce RTX 3060 Ti 2489 K NVIDIA CMP 70HX 248A K NVIDIA GeForce RTX 3080 Laptop GPU 249C K NVIDIA GeForce RTX 3060 Laptop GPU 249C 1D05 1194 K NVIDIA GeForce RTX 3070 Laptop GPU 249D K NVIDIA GeForce RTX 3070 Ti Laptop GPU 24A0 K NVIDIA GeForce RTX 3060 Laptop GPU 24A0 1D05 1192 K NVIDIA RTX A4000 24B0 1028 14AD K NVIDIA RTX A4000 24B0 103C 14AD K NVIDIA RTX A4000 24B0 10DE 14AD K NVIDIA RTX A4000 24B0 17AA 14AD K NVIDIA RTX A4000H 24B1 10DE 1658 K NVIDIA RTX A5000 Laptop GPU 24B6 K NVIDIA RTX A4000 Laptop GPU 24B7 K NVIDIA RTX A3000 Laptop GPU 24B8 K NVIDIA RTX A3000 12GB Laptop GPU 24B9 K NVIDIA RTX A4500 Laptop GPU 24BA K NVIDIA RTX A3000 12GB Laptop GPU 24BB K NVIDIA GeForce RTX 3060 24C7 K NVIDIA GeForce RTX 3060 Ti 24C9 K NVIDIA GeForce RTX 3080 Laptop GPU 24DC K NVIDIA GeForce RTX 3070 Laptop GPU 24DD K NVIDIA GeForce RTX 3070 Ti Laptop GPU 24E0 K NVIDIA RTX A4500 Embedded GPU 24FA K NVIDIA GeForce RTX 3060 2503 K NVIDIA GeForce RTX 3060 2504 K NVIDIA GeForce RTX 3050 2507 K NVIDIA GeForce RTX 3050 OEM 2508 K NVIDIA GeForce RTX 3060 Laptop GPU 2520 K NVIDIA GeForce RTX 3060 Laptop GPU 2521 K NVIDIA GeForce RTX 3050 Ti Laptop GPU 2523 K NVIDIA RTX A2000 2531 1028 151D K NVIDIA RTX A2000 2531 103C 151D K NVIDIA RTX A2000 2531 10DE 151D K NVIDIA RTX A2000 2531 17AA 151D K NVIDIA GeForce RTX 3060 2544 K NVIDIA GeForce RTX 3060 Laptop GPU 2560 K NVIDIA GeForce RTX 3050 Ti Laptop GPU 2563 K NVIDIA RTX A2000 12GB 2571 1028 1611 K NVIDIA RTX A2000 12GB 2571 103C 1611 K NVIDIA RTX A2000 12GB 2571 10DE 1611 K NVIDIA RTX A2000 12GB 2571 17AA 1611 K NVIDIA GeForce RTX 3050 2582 K NVIDIA GeForce RTX 3050 2584 K NVIDIA GeForce RTX 3050 Ti Laptop GPU 25A0 K NVIDIA GeForce RTX 3050Ti Laptop GPU 25A0 103C 8928 K NVIDIA GeForce RTX 3050Ti Laptop GPU 25A0 103C 89F9 K NVIDIA GeForce RTX 3060 Laptop GPU 25A0 1D05 1196 K NVIDIA GeForce RTX 3050 Laptop GPU 25A2 K NVIDIA GeForce RTX 3050 Ti Laptop GPU 25A2 1028 0BAF K NVIDIA GeForce RTX 3060 Laptop GPU 25A2 1D05 1195 K NVIDIA GeForce RTX 3050 Laptop GPU 25A5 K NVIDIA GeForce MX570 25A6 K NVIDIA GeForce RTX 2050 25A7 K NVIDIA GeForce RTX 2050 25A9 K NVIDIA GeForce MX570 A 25AA K NVIDIA GeForce RTX 3050 4GB Laptop GPU 25AB K NVIDIA GeForce RTX 3050 6GB Laptop GPU 25AC K NVIDIA GeForce RTX 2050 25AD K NVIDIA RTX A1000 25B0 1028 1878 K NVIDIA RTX A1000 25B0 103C 1878 K NVIDIA RTX A1000 25B0 103C 8D96 K NVIDIA RTX A1000 25B0 10DE 1878 K NVIDIA RTX A1000 25B0 17AA 1878 K NVIDIA RTX A400 25B2 1028 1879 K NVIDIA RTX A400 25B2 103C 1879 K NVIDIA RTX A400 25B2 103C 8D95 K NVIDIA RTX A400 25B2 10DE 1879 K NVIDIA RTX A400 25B2 17AA 1879 K NVIDIA A16 25B6 10DE 14A9 K NVIDIA A2 25B6 10DE 157E K NVIDIA RTX A2000 Laptop GPU 25B8 K NVIDIA RTX A1000 Laptop GPU 25B9 K NVIDIA RTX A2000 8GB Laptop GPU 25BA K NVIDIA RTX A500 Laptop GPU 25BB K NVIDIA RTX A1000 6GB Laptop GPU 25BC K NVIDIA RTX A500 Laptop GPU 25BD K NVIDIA GeForce RTX 3050 Ti Laptop GPU 25E0 K NVIDIA GeForce RTX 3050 Laptop GPU 25E2 K NVIDIA GeForce RTX 3050 Laptop GPU 25E5 K NVIDIA GeForce RTX 3050 6GB Laptop GPU 25EC K NVIDIA GeForce RTX 2050 25ED K NVIDIA RTX A1000 Embedded GPU 25F9 K NVIDIA RTX A2000 Embedded GPU 25FA K NVIDIA RTX A500 Embedded GPU 25FB K NVIDIA GeForce RTX 4090 2684 K NVIDIA GeForce RTX 4090 D 2685 K NVIDIA GeForce RTX 4070 Ti SUPER 2689 K NVIDIA RTX 6000 Ada Generation 26B1 1028 16A1 K NVIDIA RTX 6000 Ada Generation 26B1 103C 16A1 K NVIDIA RTX 6000 Ada Generation 26B1 10DE 16A1 K NVIDIA RTX 6000 Ada Generation 26B1 17AA 16A1 K NVIDIA RTX 5000 Ada Generation 26B2 1028 17FA K NVIDIA RTX 5000 Ada Generation 26B2 103C 17FA K NVIDIA RTX 5000 Ada Generation 26B2 10DE 17FA K NVIDIA RTX 5000 Ada Generation 26B2 17AA 17FA K NVIDIA RTX 5880 Ada Generation 26B3 1028 1934 K NVIDIA RTX 5880 Ada Generation 26B3 103C 1934 K NVIDIA RTX 5880 Ada Generation 26B3 10DE 1934 K NVIDIA RTX 5880 Ada Generation 26B3 17AA 1934 K NVIDIA L40 26B5 10DE 169D K NVIDIA L40 26B5 10DE 17DA K NVIDIA L40S 26B9 10DE 1851 K NVIDIA L40S 26B9 10DE 18CF K NVIDIA L20 26BA 10DE 1957 K NVIDIA L20 26BA 10DE 1990 K NVIDIA GeForce RTX 4080 SUPER 2702 K NVIDIA GeForce RTX 4080 2704 K NVIDIA GeForce RTX 4070 Ti SUPER 2705 K NVIDIA GeForce RTX 4070 2709 K NVIDIA GeForce RTX 4090 Laptop GPU 2717 K NVIDIA RTX 5000 Ada Generation Laptop GPU 2730 K NVIDIA GeForce RTX 4090 Laptop GPU 2757 K NVIDIA RTX 5000 Ada Generation Embedded GPU 2770 K NVIDIA GeForce RTX 4070 Ti 2782 K NVIDIA GeForce RTX 4070 SUPER 2783 K NVIDIA GeForce RTX 4070 2786 K NVIDIA GeForce RTX 4060 Ti 2788 K NVIDIA GeForce RTX 4080 Laptop GPU 27A0 K NVIDIA RTX 4000 SFF Ada Generation 27B0 1028 16FA K NVIDIA RTX 4000 SFF Ada Generation 27B0 103C 16FA K NVIDIA RTX 4000 SFF Ada Generation 27B0 10DE 16FA K NVIDIA RTX 4000 SFF Ada Generation 27B0 17AA 16FA K NVIDIA RTX 4500 Ada Generation 27B1 1028 180C K NVIDIA RTX 4500 Ada Generation 27B1 103C 180C K NVIDIA RTX 4500 Ada Generation 27B1 10DE 180C K NVIDIA RTX 4500 Ada Generation 27B1 17AA 180C K NVIDIA RTX 4000 Ada Generation 27B2 1028 181B K NVIDIA RTX 4000 Ada Generation 27B2 103C 181B K NVIDIA RTX 4000 Ada Generation 27B2 10DE 181B K NVIDIA RTX 4000 Ada Generation 27B2 17AA 181B K NVIDIA L2 27B6 10DE 1933 K NVIDIA L4 27B8 10DE 16CA K NVIDIA L4 27B8 10DE 16EE K NVIDIA RTX 4000 Ada Generation Laptop GPU 27BA K NVIDIA RTX 3500 Ada Generation Laptop GPU 27BB K NVIDIA GeForce RTX 4080 Laptop GPU 27E0 K NVIDIA RTX 3500 Ada Generation Embedded GPU 27FB K NVIDIA GeForce RTX 4060 Ti 2803 K NVIDIA GeForce RTX 4060 Ti 2805 K NVIDIA GeForce RTX 4060 2808 K NVIDIA GeForce RTX 4070 Laptop GPU 2820 K NVIDIA GeForce RTX 3050 A Laptop GPU 2822 K NVIDIA RTX 3000 Ada Generation Laptop GPU 2838 K NVIDIA GeForce RTX 4070 Laptop GPU 2860 K NVIDIA GeForce RTX 4060 2882 K NVIDIA GeForce RTX 4060 Laptop GPU 28A0 K NVIDIA GeForce RTX 4050 Laptop GPU 28A1 K NVIDIA GeForce RTX 3050 A Laptop GPU 28A3 K NVIDIA RTX 2000 Ada Generation 28B0 1028 1870 K NVIDIA RTX 2000 Ada Generation 28B0 103C 1870 K NVIDIA RTX 2000E Ada Generation 28B0 103C 1871 K NVIDIA RTX 2000 Ada Generation 28B0 10DE 1870 K NVIDIA RTX 2000E Ada Generation 28B0 10DE 1871 K NVIDIA RTX 2000 Ada Generation 28B0 17AA 1870 K NVIDIA RTX 2000E Ada Generation 28B0 17AA 1871 K NVIDIA RTX 2000 Ada Generation Laptop GPU 28B8 K NVIDIA RTX 1000 Ada Generation Laptop GPU 28B9 K NVIDIA RTX 500 Ada Generation Laptop GPU 28BA K NVIDIA RTX 500 Ada Generation Laptop GPU 28BB K NVIDIA GeForce RTX 4060 Laptop GPU 28E0 K NVIDIA GeForce RTX 4050 Laptop GPU 28E1 K NVIDIA GeForce RTX 3050 A Laptop GPU 28E3 K NVIDIA RTX 2000 Ada Generation Embedded GPU 28F8 K NVIDIA B200 2901 10DE 1999 - NVIDIA B200 2901 10DE 199B - NVIDIA B200 2901 10DE 20DA - HGX GB200 2941 10DE 2046 - HGX GB200 2941 10DE 20CA - HGX GB200 2941 10DE 20D5 - HGX GB200 2941 10DE 21C9 - HGX GB200 2941 10DE 21CA - NVIDIA GeForce RTX 5090 2B85 - NVIDIA GeForce RTX 5090 D 2B87 - NVIDIA GeForce RTX 5080 2C02 - NVIDIA GeForce RTX 5070 Ti 2C05 - Below are the legacy GPUs that are no longer supported in the unified driver. These GPUs will continue to be maintained through the special legacy NVIDIA GPU driver releases. The 470.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID* VDPAU features ----------------------------------------------- -------------- -------------- GeForce GTX 650 0FC6 D GeForce GT 740 0FC8 D GeForce GT 730 0FC9 C GeForce GT 755M 0FCD D GeForce GT 640M LE 0FCE C GeForce GT 650M 0FD1 D GeForce GT 640M 0FD2 D GeForce GT 640M LE 0FD2 1028 0595 C GeForce GT 640M LE 0FD2 1028 05B2 C GeForce GT 640M LE 0FD3 C GeForce GTX 660M 0FD4 D GeForce GT 650M 0FD5 D GeForce GT 640M 0FD8 D GeForce GT 645M 0FD9 D GeForce GT 740M 0FDF D GeForce GTX 660M 0FE0 D GeForce GT 730M 0FE1 D GeForce GT 745M 0FE2 D GeForce GT 745M 0FE3 D GeForce GT 745A 0FE3 103C 2B16 D GeForce GT 745A 0FE3 17AA 3675 D GeForce GT 750M 0FE4 D GeForce GT 750M 0FE9 D GeForce GT 755M 0FEA D GeForce 710A 0FEC C GeForce 820M 0FED C GeForce 810M 0FEE C Quadro K1100M 0FF6 D Quadro K500M 0FF8 D Quadro K2000D 0FF9 D Quadro K600 0FFA D Quadro K2000M 0FFB D Quadro K1000M 0FFC D NVS 510 0FFD D Quadro K2000 0FFE D Quadro 410 0FFF D GeForce GTX TITAN Z 1001 D GeForce GTX 780 1004 D GeForce GTX TITAN 1005 D GeForce GTX 780 1007 D GeForce GTX 780 Ti 1008 D GeForce GTX 780 Ti 100A D GeForce GTX TITAN Black 100C D Tesla K20Xm 1021 D Tesla K20c 1022 D Tesla K40m 1023 D Tesla K40c 1024 D Tesla K20s 1026 D Tesla K40st 1027 D Tesla K20m 1028 D Tesla K40s 1029 D Tesla K40t 102A D Tesla K80 102D D Quadro K6000 103A D Quadro K5200 103C D GeForce GTX 680 1180 D GeForce GTX 660 Ti 1183 D GeForce GTX 770 1184 D GeForce GTX 660 1185 D GeForce GTX 760 1185 10DE 106F D GeForce GTX 760 1187 D GeForce GTX 690 1188 D GeForce GTX 670 1189 D GeForce GTX 760 Ti OEM 1189 10DE 1074 D GRID K520 118A D GeForce GTX 760 (192-bit) 118E D Tesla K10 118F D GeForce GTX 760 Ti OEM 1193 D Tesla K8 1194 D GeForce GTX 660 1195 D GeForce GTX 880M 1198 D GeForce GTX 870M 1199 D GeForce GTX 760 1199 1458 D001 D GeForce GTX 860M 119A D GeForce GTX 775M 119D D GeForce GTX 780M 119E D GeForce GTX 780M 119F D GeForce GTX 680M 11A0 D GeForce GTX 670MX 11A1 D GeForce GTX 675MX 11A2 D GeForce GTX 680MX 11A3 D GeForce GTX 675MX 11A7 D Quadro K4200 11B4 D Quadro K3100M 11B6 D Quadro K4100M 11B7 D Quadro K5100M 11B8 D Quadro K5000 11BA D Quadro K5000M 11BC D Quadro K4000M 11BD D Quadro K3000M 11BE D GeForce GTX 660 11C0 D GeForce GTX 650 Ti BOOST 11C2 D GeForce GTX 650 Ti 11C3 D GeForce GTX 645 11C4 D GeForce GT 740 11C5 D GeForce GTX 650 Ti 11C6 D GeForce GTX 650 11C8 D GeForce GT 740 11CB D GeForce GTX 770M 11E0 D GeForce GTX 765M 11E1 D GeForce GTX 765M 11E2 D GeForce GTX 760M 11E3 D GeForce GTX 760A 11E3 17AA 3683 D Quadro K4000 11FA D Quadro K2100M 11FC D GeForce GT 635 1280 D GeForce GT 710 1281 D GeForce GT 640 1282 C GeForce GT 630 1284 C GeForce GT 720 1286 D GeForce GT 730 1287 C GeForce GT 720 1288 D GeForce GT 710 1289 D GeForce GT 710 128B D GeForce GT 730M 1290 D GeForce GT 730A 1290 103C 2AFA D GeForce GT 735M 1291 D GeForce GT 740M 1292 D GeForce GT 740A 1292 17AA 3675 D GeForce GT 740A 1292 17AA 367C D GeForce GT 740A 1292 17AA 3684 D GeForce GT 730M 1293 D GeForce 710M 1295 C GeForce 710A 1295 103C 2B0D C GeForce 710A 1295 103C 2B0F C GeForce 810A 1295 103C 2B20 D GeForce 810A 1295 103C 2B21 D GeForce 805A 1295 17AA 367A D GeForce 710A 1295 17AA 367C C GeForce 825M 1296 D GeForce GT 720M 1298 C GeForce 920M 1299 D GeForce 920A 1299 17AA 30BB D GeForce 920A 1299 17AA 30DA D GeForce 920A 1299 17AA 30DC D GeForce 920A 1299 17AA 30DD D GeForce 920A 1299 17AA 30DF D GeForce 920A 1299 17AA 3117 D GeForce 920A 1299 17AA 361B D GeForce 920A 1299 17AA 362D D GeForce 920A 1299 17AA 362E D GeForce 920A 1299 17AA 3630 D GeForce 920A 1299 17AA 3637 D GeForce 920A 1299 17AA 369B D GeForce 920A 1299 17AA 36A7 D GeForce 920A 1299 17AA 36AF D GeForce 920A 1299 17AA 36F0 D GeForce GT 730 1299 1B0A 01C6 C GeForce 910M 129A D Quadro K610M 12B9 D Quadro K510M 12BA D The 390.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID* VDPAU features ----------------------------------------------- -------------- -------------- GeForce GTX 480 06C0 C GeForce GTX 465 06C4 C GeForce GTX 480M 06CA C GeForce GTX 470 06CD C Tesla C2050 / C2070 06D1 C Tesla C2050 06D1 10DE 0771 C Tesla C2070 06D1 10DE 0772 C Tesla M2070 06D2 C Tesla X2070 06D2 10DE 088F C Quadro 6000 06D8 C Quadro 5000 06D9 C Quadro 5000M 06DA C Quadro 6000 06DC C Quadro 4000 06DD C Tesla T20 Processor 06DE C Tesla S2050 06DE 10DE 0773 C Tesla M2050 06DE 10DE 082F C Tesla X2070 06DE 10DE 0840 C Tesla M2050 06DE 10DE 0842 C Tesla M2050 06DE 10DE 0846 C Tesla M2050 06DE 10DE 0866 C Tesla M2050 06DE 10DE 0907 C Tesla M2050 06DE 10DE 091E C Tesla M2070-Q 06DF C GeForce GT 440 0DC0 C GeForce GTS 450 0DC4 C GeForce GTS 450 0DC5 C GeForce GTS 450 0DC6 C GeForce GT 555M 0DCD C GeForce GT 555M 0DCE C GeForce GTX 460M 0DD1 C GeForce GT 445M 0DD2 C GeForce GT 435M 0DD3 C GeForce GT 550M 0DD6 C Quadro 2000 0DD8 C Quadro 2000D 0DD8 10DE 0914 C Quadro 2000M 0DDA C GeForce GT 440 0DE0 C GeForce GT 430 0DE1 C GeForce GT 420 0DE2 C GeForce GT 635M 0DE3 C GeForce GT 520 0DE4 C GeForce GT 530 0DE5 C GeForce GT 610 0DE7 C GeForce GT 620M 0DE8 C GeForce GT 630M 0DE9 C GeForce GT 620M 0DE9 1025 0692 C GeForce GT 620M 0DE9 1025 0725 C GeForce GT 620M 0DE9 1025 0728 C GeForce GT 620M 0DE9 1025 072B C GeForce GT 620M 0DE9 1025 072E C GeForce GT 620M 0DE9 1025 0753 C GeForce GT 620M 0DE9 1025 0754 C GeForce GT 640M LE 0DE9 17AA 3977 C GeForce GT 635M 0DE9 1B0A 2210 C GeForce 610M 0DEA C GeForce 615 0DEA 17AA 365A C GeForce 615 0DEA 17AA 365B C GeForce 615 0DEA 17AA 365E C GeForce 615 0DEA 17AA 3660 C GeForce 615 0DEA 17AA 366C C GeForce GT 555M 0DEB C GeForce GT 525M 0DEC C GeForce GT 520M 0DED C GeForce GT 415M 0DEE C NVS 5400M 0DEF C GeForce GT 425M 0DF0 C GeForce GT 420M 0DF1 C GeForce GT 435M 0DF2 C GeForce GT 420M 0DF3 C GeForce GT 540M 0DF4 C GeForce GT 630M 0DF4 152D 0952 C GeForce GT 630M 0DF4 152D 0953 C GeForce GT 525M 0DF5 C GeForce GT 550M 0DF6 C GeForce GT 520M 0DF7 C Quadro 600 0DF8 C Quadro 500M 0DF9 C Quadro 1000M 0DFA C NVS 5200M 0DFC C GeForce GTX 460 0E22 C GeForce GTX 460 SE 0E23 C GeForce GTX 460 0E24 C GeForce GTX 470M 0E30 C GeForce GTX 485M 0E31 C Quadro 3000M 0E3A C Quadro 4000M 0E3B C GeForce GT 630 0F00 C GeForce GT 620 0F01 C GeForce GT 730 0F02 C GeForce GT 610 0F03 C GeForce GT 520 1040 C GeForce 510 1042 D GeForce 605 1048 D GeForce GT 620 1049 C GeForce GT 610 104A C GeForce GT 625 (OEM) 104B D GeForce GT 625 104B 1043 844C D GeForce GT 625 104B 1043 846B D GeForce GT 625 104B 1462 B590 D GeForce GT 625 104B 174B 0625 D GeForce GT 625 104B 174B A625 D GeForce GT 705 104C D GeForce GT 520M 1050 C GeForce GT 520MX 1051 D GeForce GT 520M 1052 C GeForce 410M 1054 D GeForce 410M 1055 D NVS 4200M 1056 D NVS 4200M 1057 D GeForce 610M 1058 C GeForce 610 1058 103C 2AF1 D GeForce 800A 1058 17AA 3682 D GeForce 705A 1058 17AA 3692 C GeForce 800A 1058 17AA 3695 D GeForce 800A 1058 17AA 36A8 D GeForce 800A 1058 17AA 36AC D GeForce 800A 1058 17AA 36AD D GeForce 800A 1058 705A 3682 D GeForce 610M 1059 C GeForce 610M 105A C GeForce 705M 105B C GeForce 705A 105B 103C 2AFB C GeForce 800A 105B 17AA 30B1 D GeForce 705A 105B 17AA 30F3 C GeForce 800A 105B 17AA 36A1 D NVS 315 107C D NVS 310 107D D GeForce GTX 580 1080 C GeForce GTX 570 1081 C GeForce GTX 560 Ti 1082 C GeForce GTX 560 1084 C GeForce GTX 570 1086 C GeForce GTX 560 Ti 1087 C GeForce GTX 590 1088 C GeForce GTX 580 1089 C GeForce GTX 580 108B C Tesla M2090 1091 C Tesla X2090 1091 10DE 088E C Tesla X2090 1091 10DE 0891 C Tesla X2090 1091 10DE 0974 C Tesla X2090 1091 10DE 098D C Tesla M2075 1094 C Tesla C2075 1096 C Tesla C2050 1096 10DE 0911 C Quadro 5010M 109A C Quadro 7000 109B C GeForce 820M 1140 1019 0799 C GeForce GT 720M 1140 1019 999F C GeForce GT 620M 1140 1025 0600 C GeForce GT 620M 1140 1025 0606 C GeForce GT 620M 1140 1025 064A C GeForce GT 620M 1140 1025 064C C GeForce GT 620M 1140 1025 067A C GeForce GT 620M 1140 1025 0680 C GeForce 710M 1140 1025 0686 C GeForce 710M 1140 1025 0689 C GeForce 710M 1140 1025 068B C GeForce 710M 1140 1025 068D C GeForce 710M 1140 1025 068E C GeForce 710M 1140 1025 0691 C GeForce GT 620M 1140 1025 0692 C GeForce GT 620M 1140 1025 0694 C GeForce GT 620M 1140 1025 0702 C GeForce GT 620M 1140 1025 0719 C GeForce GT 620M 1140 1025 0725 C GeForce GT 620M 1140 1025 0728 C GeForce GT 620M 1140 1025 072B C GeForce GT 620M 1140 1025 072E C GeForce GT 620M 1140 1025 0732 C GeForce GT 720M 1140 1025 0763 C GeForce 710M 1140 1025 0773 C GeForce 710M 1140 1025 0774 C GeForce GT 720M 1140 1025 0776 C GeForce 710M 1140 1025 077A C GeForce 710M 1140 1025 077B C GeForce 710M 1140 1025 077C C GeForce 710M 1140 1025 077D C GeForce 710M 1140 1025 077E C GeForce 710M 1140 1025 077F C GeForce GT 720M 1140 1025 0781 C GeForce GT 720M 1140 1025 0798 C GeForce GT 720M 1140 1025 0799 C GeForce GT 720M 1140 1025 079B C GeForce GT 720M 1140 1025 079C C GeForce GT 720M 1140 1025 0807 C GeForce 820M 1140 1025 0821 C GeForce GT 720M 1140 1025 0823 C GeForce GT 720M 1140 1025 0830 C GeForce GT 720M 1140 1025 0833 C GeForce GT 720M 1140 1025 0837 C GeForce 820M 1140 1025 083E C GeForce 710M 1140 1025 0841 C GeForce 820M 1140 1025 0853 C GeForce 820M 1140 1025 0854 C GeForce 820M 1140 1025 0855 C GeForce 820M 1140 1025 0856 C GeForce 820M 1140 1025 0857 C GeForce 820M 1140 1025 0858 C GeForce 820M 1140 1025 0863 C GeForce 820M 1140 1025 0868 C GeForce 810M 1140 1025 0869 C GeForce 820M 1140 1025 0873 C GeForce 820M 1140 1025 0878 C GeForce 820M 1140 1025 087B C GeForce 820M 1140 1025 087F C GeForce 820M 1140 1025 0881 C GeForce 820M 1140 1025 0885 C GeForce 820M 1140 1025 088A C GeForce 820M 1140 1025 089B C GeForce 820M 1140 1025 0921 C GeForce 810M 1140 1025 092E C GeForce 820M 1140 1025 092F C GeForce 820M 1140 1025 0932 C GeForce 820M 1140 1025 093A C GeForce 820M 1140 1025 093C C GeForce 820M 1140 1025 093F C GeForce 820M 1140 1025 0941 C GeForce 820M 1140 1025 0945 C GeForce 820M 1140 1025 0954 C GeForce 820M 1140 1025 0965 C GeForce GT 630M 1140 1028 054D C GeForce GT 630M 1140 1028 054E C GeForce GT 620M 1140 1028 0554 C GeForce GT 620M 1140 1028 0557 C GeForce GT 625M 1140 1028 0562 C GeForce GT 630M 1140 1028 0565 C GeForce GT 630M 1140 1028 0568 C GeForce GT 630M 1140 1028 0590 C GeForce GT 625M 1140 1028 0592 C GeForce GT 625M 1140 1028 0594 C GeForce GT 625M 1140 1028 0595 C GeForce GT 625M 1140 1028 05A2 C GeForce GT 625M 1140 1028 05B1 C GeForce GT 625M 1140 1028 05B3 C GeForce GT 630M 1140 1028 05DA C GeForce GT 720M 1140 1028 05DE C GeForce GT 720M 1140 1028 05E0 C GeForce GT 630M 1140 1028 05E8 C GeForce GT 720M 1140 1028 05F4 C GeForce GT 720M 1140 1028 060F C GeForce GT 720M 1140 1028 062F C GeForce 820M 1140 1028 064E C GeForce 820M 1140 1028 0652 C GeForce 820M 1140 1028 0653 C GeForce 820M 1140 1028 0655 C GeForce 820M 1140 1028 065E C GeForce 820M 1140 1028 0662 C GeForce 820M 1140 1028 068D C GeForce 820M 1140 1028 06AD C GeForce 820M 1140 1028 06AE C GeForce 820M 1140 1028 06AF C GeForce 820M 1140 1028 06B0 C GeForce 820M 1140 1028 06C0 C GeForce 820M 1140 1028 06C1 C GeForce GT 630M 1140 103C 18EF C GeForce GT 630M 1140 103C 18F9 C GeForce GT 630M 1140 103C 18FB C GeForce GT 630M 1140 103C 18FD C GeForce GT 630M 1140 103C 18FF C GeForce 820M 1140 103C 218A C GeForce 820M 1140 103C 21BB C GeForce 820M 1140 103C 21BC C GeForce 820M 1140 103C 220E C GeForce 820M 1140 103C 2210 C GeForce 820M 1140 103C 2212 C GeForce 820M 1140 103C 2214 C GeForce 820M 1140 103C 2218 C GeForce 820M 1140 103C 225B C GeForce 820M 1140 103C 225D C GeForce 820M 1140 103C 226D C GeForce 820M 1140 103C 226F C GeForce 820M 1140 103C 22D2 C GeForce 820M 1140 103C 22D9 C GeForce 820M 1140 103C 2335 C GeForce 820M 1140 103C 2337 C GeForce GT 720A 1140 103C 2AEF C GeForce 710A 1140 103C 2AF9 C NVS 5200M 1140 1043 10DD C NVS 5200M 1140 1043 10ED C GeForce GT 720M 1140 1043 11FD C GeForce GT 720M 1140 1043 124D C GeForce GT 720M 1140 1043 126D C GeForce GT 720M 1140 1043 131D C GeForce GT 720M 1140 1043 13FD C GeForce GT 720M 1140 1043 14C7 C GeForce GT 620M 1140 1043 1507 C GeForce 820M 1140 1043 15AD C GeForce 820M 1140 1043 15ED C GeForce 820M 1140 1043 160D C GeForce 820M 1140 1043 163D C GeForce 820M 1140 1043 165D C GeForce 820M 1140 1043 166D C GeForce 820M 1140 1043 16CD C GeForce 820M 1140 1043 16DD C GeForce 820M 1140 1043 170D C GeForce 820M 1140 1043 176D C GeForce 820M 1140 1043 178D C GeForce 820M 1140 1043 179D C GeForce GT 620M 1140 1043 2132 C NVS 5200M 1140 1043 2136 C GeForce GT 720M 1140 1043 21BA C GeForce GT 720M 1140 1043 21FA C GeForce GT 720M 1140 1043 220A C GeForce GT 720M 1140 1043 221A C GeForce GT 710M 1140 1043 223A C GeForce GT 710M 1140 1043 224A C GeForce 820M 1140 1043 227A C GeForce 820M 1140 1043 228A C GeForce 820M 1140 1043 22FA C GeForce 820M 1140 1043 232A C GeForce 820M 1140 1043 233A C GeForce 820M 1140 1043 235A C GeForce 820M 1140 1043 236A C GeForce 820M 1140 1043 238A C GeForce GT 720M 1140 1043 8595 C GeForce GT 720M 1140 1043 85EA C GeForce 820M 1140 1043 85EB C GeForce 820M 1140 1043 85EC C GeForce GT 720M 1140 1043 85EE C GeForce 820M 1140 1043 85F3 C GeForce 820M 1140 1043 860E C GeForce 820M 1140 1043 861A C GeForce 820M 1140 1043 861B C GeForce 820M 1140 1043 8628 C GeForce 820M 1140 1043 8643 C GeForce 820M 1140 1043 864C C GeForce 820M 1140 1043 8652 C GeForce 820M 1140 1043 8660 C GeForce 820M 1140 1043 8661 C GeForce GT 720M 1140 105B 0DAC C GeForce GT 720M 1140 105B 0DAD C GeForce GT 720M 1140 105B 0EF3 C GeForce GT 720M 1140 10CF 17F5 C GeForce 710M 1140 1179 FA01 C GeForce 710M 1140 1179 FA02 C GeForce 710M 1140 1179 FA03 C GeForce 710M 1140 1179 FA05 C GeForce 710M 1140 1179 FA11 C GeForce 710M 1140 1179 FA13 C GeForce 710M 1140 1179 FA18 C GeForce 710M 1140 1179 FA19 C GeForce 710M 1140 1179 FA21 C GeForce 710M 1140 1179 FA23 C GeForce 710M 1140 1179 FA2A C GeForce 710M 1140 1179 FA32 C GeForce 710M 1140 1179 FA33 C GeForce 710M 1140 1179 FA36 C GeForce 710M 1140 1179 FA38 C GeForce 710M 1140 1179 FA42 C GeForce 710M 1140 1179 FA43 C GeForce 710M 1140 1179 FA45 C GeForce 710M 1140 1179 FA47 C GeForce 710M 1140 1179 FA49 C GeForce 710M 1140 1179 FA58 C GeForce 710M 1140 1179 FA59 C GeForce 710M 1140 1179 FA88 C GeForce 710M 1140 1179 FA89 C GeForce GT 620M 1140 144D B092 C GeForce GT 630M 1140 144D C0D5 C GeForce GT 620M 1140 144D C0D7 C NVS 5200M 1140 144D C0E2 C NVS 5200M 1140 144D C0E3 C NVS 5200M 1140 144D C0E4 C GeForce 820M 1140 144D C10D C GeForce GT 620M 1140 144D C652 C GeForce 710M 1140 144D C709 C GeForce 710M 1140 144D C711 C GeForce 710M 1140 144D C736 C GeForce 710M 1140 144D C737 C GeForce 820M 1140 144D C745 C GeForce 820M 1140 144D C750 C GeForce GT 710M 1140 1462 10B8 C GeForce GT 720M 1140 1462 10E9 C GeForce 820M 1140 1462 1116 C GeForce 720M 1140 1462 AA33 C GeForce GT 720M 1140 1462 AAA2 C GeForce 820M 1140 1462 AAA3 C GeForce GT 720M 1140 1462 ACB2 C GeForce GT 720M 1140 1462 ACC1 C GeForce 720M 1140 1462 AE61 C GeForce GT 720M 1140 1462 AE65 C GeForce 820M 1140 1462 AE6A C GeForce GT 720M 1140 1462 AE71 C GeForce 820M 1140 14C0 0083 C GeForce 620M 1140 152D 0926 C GeForce GT 630M 1140 152D 0982 C GeForce GT 630M 1140 152D 0983 C GeForce GT 820M 1140 152D 1005 C GeForce 710M 1140 152D 1012 C GeForce 820M 1140 152D 1019 C GeForce GT 630M 1140 152D 1030 C GeForce 710M 1140 152D 1055 C GeForce GT 720M 1140 152D 1067 C GeForce 820M 1140 152D 1092 C NVS 5200M 1140 17AA 2200 C GeForce GT 720M 1140 17AA 2213 C GeForce GT 720M 1140 17AA 2220 C GeForce GT 720A 1140 17AA 309C C GeForce 820A 1140 17AA 30B4 C GeForce 720A 1140 17AA 30B7 C GeForce 820A 1140 17AA 30E4 C GeForce 820A 1140 17AA 361B C GeForce 820A 1140 17AA 361C C GeForce 820A 1140 17AA 361D C GeForce GT 620M 1140 17AA 3656 C GeForce 705M 1140 17AA 365A C GeForce 800M 1140 17AA 365E C GeForce 820A 1140 17AA 3661 C GeForce 800M 1140 17AA 366C C GeForce 800M 1140 17AA 3685 C GeForce 800M 1140 17AA 3686 C GeForce 705A 1140 17AA 3687 C GeForce 820A 1140 17AA 3696 C GeForce 820A 1140 17AA 369B C GeForce 820A 1140 17AA 369C C GeForce 820A 1140 17AA 369D C GeForce 820A 1140 17AA 369E C GeForce 820A 1140 17AA 36A6 C GeForce 820A 1140 17AA 36A7 C GeForce 820A 1140 17AA 36A9 C GeForce 820A 1140 17AA 36AF C GeForce 820A 1140 17AA 36B0 C GeForce 820A 1140 17AA 36B6 C GeForce GT 720M 1140 17AA 3800 C GeForce GT 720M 1140 17AA 3801 C GeForce GT 720M 1140 17AA 3802 C GeForce GT 720M 1140 17AA 3803 C GeForce GT 720M 1140 17AA 3804 C GeForce GT 720M 1140 17AA 3806 C GeForce GT 720M 1140 17AA 3808 C GeForce GT 820M 1140 17AA 380D C GeForce GT 820M 1140 17AA 380E C GeForce GT 820M 1140 17AA 380F C GeForce GT 820M 1140 17AA 3811 C GeForce 820M 1140 17AA 3812 C GeForce 820M 1140 17AA 3813 C GeForce 820M 1140 17AA 3816 C GeForce 820M 1140 17AA 3817 C GeForce 820M 1140 17AA 3818 C GeForce 820M 1140 17AA 381A C GeForce 820M 1140 17AA 381C C GeForce 820M 1140 17AA 381D C GeForce 610M 1140 17AA 3901 C GeForce 710M 1140 17AA 3902 C GeForce 710M 1140 17AA 3903 C GeForce GT 625M 1140 17AA 3904 C GeForce GT 720M 1140 17AA 3905 C GeForce 820M 1140 17AA 3907 C GeForce GT 720M 1140 17AA 3910 C GeForce GT 720M 1140 17AA 3912 C GeForce 820M 1140 17AA 3913 C GeForce 820M 1140 17AA 3915 C GeForce 610M 1140 17AA 3983 C GeForce 610M 1140 17AA 5001 C GeForce GT 720M 1140 17AA 5003 C GeForce 705M 1140 17AA 5005 C GeForce GT 620M 1140 17AA 500D C GeForce 710M 1140 17AA 5014 C GeForce 710M 1140 17AA 5017 C GeForce 710M 1140 17AA 5019 C GeForce 710M 1140 17AA 501A C GeForce GT 720M 1140 17AA 501F C GeForce 710M 1140 17AA 5025 C GeForce 710M 1140 17AA 5027 C GeForce 710M 1140 17AA 502A C GeForce GT 720M 1140 17AA 502B C GeForce 710M 1140 17AA 502D C GeForce GT 720M 1140 17AA 502E C GeForce GT 720M 1140 17AA 502F C GeForce 705M 1140 17AA 5030 C GeForce 705M 1140 17AA 5031 C GeForce 820M 1140 17AA 5032 C GeForce 820M 1140 17AA 5033 C GeForce 710M 1140 17AA 503E C GeForce 820M 1140 17AA 503F C GeForce 820M 1140 17AA 5040 C GeForce 710M 1140 1854 0177 C GeForce 710M 1140 1854 0180 C GeForce GT 720M 1140 1854 0190 C GeForce GT 720M 1140 1854 0192 C GeForce 820M 1140 1854 0224 C GeForce 820M 1140 1B0A 01C0 C GeForce GT 620M 1140 1B0A 20DD C GeForce GT 620M 1140 1B0A 20DF C GeForce 820M 1140 1B0A 210E C GeForce GT 720M 1140 1B0A 2202 C GeForce 820M 1140 1B0A 90D7 C GeForce 820M 1140 1B0A 90DD C GeForce 820M 1140 1B50 5530 C GeForce GT 720M 1140 1B6C 5031 C GeForce 820M 1140 1BAB 0106 C GeForce 810M 1140 1D05 1013 C GeForce GTX 560 Ti 1200 C GeForce GTX 560 1201 C GeForce GTX 460 SE v2 1203 C GeForce GTX 460 v2 1205 C GeForce GTX 555 1206 C GeForce GT 645 1207 C GeForce GTX 560 SE 1208 C GeForce GTX 570M 1210 C GeForce GTX 580M 1211 C GeForce GTX 675M 1212 C GeForce GTX 670M 1213 C GeForce GT 545 1241 C GeForce GT 545 1243 C GeForce GTX 550 Ti 1244 C GeForce GTS 450 1245 C GeForce GT 550M 1246 C GeForce GT 555M 1247 C GeForce GT 635M 1247 1043 212A C GeForce GT 635M 1247 1043 212B C GeForce GT 635M 1247 1043 212C C GeForce GT 555M 1248 C GeForce GTS 450 1249 C GeForce GT 640 124B C GeForce GT 555M 124D C GeForce GT 635M 124D 1462 10CC C GeForce GTX 560M 1251 C The 367.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID* VDPAU features ----------------------------------------------- -------------- -------------- GRID K340 0FEF D GRID K1 0FF2 D GRID K2 11BF D The 340.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID* VDPAU features ----------------------------------------------- -------------- -------------- GeForce 8800 GTX 0191 - GeForce 8800 GTS 0193 - GeForce 8800 Ultra 0194 - Tesla C870 0197 - Quadro FX 5600 019D - Quadro FX 4600 019E - GeForce 8600 GTS 0400 A GeForce 8600 GT 0401 A GeForce 8600 GT 0402 A GeForce 8600 GS 0403 A GeForce 8400 GS 0404 A GeForce 9500M GS 0405 A GeForce 8300 GS 0406 - GeForce 8600M GT 0407 A GeForce 9650M GS 0408 A GeForce 8700M GT 0409 A Quadro FX 370 040A A Quadro NVS 320M 040B A Quadro FX 570M 040C A Quadro FX 1600M 040D A Quadro FX 570 040E A Quadro FX 1700 040F A GeForce GT 330 0410 A GeForce 8400 SE 0420 - GeForce 8500 GT 0421 A GeForce 8400 GS 0422 A GeForce 8300 GS 0423 - GeForce 8400 GS 0424 A GeForce 8600M GS 0425 A GeForce 8400M GT 0426 A GeForce 8400M GS 0427 A GeForce 8400M G 0428 A Quadro NVS 140M 0429 A Quadro NVS 130M 042A A Quadro NVS 135M 042B A GeForce 9400 GT 042C A Quadro FX 360M 042D A GeForce 9300M G 042E A Quadro NVS 290 042F A GeForce GTX 295 05E0 A GeForce GTX 280 05E1 A GeForce GTX 260 05E2 A GeForce GTX 285 05E3 A GeForce GTX 275 05E6 A Tesla C1060 05E7 A Tesla T10 Processor 05E7 10DE 0595 A Tesla T10 Processor 05E7 10DE 068F A Tesla M1060 05E7 10DE 0697 A Tesla M1060 05E7 10DE 0714 A Tesla M1060 05E7 10DE 0743 A GeForce GTX 260 05EA A GeForce GTX 295 05EB A Quadroplex 2200 D2 05ED A Quadroplex 2200 S4 05F8 A Quadro CX 05F9 A Quadro FX 5800 05FD A Quadro FX 4800 05FE A Quadro FX 3800 05FF A GeForce 8800 GTS 512 0600 A GeForce 9800 GT 0601 A GeForce 8800 GT 0602 A GeForce GT 230 0603 A GeForce 9800 GX2 0604 A GeForce 9800 GT 0605 A GeForce 8800 GS 0606 A GeForce GTS 240 0607 A GeForce 9800M GTX 0608 A GeForce 8800M GTS 0609 A GeForce 8800 GS 0609 106B 00A7 A GeForce GTX 280M 060A A GeForce 9800M GT 060B A GeForce 8800M GTX 060C A GeForce 8800 GS 060D A GeForce GTX 285M 060F A GeForce 9600 GSO 0610 A GeForce 8800 GT 0611 A GeForce 9800 GTX/9800 GTX+ 0612 A GeForce 9800 GTX+ 0613 A GeForce 9800 GT 0614 A GeForce GTS 250 0615 A GeForce 9800M GTX 0617 A GeForce GTX 260M 0618 A Quadro FX 4700 X2 0619 A Quadro FX 3700 061A A Quadro VX 200 061B A Quadro FX 3600M 061C A Quadro FX 2800M 061D A Quadro FX 3700M 061E A Quadro FX 3800M 061F A GeForce GT 230 0621 A GeForce 9600 GT 0622 A GeForce 9600 GS 0623 A GeForce 9600 GSO 512 0625 A GeForce GT 130 0626 A GeForce GT 140 0627 A GeForce 9800M GTS 0628 A GeForce 9700M GTS 062A A GeForce 9800M GS 062B A GeForce 9800M GTS 062C A GeForce 9600 GT 062D A GeForce 9600 GT 062E A GeForce GT 130 062E 106B 0605 A GeForce 9700 S 0630 A GeForce GTS 160M 0631 A GeForce GTS 150M 0632 A GeForce 9600 GSO 0635 A GeForce 9600 GT 0637 A Quadro FX 1800 0638 A Quadro FX 2700M 063A A GeForce 9500 GT 0640 A GeForce 9400 GT 0641 A GeForce 9500 GT 0643 A GeForce 9500 GS 0644 A GeForce 9500 GS 0645 A GeForce GT 120 0646 A GeForce 9600M GT 0647 A GeForce 9600M GS 0648 A GeForce 9600M GT 0649 A GeForce GT 220M 0649 1043 202D A GeForce 9700M GT 064A A GeForce 9500M G 064B A GeForce 9650M GT 064C A GeForce G 110M 0651 A GeForce GT 130M 0652 A GeForce GT 240M LE 0652 152D 0850 A GeForce GT 120M 0653 A GeForce GT 220M 0654 A GeForce GT 320M 0654 1043 14A2 A GeForce GT 320M 0654 1043 14D2 A GeForce GT 120 0655 106B 0633 A GeForce GT 120 0656 106B 0693 A Quadro FX 380 0658 A Quadro FX 580 0659 A Quadro FX 1700M 065A A GeForce 9400 GT 065B A Quadro FX 770M 065C A GeForce 9300 GE 06E0 B GeForce 9300 GS 06E1 B GeForce 8400 06E2 B GeForce 8400 SE 06E3 - GeForce 8400 GS 06E4 A GeForce 9300M GS 06E5 B GeForce G100 06E6 B GeForce 9300 SE 06E7 - GeForce 9200M GS 06E8 B GeForce 9200M GE 06E8 103C 360B B GeForce 9300M GS 06E9 B Quadro NVS 150M 06EA B Quadro NVS 160M 06EB B GeForce G 105M 06EC B GeForce G 103M 06EF B GeForce G105M 06F1 B Quadro NVS 420 06F8 B Quadro FX 370 LP 06F9 B Quadro FX 370 Low Profile 06F9 10DE 060D B Quadro NVS 450 06FA B Quadro FX 370M 06FB B Quadro NVS 295 06FD B HICx16 + Graphics 06FF B HICx8 + Graphics 06FF 10DE 0711 B GeForce 8200M 0840 B GeForce 9100M G 0844 B GeForce 8200M G 0845 B GeForce 9200 0846 B GeForce 9100 0847 B GeForce 8300 0848 B GeForce 8200 0849 B nForce 730a 084A B GeForce 9200 084B B nForce 980a/780a SLI 084C B nForce 750a SLI 084D B GeForce 8100 / nForce 720a 084F - GeForce 9400 0860 B GeForce 9400 0861 B GeForce 9400M G 0862 B GeForce 9400M 0863 B GeForce 9300 0864 B ION 0865 B GeForce 9400M G 0866 B GeForce 9400M 0866 106B 00B1 B GeForce 9400 0867 B nForce 760i SLI 0868 B GeForce 9400 0869 B GeForce 9400 086A B GeForce 9300 / nForce 730i 086C B GeForce 9200 086D B GeForce 9100M G 086E B GeForce 8200M G 086F B GeForce 9400M 0870 B GeForce 9200 0871 B GeForce G102M 0872 B GeForce G205M 0872 1043 1C42 B GeForce G102M 0873 B GeForce G205M 0873 1043 1C52 B ION 0874 B ION 0876 B GeForce 9400 087A B ION 087D B ION LE 087E B ION LE 087F B GeForce 320M 08A0 C GeForce 320M 08A2 C GeForce 320M 08A3 C GeForce 320M 08A4 C GeForce 320M 08A5 C GeForce GT 220 0A20 C GeForce 315 0A22 - GeForce 210 0A23 C GeForce 405 0A26 C GeForce 405 0A27 C GeForce GT 230M 0A28 C GeForce GT 330M 0A29 C GeForce GT 230M 0A2A C GeForce GT 330M 0A2B C NVS 5100M 0A2C C GeForce GT 320M 0A2D A GeForce GT 415 0A32 C GeForce GT 240M 0A34 C GeForce GT 325M 0A35 C Quadro 400 0A38 C Quadro FX 880M 0A3C C GeForce G210 0A60 C GeForce 205 0A62 C GeForce 310 0A63 C Second Generation ION 0A64 C GeForce 210 0A65 C GeForce 310 0A66 C GeForce 315 0A67 - GeForce G105M 0A68 B GeForce G105M 0A69 B NVS 2100M 0A6A C NVS 3100M 0A6C C GeForce 305M 0A6E C Second Generation ION 0A6E 17AA 3607 C Second Generation ION 0A6F C GeForce 310M 0A70 C Second Generation ION 0A70 17AA 3605 C Second Generation ION 0A70 17AA 3617 C GeForce 305M 0A71 C GeForce 310M 0A72 C GeForce 305M 0A73 C Second Generation ION 0A73 17AA 3607 C Second Generation ION 0A73 17AA 3610 C GeForce G210M 0A74 C GeForce G210 0A74 17AA 903A C GeForce 310M 0A75 C Second Generation ION 0A75 17AA 3605 C Second Generation ION 0A76 C Quadro FX 380 LP 0A78 C GeForce 315M 0A7A C GeForce 405 0A7A 1462 AA51 C GeForce 405 0A7A 1462 AA58 C GeForce 405 0A7A 1462 AC71 C GeForce 405 0A7A 1462 AC82 C GeForce 405 0A7A 1642 3980 C GeForce 405M 0A7A 17AA 3950 C GeForce 405M 0A7A 17AA 397D C GeForce 405 0A7A 1B0A 90B4 C GeForce 405 0A7A 1BFD 0003 C GeForce 405 0A7A 1BFD 8006 C Quadro FX 380M 0A7C C GeForce GT 330 0CA0 A GeForce GT 320 0CA2 C GeForce GT 240 0CA3 C GeForce GT 340 0CA4 C GeForce GT 220 0CA5 C GeForce GT 330 0CA7 A GeForce GTS 260M 0CA8 C GeForce GTS 250M 0CA9 C GeForce GT 220 0CAC C GeForce GT 335M 0CAF C GeForce GTS 350M 0CB0 C GeForce GTS 360M 0CB1 C Quadro FX 1800M 0CBC C GeForce 9300 GS 10C0 B GeForce 8400GS 10C3 A GeForce 405 10C5 C NVS 300 10D8 C The 304.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID ----------------------------------------------- -------------- GeForce 6800 Ultra 0040 GeForce 6800 0041 GeForce 6800 LE 0042 GeForce 6800 XE 0043 GeForce 6800 XT 0044 GeForce 6800 GT 0045 GeForce 6800 GT 0046 GeForce 6800 GS 0047 GeForce 6800 XT 0048 Quadro FX 4000 004E GeForce 7800 GTX 0090 GeForce 7800 GTX 0091 GeForce 7800 GT 0092 GeForce 7800 GS 0093 GeForce 7800 SLI 0095 GeForce Go 7800 0098 GeForce Go 7800 GTX 0099 Quadro FX 4500 009D GeForce 6800 GS 00C0 GeForce 6800 00C1 GeForce 6800 LE 00C2 GeForce 6800 XT 00C3 GeForce Go 6800 00C8 GeForce Go 6800 Ultra 00C9 Quadro FX Go1400 00CC Quadro FX 3450/4000 SDI 00CD Quadro FX 1400 00CE GeForce 6600 GT 00F1 GeForce 6600 00F2 GeForce 6200 00F3 GeForce 6600 LE 00F4 GeForce 7800 GS 00F5 GeForce 6800 GS 00F6 Quadro FX 3400/Quadro FX 4000 00F8 GeForce 6800 Ultra 00F9 GeForce 6600 GT 0140 GeForce 6600 0141 GeForce 6600 LE 0142 GeForce 6600 VE 0143 GeForce Go 6600 0144 GeForce 6610 XL 0145 GeForce Go 6600 TE/6200 TE 0146 GeForce 6700 XL 0147 GeForce Go 6600 0148 GeForce Go 6600 GT 0149 Quadro NVS 440 014A Quadro FX 540M 014C Quadro FX 550 014D Quadro FX 540 014E GeForce 6200 014F GeForce 6500 0160 GeForce 6200 TurboCache(TM) 0161 GeForce 6200SE TurboCache(TM) 0162 GeForce 6200 LE 0163 GeForce Go 6200 0164 Quadro NVS 285 0165 GeForce Go 6400 0166 GeForce Go 6200 0167 GeForce Go 6400 0168 GeForce 6250 0169 GeForce 7100 GS 016A GeForce 7350 LE 01D0 GeForce 7300 LE 01D1 GeForce 7550 LE 01D2 GeForce 7300 SE/7200 GS 01D3 GeForce Go 7200 01D6 GeForce Go 7300 01D7 GeForce Go 7400 01D8 Quadro NVS 110M 01DA Quadro NVS 120M 01DB Quadro FX 350M 01DC GeForce 7500 LE 01DD Quadro FX 350 01DE GeForce 7300 GS 01DF GeForce 6800 0211 GeForce 6800 LE 0212 GeForce 6800 GT 0215 GeForce 6800 XT 0218 GeForce 6200 0221 GeForce 6200 A-LE 0222 GeForce 6150 0240 GeForce 6150 LE 0241 GeForce 6100 0242 GeForce Go 6150 0244 Quadro NVS 210S / GeForce 6150LE 0245 GeForce Go 6100 0247 GeForce 7900 GTX 0290 GeForce 7900 GT/GTO 0291 GeForce 7900 GS 0292 GeForce 7950 GX2 0293 GeForce 7950 GX2 0294 GeForce 7950 GT 0295 GeForce Go 7950 GTX 0297 GeForce Go 7900 GS 0298 Quadro NVS 510M 0299 Quadro FX 2500M 029A Quadro FX 1500M 029B Quadro FX 5500 029C Quadro FX 3500 029D Quadro FX 1500 029E Quadro FX 4500 X2 029F GeForce 7600 GT 02E0 GeForce 7600 GS 02E1 GeForce 7300 GT 02E2 GeForce 7900 GS 02E3 GeForce 7950 GT 02E4 GeForce 7650 GS 038B GeForce 7650 GS 0390 GeForce 7600 GT 0391 GeForce 7600 GS 0392 GeForce 7300 GT 0393 GeForce 7600 LE 0394 GeForce 7300 GT 0395 GeForce Go 7700 0397 GeForce Go 7600 0398 GeForce Go 7600 GT 0399 Quadro FX 560M 039C Quadro FX 560 039E GeForce 6150SE nForce 430 03D0 GeForce 6100 nForce 405 03D1 GeForce 6100 nForce 400 03D2 GeForce 6100 nForce 420 03D5 GeForce 7025 / nForce 630a 03D6 GeForce 7150M / nForce 630M 0531 GeForce 7000M / nForce 610M 0533 GeForce 7050 PV / nForce 630a 053A GeForce 7050 PV / nForce 630a 053B GeForce 7025 / nForce 630a 053E GeForce 7150 / nForce 630i 07E0 GeForce 7100 / nForce 630i 07E1 GeForce 7050 / nForce 630i 07E2 GeForce 7050 / nForce 610i 07E3 GeForce 7050 / nForce 620i 07E5 The 173.14.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID ----------------------------------------------- -------------- GeForce PCX 5750 00FA GeForce PCX 5900 00FB Quadro FX 330/GeForce PCX 5300 00FC Quadro FX 330/Quadro NVS 280 PCI-E 00FD Quadro FX 1300 00FE GeForce FX 5800 Ultra 0301 GeForce FX 5800 0302 Quadro FX 2000 0308 Quadro FX 1000 0309 GeForce FX 5600 Ultra 0311 GeForce FX 5600 0312 GeForce FX 5600XT 0314 GeForce FX Go5600 031A GeForce FX Go5650 031B Quadro FX Go700 031C GeForce FX 5200 0320 GeForce FX 5200 Ultra 0321 GeForce FX 5200 0322 GeForce FX 5200LE 0323 GeForce FX Go5200 0324 GeForce FX Go5250 0325 GeForce FX 5500 0326 GeForce FX 5100 0327 GeForce FX Go5200 32M/64M 0328 Quadro NVS 55/280 PCI 032A Quadro FX 500/FX 600 032B GeForce FX Go53xx 032C GeForce FX Go5100 032D GeForce FX 5900 Ultra 0330 GeForce FX 5900 0331 GeForce FX 5900XT 0332 GeForce FX 5950 Ultra 0333 GeForce FX 5900ZT 0334 Quadro FX 3000 0338 Quadro FX 700 033F GeForce FX 5700 Ultra 0341 GeForce FX 5700 0342 GeForce FX 5700LE 0343 GeForce FX 5700VE 0344 GeForce FX Go5700 0347 GeForce FX Go5700 0348 Quadro FX Go1000 034C Quadro FX 1100 034E The 96.43.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID ----------------------------------------------- -------------- GeForce2 MX/MX 400 0110 GeForce2 MX 100/200 0111 GeForce2 Go 0112 Quadro2 MXR/EX/Go 0113 GeForce4 MX 460 0170 GeForce4 MX 440 0171 GeForce4 MX 420 0172 GeForce4 MX 440-SE 0173 GeForce4 440 Go 0174 GeForce4 420 Go 0175 GeForce4 420 Go 32M 0176 GeForce4 460 Go 0177 Quadro4 550 XGL 0178 GeForce4 440 Go 64M 0179 Quadro NVS 400 017A Quadro4 500 GoGL 017C GeForce4 410 Go 16M 017D GeForce4 MX 440 with AGP8X 0181 GeForce4 MX 440SE with AGP8X 0182 GeForce4 MX 420 with AGP8X 0183 GeForce4 MX 4000 0185 Quadro4 580 XGL 0188 Quadro NVS 280 SD 018A Quadro4 380 XGL 018B Quadro NVS 50 PCI 018C GeForce2 Integrated GPU 01A0 GeForce4 MX Integrated GPU 01F0 GeForce3 0200 GeForce3 Ti 200 0201 GeForce3 Ti 500 0202 Quadro DCC 0203 GeForce4 Ti 4600 0250 GeForce4 Ti 4400 0251 GeForce4 Ti 4200 0253 Quadro4 900 XGL 0258 Quadro4 750 XGL 0259 Quadro4 700 XGL 025B GeForce4 Ti 4800 0280 GeForce4 Ti 4200 with AGP8X 0281 GeForce4 Ti 4800 SE 0282 GeForce4 4200 Go 0286 Quadro4 980 XGL 0288 Quadro4 780 XGL 0289 Quadro4 700 GoGL 028C The 71.86.xx driver supports the following set of GPUs: NVIDIA GPU product Device PCI ID ----------------------------------------------- -------------- RIVA TNT 0020 RIVA TNT2/TNT2 Pro 0028 RIVA TNT2 Ultra 0029 Vanta/Vanta LT 002C RIVA TNT2 Model 64/Model 64 Pro 002D Aladdin TNT2 00A0 GeForce 256 0100 GeForce DDR 0101 Quadro 0103 GeForce2 GTS/GeForce2 Pro 0150 GeForce2 Ti 0151 GeForce2 Ultra 0152 Quadro2 Pro 0153 * If three IDs are listed, the first is the PCI Device ID, the second is the PCI Subsystem Vendor ID, and the third is the PCI Subsystem Device ID. ______________________________________________________________________________ Appendix B. X Config Options ______________________________________________________________________________ The following driver options are supported by the NVIDIA X driver. They may be specified either in the Screen or Device sections of the X config file. X Config Options Option "Accel" "boolean" Controls whether the X driver uses the GPU for graphics processing. Disabling acceleration is useful when another component, such as CUDA, requires exclusive use of the GPU's processing cores. Performance of the X server will be reduced when acceleration is disabled, and some features may not be available. OpenGL and VDPAU are not supported when Accel is disabled. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Default: acceleration is enabled. Option "RenderAccel" "boolean" Enable or disable hardware acceleration of the RENDER extension. Default: hardware acceleration of the RENDER extension is enabled. Option "NoRenderExtension" "boolean" Disable the RENDER extension. Other than recompiling it, the X server does not seem to have another way of disabling this. Fortunately, we can control this from the driver so we export this option. This is useful in depth 8 where RENDER would normally steal most of the default colormap. Default: RENDER is offered when possible. Option "UBB" "boolean" Enable or disable the Unified Back Buffer on NVIDIA RTX/Quadro-based GPUs (Quadro NVS excluded); see Chapter 19 for a description of UBB. This option has no effect on non-NVIDIA RTX/Quadro GPU products. Default: UBB is on for NVIDIA RTX/Quadro GPUs. Option "NoFlip" "boolean" Disable OpenGL flipping; see Chapter 19 for a description. Default: OpenGL will swap by flipping when possible. Option "GLShaderDiskCache" "boolean" This option controls whether the OpenGL driver will utilize a disk cache to save and reuse compiled shaders. See the description of the __GL_SHADER_DISK_CACHE and __GL_SHADER_DISK_CACHE_PATH environment variables in Chapter 11 for more details. Option "Overlay" "boolean" Enables RGB workstation overlay visuals. This is only supported on NVIDIA RTX/Quadro GPUs (Quadro NVS GPUs excluded) in depth 24. This option causes the server to advertise the SERVER_OVERLAY_VISUALS root window property and GLX will report single- and double-buffered, Z-buffered 16-bit overlay visuals. The transparency key is pixel 0x0000 (hex). There is no gamma correction support in the overlay plane. RGB workstation overlays are not supported when the Composite extension is enabled. UBB must be enabled when overlays are enabled (this is the default behavior). Option "CIOverlay" "boolean" Enables Color Index workstation overlay visuals with identical restrictions to Option "Overlay" above. This option causes the server to advertise the SERVER_OVERLAY_VISUALS root window property. Some of the visuals advertised that way may be listed in the main plane (layer 0) for compatibility purposes. They however belong to the overlay (layer 1). The server will offer visuals both with and without a transparency key. These are depth 8 PseudoColor visuals. Color Index workstation overlays are not supported when the Composite extension is enabled. Default: off. UBB must be enabled when overlays are enabled (this is the default behavior). Option "TransparentIndex" "integer" When color index overlays are enabled, use this option to choose which pixel is used for the transparent pixel in visuals featuring transparent pixels. This value is clamped between 0 and 255 (Note: some applications such as Alias's Maya require this to be zero in order to work correctly). Default: 0. Option "OverlayDefaultVisual" "boolean" When overlays are used, this option sets the default visual to an overlay visual thereby putting the root window in the overlay. This option is not recommended for RGB overlays. Default: off. Option "EmulatedOverlaysTimerMs" "integer" Enables the use of a timer within the X server to perform the updates to the emulated overlay or CI overlay. This option can be used to improve the performance of the emulated or CI overlays by reducing the frequency of the updates. The value specified indicates the desired number of milliseconds between overlay updates. To disable the use of the timer either leave the option unset or set it to 0. Default: off. Option "EmulatedOverlaysThreshold" "boolean" Enables the use of a threshold within the X server to perform the updates to the emulated overlay or CI overlay. The emulated or CI overlay updates can be deferred but this threshold will limit the number of deferred OpenGL updates allowed before the overlay is updated. This option can be used to trade off performance and animation quality. Default: on. Option "EmulatedOverlaysThresholdValue" "integer" Controls the threshold used in updating the emulated or CI overlays. This is used in conjunction with the EmulatedOverlaysThreshold option to trade off performance and animation quality. Higher values for this option favor performance over quality. Setting low values of this option will not cause the overlay to be updated more often than the frequence specified by the EmulatedOverlaysTimerMs option. Default: 5. Option "SWCursor" "boolean" Enable or disable software rendering of the X cursor. Default: off. Option "HWCursor" "boolean" Enable or disable hardware rendering of the X cursor. Default: on. Option "ConnectedMonitor" "string" Allows you to override what the NVIDIA kernel module detects is connected to your graphics card. This may be useful, for example, if you use a KVM (keyboard, video, mouse) switch and you are switched away when X is started. It may also be useful in situations where you wish to simulate the presence of a monitor when none are connected. When simulating the presence of a monitor, it may be useful to combine this option with the "CustomEDID" option to provide the simulated display device with an EDID. Valid values for this option are "CRT" (cathode ray tube) or "DFP" (digital flat panel); if using multiple display devices, this option may be a comma-separated list of display devices; e.g.: "CRT, CRT" or "CRT, DFP". The display device name used in this option may also include a numeric suffix (e.g., "DFP-1") to refer to a specific display connector; check your X log to determine which display device identifiers are valid for your particular GPU. It is generally recommended to not use this option, but instead use the "UseDisplayDevice" option. NOTE: anything attached to a 15 pin VGA connector is regarded by the driver as a CRT. "DFP" should only be used to refer to digital flat panels connected via DVI, HDMI, or DisplayPort. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Default: string is NULL (the NVIDIA driver will detect the connected display devices). Option "UseDisplayDevice" "string" The "UseDisplayDevice" X configuration option is a list of one or more display devices, which limits the display devices the NVIDIA X driver will consider for an X screen. The display device names used in the option may be either specific (with a numeric suffix; e.g., "DFP-1") or general (without a numeric suffix; e.g., "DFP"). When assigning display devices to X screens, the NVIDIA X driver walks through the list of all (not already assigned) display devices detected as connected. When the "UseDisplayDevice" X configuration option is specified, the X driver will only consider connected display devices which are also included in the "UseDisplayDevice" list. This can be thought of as a "mask" against the connected (and not already assigned) display devices. Note the subtle difference between this option and the "ConnectedMonitor" option: the "ConnectedMonitor" option overrides which display devices are actually detected, while the "UseDisplayDevice" option controls which of the detected display devices will be used on this X screen. Of the list of display devices considered for this X screen (either all connected display devices, or a subset limited by the "UseDisplayDevice" option), the NVIDIA X driver first looks at CRTs, then at DFPs. For example, if both a CRT and a DFP are connected, by default the X driver would assign the CRT to this X screen. However, by specifying: Option "UseDisplayDevice" "DFP" the X screen would use the DFP instead. Or, if CRT-0, DFP-0, and DFP-1 are connected, the X driver would assign CRT-0 and DFP-0 to the X screen. However, by specifying: Option "UseDisplayDevice" "CRT-0, DFP-1" the X screen would use CRT-0 and DFP-1 instead. Additionally, the special value "none" can be specified for the "UseDisplayDevice" option. When this value is given, any programming of the display hardware is disabled. The NVIDIA driver will not perform any mode validation or mode setting for this X screen. This is intended for use in conjunction with CUDA or in remote graphics solutions such as VNC or Hewlett Packard's Remote Graphics Software (RGS); however, note that this configuration option may not interact well with some desktop environments or window managers (See "Q. My remote desktop software doesn't work when no displays are attached." in Chapter 8 for more information). "UseDisplayDevice" defaults to "none" on GPUs that have no display capabilities, such as some Tesla GPUs and some mobile GPUs used in Optimus notebook configurations. Note the following restrictions for setting the "UseDisplayDevice" to "none": o OpenGL SyncToVBlank will have no effect. o None of Stereo, Overlay, CIOverlay, or SLI are allowed when "UseDisplayDevice" is set to "none". Option "UseEdidFreqs" "boolean" This option controls whether the NVIDIA X driver will use the HorizSync and VertRefresh ranges given in a display device's EDID, if any. When UseEdidFreqs is set to True, EDID-provided range information will override the HorizSync and VertRefresh ranges specified in the Monitor section. If a display device does not provide an EDID, or the EDID does not specify an hsync or vrefresh range, then the X server will default to the HorizSync and VertRefresh ranges specified in the Monitor section of your X config file. These frequency ranges are used when validating modes for your display device. Default: True (EDID frequencies will be used) Option "UseEDID" "boolean" By default, the NVIDIA X driver makes use of a display device's EDID, when available, during construction of its mode pool. The EDID is used as a source for possible modes, for valid frequency ranges, and for collecting data on the physical dimensions of the display device for computing the DPI (see Appendix E). However, if you wish to disable the driver's use of the EDID, you can set this option to False: Option "UseEDID" "FALSE" Note that, rather than globally disable all uses of the EDID, you can individually disable each particular use of the EDID; e.g., Option "UseEDIDFreqs" "FALSE" Option "UseEDIDDpi" "FALSE" Option "ModeValidation" "NoEdidModes" When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Default: True (use EDID). Option "MetaModeOrientation" "string" Controls the default relationship between display devices when using multiple display devices on a single X screen. Takes one of the following values: "RightOf" "LeftOf" "Above" "Below" "SamePositionAs". For backwards compatibility, "TwinViewOrientation" is a synonym for "MetaModeOrientation", and "Clone" is a synonym for "SamePositionAs". See Chapter 12 for details. Default: string is NULL. Option "MetaModes" "string" This option describes the combination of modes to use on each monitor. See Chapter 12 and Chapter 31 for details. Default: string is NULL. Option "nvidiaXineramaInfo" "boolean" The NVIDIA X driver normally provides a Xinerama extension that X clients (such as window managers) can use to discover the current layout of display devices within an X screen. Some window mangers get confused by this information, so this option is provided to disable this behavior. Default: true (NVIDIA Xinerama information is provided). On X servers with RandR 1.2 support, the X server's RandR implementation may provide its own Xinerama implementation if NVIDIA Xinerama information is not provided. So, on X servers with RandR 1.2, disabling "nvidiaXineramaInfo" causes the NVIDIA X driver to still register its Xinerama implementation but report a single screen-sized region. On X servers without RandR 1.2 support, disabling "nvidiaXineramaInfo" causes the NVIDIA X driver to not register its Xinerama implementation. For backwards compatibility, "NoTwinViewXineramaInfo" is a synonym for disabling "nvidiaXineramaInfo". Option "nvidiaXineramaInfoOrder" "string" When the NVIDIA X driver provides nvidiaXineramaInfo (see the nvidiaXineramaInfo X config option), it by default reports the currently enabled display devices in the order "CRT, DFP". The nvidiaXineramaInfoOrder X config option can be used to override this order. The option string is a comma-separated list of display device names. The display device names can either be general (e.g, "CRT", which identifies all CRTs), or specific (e.g., "CRT-1", which identifies a particular CRT). Not all display devices need to be identified in the option string; display devices that are not listed will be implicitly appended to the end of the list, in their default order. Note that nvidiaXineramaInfoOrder tracks all display devices that could possibly be connected to the GPU, not just the ones that are currently enabled. When reporting the Xinerama information, the NVIDIA X driver walks through the display devices in the order specified, only reporting enabled display devices. Some high-resolution "tiled" monitors are represented internally as multiple display devices. These will be combined by default into a single Xinerama screen. The position of the device in nvidiaXineramaInfoOrder corresponding to the top-left tile will determine when the screen will be reported. Examples: "DFP" "DFP-1, DFP-0, CRT" In the first example, any enabled DFPs would be reported first (any enabled CRTs would be reported afterwards). In the second example, if DFP-1 were enabled, it would be reported first, then DFP-0, and then any enabled CRTs; finally, any other enabled DFPs would be reported. For backwards compatibility, "TwinViewXineramaInfoOrder" is a synonym for "nvidiaXineramaInfoOrder". Default: "CRT, DFP" Option "nvidiaXineramaInfoOverride" "string" This option overrides the values reported by the NVIDIA X driver's nvidiaXineramaInfo implementation. This disregards the actual display devices used by the X screen and any order specified in nvidiaXineramaInfoOrder. The option string is interpreted as a comma-separated list of regions, specified as '[width]x[height]+[x-offset]+[y-offset]'. The regions' sizes and offsets are not validated against the X screen size, but are directly reported to any Xinerama client. Examples: "1600x1200+0+0, 1600x1200+1600+0" "1024x768+0+0, 1024x768+1024+0, 1024x768+0+768, 1024x768+1024+768" For backwards compatibility, "TwinViewXineramaInfoOverride" is a synonym for "nvidiaXineramaInfoOverride". Option "Stereo" "integer" Enable offering of quad-buffered stereo visuals on NVIDIA RTX/Quadro. Integer indicates the type of stereo equipment being used: Value Equipment -------------- --------------------------------------------------- 3 Onboard stereo support. This is usually only found on professional cards. The glasses connect via a DIN connector on the back of the graphics card. 4 One-eye-per-display passive stereo. This mode allows each display to be configured to statically display either left or right eye content. This can be especially useful with multi-display configurations (TwinView or SLI Mosaic). For example, this is commonly used in conjunction with special projectors to produce 2 polarized images which are then viewed with polarized glasses. To use this stereo mode, it is recommended that you configure TwinView (or pairs of displays in SLI Mosaic) in clone mode with the same resolution, panning offset, and panning domains on each display. See Chapter 12 for more information about configuring multiple displays. 5 Vertical interlaced stereo mode, for use with SeeReal Stereo Digital Flat Panels. 6 Color interleaved stereo mode, for use with Sharp3D Stereo Digital Flat Panels. 7 Horizontal interlaced stereo mode, for use with Arisawa, Hyundai, Zalman, Pavione, and Miracube Digital Flat Panels. 8 Checkerboard pattern stereo mode, for use with 3D DLP Display Devices. 9 Inverse checkerboard pattern stereo mode, for use with 3D DLP Display Devices. 10 NVIDIA 3D Vision mode for use with NVIDIA 3D Vision glasses. The NVIDIA 3D Vision infrared emitter must be connected to a USB port of your computer, and to the 3-pin DIN connector of an NVIDIA RTX/Quadro graphics board before starting the X server. Hot-plugging the USB infrared stereo emitter is not yet supported. Also, 3D Vision Stereo Linux support requires a Linux kernel built with USB device filesystem (usbfs) and USB 2.0 support. Not presently supported on FreeBSD or Solaris. 11 NVIDIA 3D VisionPro mode for use with NVIDIA 3D VisionPro glasses. The NVIDIA 3D VisionPro RF hub must be connected to a USB port of your computer, and to the 3-pin DIN connector of an NVIDIA RTX/Quadro graphics board before starting the X server. Hot-plugging the USB RF hub is not yet supported. Also, 3D VisionPro Stereo Linux support requires a Linux kernel built with USB device filesystem (usbfs) and USB 2.0 support. When RF hub is connected and X is started in NVIDIA 3D VisionPro stereo mode, a new page will be available in nvidia-settings for various configuration settings. Some of these settings can also be done via nvidia-settings command line interface. Refer to the corresponding Help section in nvidia-settings for further details. Not presently supported on FreeBSD or Solaris. 12 HDMI 3D mode for use with HDMI 3D compatible display devices with their own stereo emitters. 13 Tridelity SL stereo mode, for use with Tridelity SL display devices. 14 Generic active stereo with in-band DisplayPort stereo signaling, for use with DisplayPort display devices with their own stereo emitters. See the documentation for the "InbandStereoSignaling" X config option for more details. Default: 0 (Stereo is not enabled). Stereo options 3, 10, 11, 12, and 14 are known as "active" stereo. Other options are known as "passive" stereo. When active stereo is used with multiple display devices, it is recommended that modes within each MetaMode have identical timing values (modelines). See Chapter 18 for suggestions on making sure the modes within your MetaModes are identical. The following table summarizes the available stereo modes, their supported GPUs, and their intended display devices: Stereo mode (value) Graphics card Display supported supported [1] -------------------- -------------------- -------------------- Onboard DIN (3) NVIDIA RTX/Quadro Displays with high graphics cards refresh rate One-eye-per-display NVIDIA RTX/Quadro Any (4) graphics cards Vertical Interlaced NVIDIA RTX/Quadro SeeReal Stereo DFP (5) graphics cards Color Interleaved NVIDIA RTX/Quadro Sharp3D stereo DFP (6) graphics cards Horizontal NVIDIA RTX/Quadro Arisawa, Hyundai, Interlaced (7) graphics cards Zalman, Pavione, and Miracube Checkerboard NVIDIA RTX/Quadro 3D DLP display Pattern (8) graphics cards devices Inverse NVIDIA RTX/Quadro 3D DLP display Checkerboard (9) graphics cards devices NVIDIA 3D Vision NVIDIA RTX/Quadro Supported 3D Vision (10) graphics cards [2] ready displays [3] NVIDIA 3D VisionPro NVIDIA RTX/Quadro Supported 3D Vision (11) graphics cards [2] ready displays [3] HDMI 3D (12) NVIDIA RTX/Quadro Supported HDMI 3D graphics cards displays [4] Tridelity SL (13) NVIDIA RTX/Quadro Tridelity SL DFP graphics cards Generic active NVIDIA RTX/Quadro DisplayPort stereo (in-band DP) graphics cards displays with (14) in-band stereo support [1] NVIDIA RTX/Quadro graphics cards excluding Quadro NVS cards. [2] http://www.nvidia.com/object/quadro_pro_graphics_boards_linux.html [3] http://www.nvidia.com/object/3D_Vision_Requirements.html [4] Supported 3D TVs, Projectors, and Home Theater Receivers listed on http://www.nvidia.com/object/3dtv-play-system-requirements.html and Desktop LCD Monitors with 3D Vision HDMI support listed on http://www.nvidia.com/object/3D_Vision_Requirements.html UBB must be enabled when stereo is enabled (this is the default behavior). Active stereo can be enabled on digital display devices (connected via DVI, HDMI, or DisplayPort). However, some digital display devices might not behave as desired with active stereo: o Some digital display devices may not be able to toggle pixel colors quickly enough when flipping between eyes on every vblank. o Some digital display devices may have an optical polarization that interferes with stereo goggles. o Active stereo requires high refresh rates, because a vertical refresh is needed to display each eye. Some digital display devices have a low refresh rate, which will result in flickering when used for active stereo. o Some digital display devices might internally convert from other refresh rates to their native refresh rate (e.g., 60Hz), resulting in incompatible rates between the stereo glasses and stereo displayed on screen. These limitations do not apply to any display devices suitable for stereo options 10, 11, or 12. Interlaced modes are disabled when active stereo is enabled. Stereo option 12 (HDMI 3D) is also known as HDMI Frame Packed Stereo mode, where the left and right eye images are stacked into a single frame with a doubled pixel clock and refresh rate. This doubled refresh rate is used for Frame Lock and in refresh rate queries through NV-CONTROL clients, and the doubled pixel clock and refresh rate are used in mode validation. Interlaced modes are not supported with this stereo mode. The following nvidia-settings command line can be used to determine whether a display's current mode is an HDMI 3D mode with a doubled refresh rate: nvidia-settings --query=Hdmi3D Stereo applies to an entire X screen, so it will apply to all display devices on that X screen, whether or not they all support the selected Stereo mode. Option "ForceStereoFlipping" "boolean" Stereo flipping is the process by which left and right eyes are displayed on alternating vertical refreshes. Normally, stereo flipping is only performed when a stereo drawable is visible. This option forces stereo flipping even when no stereo drawables are visible. This is to be used in conjunction with the "Stereo" option. If "Stereo" is 0, the "ForceStereoFlipping" option has no effect. If otherwise, the "ForceStereoFlipping" option will force the behavior indicated by the "Stereo" option, even if no stereo drawables are visible. This option is useful in a multiple-screen environment in which a stereo application is run on a different screen than the stereo master. Possible values: Value Behavior -------------- --------------------------------------------------- 0 Stereo flipping is not forced. The default behavior as indicated by the "Stereo" option is used. 1 Stereo flipping is forced. Stereo is running even if no stereo drawables are visible. The stereo mode depends on the value of the "Stereo" option. Default: 0 (Stereo flipping is not forced). Option "XineramaStereoFlipping" "boolean" By default, when using Stereo with Xinerama, all physical X screens having a visible stereo drawable will stereo flip. Use this option to allow only one physical X screen to stereo flip at a time. This is to be used in conjunction with the "Stereo" and "Xinerama" options. If "Stereo" is 0 or "Xinerama" is 0, the "XineramaStereoFlipping" option has no effect. If you wish to have all X screens stereo flip all the time, see the "ForceStereoFlipping" option. Possible values: Value Behavior -------------- --------------------------------------------------- 0 Stereo flipping is enabled on one X screen at a time. Stereo is enabled on the first X screen having the stereo drawable. 1 Stereo flipping in enabled on all X screens. Default: 1 (Stereo flipping is enabled on all X screens). Option "MultisampleCompatibility" "boolean" Enable or disable the use of separate front and back multisample buffers. Enabling this will consume more memory but is necessary for correct output when rendering to both the front and back buffers of a multisample or FSAA drawable. This option is necessary for correct operation of SoftImage XSI. Default: false (a single multisample buffer is shared between the front and back buffers). Option "NoPowerConnectorCheck" "boolean" The NVIDIA X driver will fail initialization on a GPU if it detects that the GPU that requires an external power connector does not have an external power connector plugged in. This option can be used to bypass this test. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Default: false (the power connector test is performed). Option "ThermalConfigurationCheck" "boolean" The NVIDIA X driver will fail initialization on a GPU if it detects that the GPU has a bad thermal configuration. This may indicate a problem with how your graphics board was built, or simply a driver bug. It is recommended that you contact your graphics board vendor if you encounter this problem. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. This option can be set to False to bypass this test. Default: true (the thermal configuration test is performed). Option "DamageEvents" "boolean" Use OS-level events to efficiently notify X when a client has performed direct rendering to a window that needs to be composited. This will significantly improve performance and interactivity when using GLX applications with a composite manager running. It will also affect applications using GLX when rotation is enabled. Enabled by default. Option "ExactModeTimingsDVI" "boolean" Forces the initialization of the X server with the exact timings specified in the ModeLine. Default: false (for DVI devices, the X server initializes with the closest mode in the EDID list). The "AllowNonEdidModes" token in the "ModeValidation" X configuration option has the same effect as "ExactModeTimingsDVI", but "AllowNonEdidModes" has per-display device granularity. Option "Coolbits" "integer" Enables various unsupported features, such as support for GPU clock manipulation in the NV-CONTROL X extension. This option accepts a bit mask of features to enable. WARNING: this may cause system damage and void warranties. This utility can run your computer system out of the manufacturer's design specifications, including, but not limited to: higher system voltages, above normal temperatures, excessive frequencies, and changes to BIOS that may corrupt the BIOS. Your computer's operating system may hang and result in data loss or corrupted images. Depending on the manufacturer of your computer system, the computer system, hardware and software warranties may be voided, and you may not receive any further manufacturer support. NVIDIA does not provide customer service support for the Coolbits option. It is for these reasons that absolutely no warranty or guarantee is either express or implied. Before enabling and using, you should determine the suitability of the utility for your intended use, and you shall assume all responsibility in connection therewith. When "8" (Bit 3) is set in the "Coolbits" option value, the PowerMizer page in the nvidia-settings control panel will display a table that allows setting per-clock domain and per-performance level offsets to apply to clock values. This is allowed on certain GeForce GPUs. Not all clock domains or performance levels may be modified. On GPUs based on the Pascal architecture the offset is applied to all performance levels. When "16" (Bit 4) is set in the "Coolbits" option value, the nvidia-settings command line interface allows setting GPU overvoltage. This is allowed on certain GeForce GPUs. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. The default for this option is 0 (unsupported features are disabled). Option "SLI" "string" This option controls the configuration of SLI rendering in supported configurations. Value Behavior -------------------------------- -------------------------------- 0, no, off, false, Single Use only a single GPU when rendering Mosaic Enable SLI and use SLI Mosaic Mode. Use this in conjunction with the MetaModes X configuration option to specify the combination of mode(s) used on each display. Option "TripleBuffer" "boolean" Enable or disable the use of triple buffering. If this option is enabled, OpenGL windows that sync to vblank and are double-buffered will be given a third buffer. This decreases the time an application stalls while waiting for vblank events, but increases latency slightly (delay between user input and displayed result). Option "DPI" "string" This option specifies the Dots Per Inch for the X screen; for example: Option "DPI" "75 x 85" will set the horizontal DPI to 75 and the vertical DPI to 85. By default, the X driver will compute the DPI of the X screen from the EDID of any connected display devices. See Appendix E for details. Default: string is NULL (disabled). Option "UseEdidDpi" "string" By default, the NVIDIA X driver computes the DPI of an X screen based on the physical size of the display device, as reported in the EDID, and the size in pixels of the first mode to be used on the display device. If multiple display devices are used by the X screen, then the NVIDIA X screen will choose which display device to use. This option can be used to specify which display device to use. The string argument can be a display device name, such as: Option "UseEdidDpi" "DFP-0" or the argument can be "FALSE" to disable use of EDID-based DPI calculations: Option "UseEdidDpi" "FALSE" See Appendix E for details. Default: string is NULL (the driver computes the DPI from the EDID of a display device and selects the display device). Option "ConstantDPI" "boolean" By default on X.Org 6.9 or newer, the NVIDIA X driver recomputes the size in millimeters of the X screen whenever the size in pixels of the X screen is changed using XRandR, such that the DPI remains constant. This behavior can be disabled (which means that the size in millimeters will not change when the size in pixels of the X screen changes) by setting the "ConstantDPI" option to "FALSE"; e.g., Option "ConstantDPI" "FALSE" ConstantDPI defaults to True. On X releases older than X.Org 6.9, the NVIDIA X driver cannot change the size in millimeters of the X screen. Therefore the DPI of the X screen will change when XRandR changes the size in pixels of the X screen. The driver will behave as if ConstantDPI was forced to FALSE. Option "CustomEDID" "string" This option forces the X driver to use the EDID specified in a file rather than the display's EDID. You may specify a semicolon separated list of display names and filename pairs. Valid display device names include "CRT-0", "CRT-1", "DFP-0", "DFP-1", "TV-0", "TV-1", or one of the generic names "CRT", "DFP", "TV", which apply the EDID to all devices of the specified type. Additionally, if SLI Mosaic is enabled, this name can be prefixed by a GPU name (e.g., "GPU-0.CRT-0"). The file contains a raw EDID (e.g., a file generated by nvidia-settings). For example: Option "CustomEDID" "CRT-0:/tmp/edid1.bin; DFP-0:/tmp/edid2.bin" will assign the EDID from the file /tmp/edid1.bin to the display device CRT-0, and the EDID from the file /tmp/edid2.bin to the display device DFP-0. Note that a display device name must always be specified even if only one EDID is specified. Caution: Specifying an EDID that doesn't exactly match your display may damage your hardware, as it allows the driver to specify timings beyond the capabilities of your display. Use with care. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Option "IgnoreEDIDChecksum" "string" This option forces the X driver to accept an EDID even if the checksum is invalid. You may specify a comma separated list of display names. Valid display device names include "CRT-0", "CRT-1", "DFP-0", "DFP-1", "TV-0", "TV-1", or one of the generic names "CRT", "DFP", "TV", which ignore the EDID checksum on all devices of the specified type. Additionally, if SLI Mosaic is enabled, this name can be prefixed by a GPU name (e.g., "GPU-0.CRT-0"). For example: Option "IgnoreEDIDChecksum" "CRT, DFP-0" will cause the nvidia driver to ignore the EDID checksum for all CRT monitors and the displays DFP-0 and TV-0. Caution: An invalid EDID checksum may indicate a corrupt EDID. A corrupt EDID may have mode timings beyond the capabilities of your display, and using it could damage your hardware. Use with care. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Option "ModeValidation" "string" This option provides fine-grained control over each stage of the mode validation pipeline, disabling individual mode validation checks. This option should only very rarely be used. The option string is a semicolon-separated list of comma-separated lists of mode validation arguments. Each list of mode validation arguments can optionally be prepended with a display device name and GPU specifier. ": , ; : , , ; ..." Possible arguments: o "NoMaxPClkCheck": each mode has a pixel clock; this pixel clock is validated against the maximum pixel clock of the hardware (for a DFP, this is the maximum pixel clock of the TMDS encoder, for a CRT, this is the maximum pixel clock of the DAC). This argument disables the maximum pixel clock checking stage of the mode validation pipeline. o "NoEdidMaxPClkCheck": a display device's EDID can specify the maximum pixel clock that the display device supports; a mode's pixel clock is validated against this pixel clock maximum. This argument disables this stage of the mode validation pipeline. o "NoMaxSizeCheck": each NVIDIA GPU has a maximum resolution that it can drive; this argument disables this stage of the mode validation pipeline. o "NoHorizSyncCheck": a mode's horizontal sync is validated against the range of valid horizontal sync values; this argument disables this stage of the mode validation pipeline. o "NoVertRefreshCheck": a mode's vertical refresh rate is validated against the range of valid vertical refresh rate values; this argument disables this stage of the mode validation pipeline. o "NoVirtualSizeCheck": if the X configuration file requests a specific virtual screen size, a mode cannot be larger than that virtual size; this argument disables this stage of the mode validation pipeline. o "NoVesaModes": when constructing the mode pool for a display device, the X driver uses a built-in list of VESA modes as one of the mode sources; this argument disables use of these built-in VESA modes. o "NoEdidModes": when constructing the mode pool for a display device, the X driver uses any modes listed in the display device's EDID as one of the mode sources; this argument disables use of EDID-specified modes. o "NoXServerModes": when constructing the mode pool for a display device, the X driver uses the built-in modes provided by the core Xorg X server as one of the mode sources; this argument disables use of these modes. Note that this argument does not disable custom ModeLines specified in the X config file; see the "NoCustomModes" argument for that. o "NoCustomModes": when constructing the mode pool for a display device, the X driver uses custom ModeLines specified in the X config file (through the "Mode" or "ModeLine" entries in the Monitor Section) as one of the mode sources; this argument disables use of these modes. o "NoPredefinedModes": when constructing the mode pool for a display device, the X driver uses additional modes predefined by the NVIDIA X driver; this argument disables use of these modes. o "NoUserModes": additional modes can be added to the mode pool dynamically, using the NV-CONTROL X extension; this argument prohibits user-specified modes via the NV-CONTROL X extension. o "NoExtendedGpuCapabilitiesCheck": allow mode timings that may exceed the GPU's extended capability checks. o "ObeyEdidContradictions": an EDID may contradict itself by listing a mode as supported, but the mode may exceed an EDID-specified valid frequency range (HorizSync, VertRefresh, or maximum pixel clock). Normally, the NVIDIA X driver prints a warning in this scenario, but does not invalidate an EDID-specified mode just because it exceeds an EDID-specified valid frequency range. However, the "ObeyEdidContradictions" argument instructs the NVIDIA X driver to invalidate these modes. o "NoTotalSizeCheck": allow modes in which the individual visible or sync pulse timings exceed the total raster size. o "NoDualLinkDVICheck": for mode timings used on dual link DVI DFPs, the driver must perform additional checks to ensure that the correct pixels are sent on the correct link. For some of these checks, the driver will invalidate the mode timings; for other checks, the driver will implicitly modify the mode timings to meet the GPU's dual link DVI requirements. This token disables this dual link DVI checking. o "NoDisplayPortBandwidthCheck": for mode timings used on DisplayPort devices, the driver must verify that the DisplayPort link can be configured to carry enough bandwidth to support a given mode's pixel clock. For example, some DisplayPort-to-VGA adapters only support 2 DisplayPort lanes, limiting the resolutions they can display. This token disables this DisplayPort bandwidth check. o "AllowNon3DVisionModes": modes that are not optimized for NVIDIA 3D Vision are invalidated, by default, when 3D Vision (stereo mode 10) or 3D Vision Pro (stereo mode 11) is enabled. This token allows the use of non-3D Vision modes on a 3D Vision monitor. (Stereo behavior of non-3D Vision modes on 3D Vision monitors is undefined.) o "AllowNonHDMI3DModes": modes that are incompatible with HDMI 3D are invalidated, by default, when HDMI 3D (stereo mode 12) is enabled. This token allows the use of non-HDMI 3D modes when HDMI 3D is selected. HDMI 3D will be disabled when a non-HDMI 3D mode is in use. o "AllowNonEdidModes": if a mode is not listed in a display device's EDID mode list, then the NVIDIA X driver will discard the mode if the EDID 1.3 "GTF Supported" flag is unset, if the EDID 1.4 "Continuous Frequency" flag is unset, or if the display device is connected to the GPU by a digital protocol (e.g., DVI, DP, etc). This token disables these checks for non-EDID modes. o "NoEdidHDMI2Check": HDMI 2.0 adds support for 4K@60Hz modes with either full RGB 4:4:4 pixel encoding or YUV (also known as YCbCr) 4:2:0 pixel encoding. Using these modes with RGB 4:4:4 pixel encoding requires GPU support as well as display support indicated in the display device's EDID. This token allows the use of these modes at RGB 4:4:4 as long as the GPU supports them, even if the display device's EDID does not indicate support. Otherwise, these modes will be displayed in the YUV 4:2:0 color space. o "AllowDpInterlaced": When driving interlaced modes over DisplayPort protocol, NVIDIA GPUs do not provide all the spec-mandated metadata. Some DisplayPort monitors are tolerant of this missing metadata. But, in the interest of DisplayPort specification compliance, the NVIDIA driver prohibits interlaced modes over DisplayPort protocol by default. Use this mode validation token to allow interlaced modes over DisplayPort protocol anyway. o "NoInterlacedModes": Use this mode validation token to prohibit interlaced modes. o "MaxOneHardwareHead": The display hardware of the GPU has a pipeline that produces the images that are sent to each display device. One unit within this pipeline is the "head", which generates the mode timings for the display device. Some extremely high pixel clock mode timings, such as 8K @ 60Hz, exceed the capabilities of a single head. On Turing and later GPUs, the NVIDIA GPU driver can use multiple heads within the GPU to produce the mode timings for one of these display devices; the driver will do so implicitly when needed. In the "Display Device Information" pages within nvidia-settings, this case will be indicated by reporting "Number of Hardware Heads Used" as "2" rather than "1". However, NVIDIA GPUs have a total of four hardware heads. Using two of them to drive one high pixel clock display device will mean that there are only two available to drive other display devices in the desktop (thus, the GPU may only be able to drive a total of three display devices if one of them requires the use of two hardware heads, or the GPU may only be able to drive a total of two display devices if each of them requires the use of two hardware heads). Set the "MaxOneHardwareHead" ModeValidation token to force the driver to use at most one hardware head to drive the display device, falling back to a lower refresh rate or resolution, if necessary. Note that the ability to use two hardware heads to drive one display device requires hardware support only available in Turing and later GPUs. Also, the use of two hardware heads with one display device is not currently compatible with Quadro Sync. Examples: Option "ModeValidation" "NoMaxPClkCheck" disable the maximum pixel clock check when validating modes on all display devices. Option "ModeValidation" "CRT-0: NoEdidModes, NoMaxPClkCheck; GPU-0.DFP-0: NoVesaModes" do not use EDID modes and do not perform the maximum pixel clock check on CRT-0, and do not use VESA modes on DFP-0 of GPU-0. Option "ColorSpace" "string" This option sets the preferred color space for all or a subset of the connected flat panels. The option string is a semicolon-separated list of device specific options. Each option can optionally be prepended with a display device name and a GPU specifier. ": ; : ; ..." Possible arguments: o "RGB": sets color space to RGB. RGB color space supports two valid color ranges; full and limited. By default, full color range is set when the color space is RGB. o "YCbCr444": sets color space to YCbCr 4:4:4. YCbCr supports only limited color range. It is not possible to set this color space if the GPU or display is not capable of limited range. If the ColorSpace option is not specified, or is incorrectly specified, then the color space is set to RGB by default. If driving the current mode in the RGB 4:4:4 color space would require a pixel clock that exceeds the display's or GPU's capabilities, and the display and GPU are capable of driving that mode in the YCbCr 4:2:0 color space, then the color space will be overridden to YCbCr 4:2:0. The current actual color space in use on the display can be queried with the following nvidia-settings command line: nvidia-settings --query=CurrentColorSpace Examples: Option "ColorSpace" "YCbCr444" set the color space to YCbCr 4:4:4 on all flat panels. Option "ColorSpace" "GPU-0.DFP-0: YCbCr444" set the color space to YCbCr 4:4:4 on DFP-0 of GPU-0. Option "ColorRange" "string" This option sets the preferred color range for all or a subset of the connected flat panels. The option string is a semicolon-separated list of device specific options. Each option can optionally be prepended with a display device name and a GPU specifier. ": ; : ; ..." Either full or limited color range may be selected as the preferred color range. The actual color range depends on the current color space, and will be overridden to limited color range if the current color space requires it. The current actual color range in use on the display can be queried with the following nvidia-settings command line: nvidia-settings --query=CurrentColorRange Possible arguments: o "Full": sets color range to full range. By default, full color range is set when the color space is RGB. o "Limited": sets color range to limited range. YCbCr444 supports only limited color range. Consequently, limited range is selected by the driver when color space is set to YCbCr444, and can not be changed. If the ColorRange option is not specified, or is incorrectly specified, then an appropriate default value is selected based on the selected color space. Examples: Option "ColorRange" "Limited" set the color range to limited on all flat panels. Option "ColorRange" "GPU-0.DFP-0: Limited" set the color range to limited on DFP-0 of GPU-0. Option "ModeDebug" "boolean" This option causes the X driver to print verbose details about mode validation to the X log file. Note that this option is applied globally: setting this option to TRUE will enable verbose mode validation logging for all NVIDIA X screens in the X server. Option "FlatPanelProperties" "string" This option requests particular properties for all or a subset of the connected flat panels. The option string is a semicolon-separated list of comma-separated property=value pairs. Each list of property=value pairs can optionally be prepended with a flat panel name and GPU specifier. ": , ; : ; ..." Recognized properties: o "Dithering": controls the flat panel dithering configuration; possible values are: 'Auto' (the driver will decide when to dither), 'Enabled' (the driver will always dither, if possible), and 'Disabled' (the driver will never dither). o "DitheringMode": controls the flat panel dithering mode; possible values are: 'Auto' (the driver will choose possible default mode), 'Dynamic-2x2' (a 2x2 dithering pattern is updated for every frame), 'Static-2x2' (a 2x2 dithering pattern remains constant throughout the frames), and 'Temporal' (a pseudo-random dithering algorithm is used). Examples: Option "FlatPanelProperties" "DitheringMode = Static-2x2" set the flat panel dithering mode to Static-2x2 on all flat panels. Option "FlatPanelProperties" "GPU-0.DFP-0: Dithering = Disabled; DFP-1: Dithering = Enabled, DitheringMode = Static-2x2" set dithering to disabled on DFP-0 on GPU-0, set DFP-1's dithering to enabled and dithering mode to static 2x2. Option "ProbeAllGpus" "boolean" When the NVIDIA X driver initializes, it probes all GPUs in the system, even if no X screens are configured on them. This is done so that the X driver can report information about all the system's GPUs through the NV-CONTROL X extension. This option can be set to FALSE to disable this behavior, such that only GPUs with X screens configured on them will be probed. Note that disabling this option may affect configurability through nvidia-settings, since the X driver will not know about GPUs that aren't currently being used or the display devices attached to them. Default: all GPUs in the system are probed. Option "IncludeImplicitMetaModes" "boolean" When the X server starts, a mode pool is created per display device, containing all the mode timings that the NVIDIA X driver determined to be valid for the display device. However, the only MetaModes that are made available to the X server are the ones explicitly requested in the X configuration file. It is convenient for fullscreen applications to be able to change between the modes in the mode pool, even if a given target mode was not explicitly requested in the X configuration file. To facilitate this, the NVIDIA X driver will implicitly add MetaModes for all modes in the primary display device's mode pool. This makes all the modes in the mode pool available to full screen applications that use the XF86VidMode extension or RandR 1.0/1.1 requests. Further, to make sure that fullscreen applications have a reasonable set of MetaModes available to them, the NVIDIA X driver will also add implicit MetaModes for common resolutions: 1920x1200, 1920x1080, 1600x1200, 1280x1024, 1280x720, 1024x768, 800x600, 640x480. For these common resolution implicit MetaModes, the common resolution will be the ViewPortIn, and nvidia-auto-select will be the mode. The ViewPortOut will be configured such that the ViewPortIn is aspect scaled within the mode. Each common resolution implicit MetaMode will be added if there is not already a MetaMode with that resolution, and if the resolution is not larger than the nvidia-auto-select mode of the display device. See Chapter 12 for details of the relationship between ViewPortIn, ViewPortOut, and the mode within a MetaMode. The IncludeImplicitMetaModes X configuration option can be used to disable the addition of implicit MetaModes. Or, it can be used to alter how implicit MetaModes are added. The option can have either a boolean value or a comma-separated list of token=value pairs, where the possible tokens are: o "DisplayDevice": specifies the display device for which the implicit MetaModes should be created. Any name that can be used to identify a display device can be used here; see Appendix C for details. o "Mode": specifies the name of the mode to use with the common resolution-based implicit MetaModes. The default is "nvidia-auto-select". Any mode in the display device's mode pool can be used here. o "Scaling": specifies how the ViewPortOut should be configured between the ViewPortIn and the mode for the common resolution-based implicit MetaModes. Possible values are "Scaled", "Aspect-Scaled", or "Centered". The default is "Aspect-Scaled". o "UseModePool": specifies whether modes from the display device's mode pool should be used to create implicit MetaModes. The default is "true". o "UseCommonResolutions": specifies whether the common resolution list should be used to create implicit MetaModes. The default is "true". o "Derive16x9Mode": specifies whether to create an implicit MetaMode with a resolution whose aspect ratio is 16:9, using the width of nvidia-auto-select. E.g., using a 2560x1600 monitor, this would create an implicit MetaMode of 2560x1440. The default is "true". o "ExtraResolutions": a comma-separated list of additional resolutions to use for creating implicit MetaModes. These will be created in the same way as the common resolution implicit MetaModes: the resolution will be used as the ViewPortIn, the nvidia-auto-select mode will be used as the mode, and the ViewPortOut will be computed to aspect scale the resolution within the mode. Note that the list of resolutions must be enclosed in parentheses, so that the commas are not interpreted as token=value pair separators. Some examples: Option "IncludeImplicitMetaModes" "off" Option "IncludeImplicitMetaModes" "on" (the default) Option "IncludeImplicitMetaModes" "DisplayDevice = DVI-I-2, Scaling=Aspect-Scaled, UseModePool = false" Option "IncludeImplicitMetaModes" "ExtraResolutions = ( 2560x1440, 320x200 ), DisplayDevice = DVI-I-0" Option "AllowSHMPixmaps" "boolean" This option controls whether applications can use the MIT-SHM X extension to create pixmaps whose contents are shared between the X server and the client. These pixmaps prevent the NVIDIA driver from performing a number of optimizations and degrade performance in many circumstances. Disabling this option disables only shared memory pixmaps. Applications can still use the MIT-SHM extension to transfer data to the X server through shared memory using XShmPutImage. Default: off (shared memory pixmaps are not allowed). Option "SoftwareRenderCacheSize" "boolean" This option controls the size of a cache in system memory used to accelerate software rendering. The size is specified in bytes, but may be rounded or capped based on inherent limits of the cache. Default: 0x800000 (8 Megabytes). Option "AllowIndirectGLXProtocol" "boolean" There are two ways that GLX applications can render on an X screen: direct and indirect. Direct rendering is generally faster and more featureful, but indirect rendering may be used in more configurations. Direct rendering requires that the application be running on the same machine as the X server, and that the OpenGL library have sufficient permissions to access the kernel driver. Indirect rendering works with remote X11 connections as well as unprivileged clients like those in a chroot with no access to device nodes. For those who wish to disable the use of indirect GLX protocol on a given X screen, setting the "AllowIndirectGLXProtocol" to a true value will cause GLX CreateContext requests with the "direct" parameter set to "False" to fail with a BadValue error. Starting with X.Org server 1.16, there are also command-line switches to enable or disable use of indirect GLX contexts. "-iglx" disables use of indirect GLX protocol, and "+iglx" enables use of indirect GLX protocol. +iglx is the default in server 1.16. -iglx is the default in server 1.17 and newer. The NVIDIA GLX implementation will prohibit creation of indirect GLX contexts if the AllowIndirectGLXProtocol option is set to False, or the -iglx switch was passed to the X server (X.Org server 1.16 or higher), or the X server defaulted to '-iglx'. Default: enabled (indirect protocol is allowed, unless disabled by the server). Option "AllowUnofficialGLXProtocol" "boolean" By default, the NVIDIA GLX implementation will not expose GLX protocol for GL commands if the protocol is not considered complete. Protocol could be considered incomplete for a number of reasons. The implementation could still be under development and contain known bugs, or the protocol specification itself could be under development or going through review. If users would like to test the server-side portion of such protocol when using indirect rendering, they can enable this option. If any X screen enables this option, it will enable protocol on all screens in the server. When an NVIDIA GLX client is used, the related environment variable "__GL_ALLOW_UNOFFICIAL_PROTOCOL" will need to be set as well to enable support in the client. Option "PanAllDisplays" "boolean" When this option is enabled, all displays in the current MetaMode will pan as the pointer is moved. If disabled, only the displays whose panning domain contains the pointer (at its new location) are panned. Default: enabled (all displays are panned when the pointer is moved). Option "Interactive" "boolean" This option controls the behavior of the driver's watchdog, which attempts to detect and terminate GPU programs that get stuck, in order to ensure that the GPU remains available for other processes. GPU compute applications, however, often have long-running GPU programs, and killing them would be undesirable. If you are using GPU compute applications and they are getting prematurely terminated, try turning this option off. When this option is set for an X screen, it will be applied to all X screens running on the same GPU. Default: on. The driver will attempt to detect and terminate GPU programs that cause excessive delays for other processes using the GPU. Option "BaseMosaic" "boolean" This option can be used to extend a single X screen transparently across display outputs on each GPU. This is like SLI Mosaic mode except that it does not require a video bridge or nvlink bridge connected to the graphics cards. Due to this Base Mosaic does not guarantee there will be no tearing between the display boundaries. Use this in conjunction with the MetaModes X configuration option to specify the combination of mode(s) used on each display. nvidia-xconfig can be used to configure Base Mosaic via a command like 'nvidia-xconfig --base-mosaic --metamodes=METAMODES' where the METAMODES string specifies the desired grid configuration. For example, to configure four DFPs in a 2x2 configuration, each running at 1920x1024, with two DFPs connected to two cards, the command would be: nvidia-xconfig --base-mosaic --metamodes="GPU-0.DFP-0: 1920x1024+0+0, GPU-0.DFP-1: 1920x1024+1920+0, GPU-1.DFP-0: 1920x1024+0+1024, GPU-1.DFP-1: 1920x1024+1920+1024" Option "ConstrainCursor" "boolean" When this option is enabled, the mouse cursor will be constrained to the region of the desktop that is visible within the union of all displays' panning domains in the current MetaMode. When it is disabled, it may be possible to move the cursor to regions of the X screen that are not visible on any display. Note that if this would make a display's panning domain inaccessible (in other words, if the union of all panning domains is disjoint), then the cursor will not be constrained. This option has no effect if the X server doesn't support cursor constraint. This support was added in X.Org server version 1.10. Default: on, if the X server supports it. The cursor will be constrained to the panning domain of each monitor, when possible. Option "UseHotplugEvents" "boolean" When this option is enabled, the NVIDIA X driver will generate RandR display changed events when displays are plugged into or unplugged from an NVIDIA GPU. Some desktop environments will listen for these events and dynamically reconfigure the desktop when displays are added or removed. Disabling this option suppresses the generation of these RandR events for non-DisplayPort displays, i.e., ones connected via VGA, DVI, or HDMI. Hotplug events cannot be suppressed for displays connected via DisplayPort. Note that probing the display configuration (e.g. with xrandr or nvidia-settings) may cause RandR display changed events to be generated, regardless of whether this option is enabled or disabled. Additionally, some VGA ports are incapable of hotplug detection: on such ports, the addition or removal of displays can only be detected by re-probing the display configuration. Default: on. The driver will generate RandR events when displays are added or removed. Option "AllowEmptyInitialConfiguration" "boolean" Normally, the NVIDIA X driver will start even if there are no display devices connected to the NVIDIA GPU. Manually disabling AllowEmptyInitialConfiguration will prevent the NVIDIA X driver from loading if it cannot find any display devices. Disabling this option makes sense in configurations where X should not start without any displays connected. It would also make sense to disable this option to preserve behavior on older systems. Default: on. The driver will start even if it cannot find any connected display devices. Option "InbandStereoSignaling" "boolean" This option can be used to enable the DisplayPort in-band stereo signaling done via the MISC1 field in the main stream attribute (MSA) data that's sent once per frame during the vertical blanking period of the main video stream. DisplayPort in-band stereo signaling is only available on certain NVIDIA RTX/Quadro boards. This option is implied by stereo mode 14 (Generic active stereo with in-band DP), and selecting that stereo mode will override this option. Default: off. DisplayPort in-band stereo signaling will be disabled. Option "UseSysmemPixmapAccel" "boolean" For discrete GPUs with dedicated video memory, this option allows the GPU to accelerate X drawing operations using system memory in addition to memory on the GPU. Disabling this option is generally not recommended, but it may reduce X driver memory usage in some situations at the cost of some performance. This option does not affect the usage of GPU acceleration for pixmaps bound to GLX drawables, EGL surfaces, or EGL images. GPU acceleration of such pixmaps is critical for interactive performance. Default: on. When video memory is unavailable, the GPU will still attempt to accelerate X drawing operations on pixmaps allocated in system memory. This option is not applicable for Tegra integrated graphics as it does not have dedicated video memory. Option "ForceCompositionPipeline" "string" The NVIDIA X driver can use a composition pipeline to apply X screen transformations and rotations. Normally, this composition pipeline is enabled implicitly when necessary, or when the MetaMode token "ForceCompositionPipeline" is specified. This X configuration option can be used to explicitly enable the composition pipeline, even if the corresponding MetaMode token is not specified. The option value is a comma-separated list of display device names. The composition pipeline will be forced on for all display devices in the comma-separated list. Alternatively, the option value can be any boolean true string ("1", "on", "true", "yes"), in which case all display devices will have their composition pipeline enabled. By default, the option value is NULL. Option "ForceFullCompositionPipeline" "string" This option has the same possible values and semantics as "ForceCompositionPipeline", but it additionally makes use of the composition pipeline to apply ViewPortOut scaling. Option "AllowHMD" "string" Most Virtual Reality Head Mounted Displays (HMDs), such as the HTC VIVE, require special image processing. This means it is usually undesirable to display the X11 desktop on an HMD. By default, the NVIDIA X driver will treat any detected HMDs as disconnected. To override this behavior, set the X configuration option "AllowHMD" to "yes", or explicitly list the HMDs to allow (any other HMDs will continue to be ignored). Examples: Option "AllowHMD" "yes" Option "AllowHMD" "HDMI-0, HDMI-1" Option "TegraReserveDisplayBandwidth" "boolean" The NVIDIA X driver on Tegra dynamically calculates the amount of display bandwidth required for the current configuration of monitors, resolutions, and refresh rates. It uses this calculation to reserve only as much display bandwidth as is necessary for the current configuration, potentially allowing the system to run at lower clocks and conserve power. However, in rare cases where other programs also use display without the X driver's knowledge, the calculated display bandwidth may be insufficient, resulting in underflow. This option can be used to disable the X driver's dynamic display bandwidth reservation, causing the system to assume worst case display usage. This may result in higher power usage, but will avoid the risk of underflow. Default: true (the X driver will only reserve as much display bandwidth as it calculates is necessary). Option "AllowExternalGpus" "boolean" This option allows the NVIDIA X driver to configure X screens on external GPUs, also known as eGPUs. Note that this option is applied globally: setting this option to true will enable the use of all eGPUs. "AllowExternalGpus" defaults to false, to avoid putting the X server in a situation where a GPU it is actively using can be hot-unplugged. External GPUs are often used in short-running compute scenarios, which better tolerate the eGPU being hot-unplugged. In such cases, a different GPU may be used to display the X11 desktop. In addition to eGPUs, "AllowExternalGpus" set to false may prevent the NVIDIA X driver from configuring X screens on GPUs attached to internal PCIe slots with surprise removal/hot-unplug support, such as in some enterprise systems. This option can be placed either in the "Device" or "ServerLayout" section of the X.Org configuration file. Default: false. The NVIDIA X driver will not configure X screens on eGPUs. Option "LimitFrameRateWhenHeadless" "boolean" When there are no active displays, the resulting configuration is "headless". In a headless configuration, OpenGL and Vulkan applications are not able to sync to vblank, and without framerate limiting will run unbounded, potentially wasting system resources and power. To prevent this from happening unintentionally, OpenGL and Vulkan applications that sync to vblank are restricted to 1 frame per second when there are no active displays. If X starts without displays, framerate limiting is disabled to avoid interfering with intended headless configurations. Default: true. OpenGL and Vulkan applications will be limited to 1 frame per second when there are no active displays. If no displays are configured when the NVIDIA X driver initializes, the option will be forced to false. Option "HardDPMS" "boolean" By default, the NVIDIA X driver puts displays to sleep using modesets rather than the VESA DPMS ("Display Power Management Signaling") standard. There should be no visible difference to users of LCDs -- DPMS was originally designed for CRT monitors, and includes intermediate states that are redundant for LCDs. When displays are put to sleep using modesets, they are completely shut down. As a result, OpenGL and Vulkan applications will behave as though the system were headless, and will not be able to sync to vblank, instead being subject to automatic framerate limiting where applicable; see LimitFrameRateWhenHeadless. When "HardDPMS" is set to false, the NVIDIA X driver will put displays to sleep using the VESA DPMS standard rather than modesets. Some LCDs may fail to respond correctly to VESA DPMS, which leaves them powered on when they should be in a sleep state. Default: true. The NVIDIA X driver will put displays to sleep using modesets. Option "SidebandSocketPath" "string" The NVIDIA X driver uses a UNIX domain socket to pass information to other driver components. If unable to connect to this socket, some driver features, such as G-Sync, may not work correctly. The socket will be bound to a file with a name unique to the X server instance created in the directory specified by this option. Note that on Linux, an additional abstract socket (not associated with a file) will also be created, with this pathname socket serving as a fallback if connecting to the abstract socket fails. Default: /var/run bind the pathname socket to a file created in this directory. Option "ConnectToAcpid" "boolean" The ACPI daemon (acpid) receives information about ACPI events like AC/Battery power, docking, etc. acpid will deliver these events to the NVIDIA X driver via a UNIX domain socket connection. By default, the NVIDIA X driver will attempt to connect to acpid to receive these events. Set this option to "off" to prevent the NVIDIA X driver from connecting to acpid. Default: on (the NVIDIA X driver will attempt to connect to acpid). Option "AcpidSocketPath" "string" The NVIDIA X driver attempts to connect to the ACPI daemon (acpid) via a UNIX domain socket. The default path to this socket is "/var/run/acpid.socket". Set this option to specify an alternate path to acpid's socket. Default: "/var/run/acpid.socket". Option "EnableACPIBrightnessHotkeys" "boolean" Enable or disable handling of ACPI brightness change hotkey events. Default: enabled Option "3DVisionUSBPath" "string" When NVIDIA 3D Vision is enabled, the X driver searches through the usbfs to find the connected USB dongle. Set this option to specify the sysfs path of the dongle, from which the X driver will infer the usbfs path. Example: Option "3DVisionUSBPath" "/sys/bus/usb/devices/1-1" Option "3DVisionProConfigFile" "string" NVIDIA 3D VisionPro provides various configuration options and pairs various glasses to sync to the hub. It is convenient to store this configuration information to re-use when X restarts. Filename provided in this option is used by NVIDIA X driver to store this information. Ensure that X server has read and write access permissions to the filename provided. Default: No configuration is stored. Example: Option "3DVisionProConfigFile" "/etc/nvidia_3d_vision_pro_config_filename" Option "3DVisionDisplayType" "integer" When NVIDIA 3D Vision is enabled with a non 3D Vision ready display, use this option to specify the display type. Value Behavior -------------- --------------------------------------------------- 1 Assume it is a CRT. 2 Assume it is a DLP. 3 Assume it is a DLP TV and enable the checkerboard output. Default: 1 Example: Option "3DVisionDisplayType" "1" Option "3DVisionProHwButtonPairing" "boolean" When NVIDIA 3D Vision Pro is enabled, use this option to disable hardware button based pairing. Single click button on the hub to enter into pairing mode which pairs single pair of glasses at a time. Double click button on the hub to enter into a pairing mode which pairs multiple pairs of glasses at a time. Default: True Example: Option "3DVisionProHwButtonPairing" "False" Option "3DVisionProHwSinglePairingTimeout" "integer" When NVIDIA 3D Vision Pro and hardware button based pairing are enabled, use this option to set timeout in seconds for pairing single pair of glasses. Default: 6 Example: Option "3DVisionProHwSinglePairingTimeout" "10" Option "3DVisionProHwMultiPairingTimeout" "integer" When NVIDIA 3D Vision Pro and hardware button based pairing is enabled, use this option to set timeout in seconds for pairing multiple pairs of glasses. Default: 10 Example: Option "3DVisionProHwMultiPairingTimeout" "10" Option "3DVisionProHwDoubleClickThreshold" "integer" When NVIDIA 3D Vision Pro and hardware button based pairing is enabled, use this option to set the threshold for detecting double click event of the button. Threshold is time in ms. within which user has to click the button twice to generate double click event. Default: 1000 ms Example: Option "3DVisionProHwDoubleClickThreshold" "1500" Option "DisableBuiltin3DVisionEmitter" "boolean" This option can be used to disable the NVIDIA 3D Vision infrared emitter that is built into some 3D Vision ready display panels. This can be useful when an external NVIDIA 3D Vision emitter needs to be used with such a panel. Default: False Example: Option "DisableBuiltin3DVisionEmitter" "True" The following NVIDIA X driver options are global to the X server, and should be specified in the "ServerLayout" section of the X configuration file. Option "AllowNVIDIAGPUScreens" "boolean" When this option is enabled, and the X server version supports GPU screens (xorg-server 1.13 or newer), the NVIDIA X driver will allow GPU screens to be created for each NVIDIA GPU detected in the system for which there is no X screen configured or auto-configured. GPU screens within the X server are generally not directly addressable by X11 applications: the 'screen' index in any X protocol request can only specify X screens, not GPU screens. But, GPU screens are useful for PRIME render offload configurations. Default: The NVIDIA X driver will allow GPU screens on X.Org xserver version 1.20.7 and higher. Option "AllowPRIMEDisplayOffloadSink" "boolean" RandR PRIME Display Offload Sink support, also known as "Reverse PRIME", is disabled by default for NVIDIA RandR providers on X.Org X servers prior to version 1.20.7 due to latent bugs that can result in crashes when used with the NVIDIA driver. X server version 1.20.6 contains fixes for the crashing, but cannot be detected automatically. This option overrides the default behavior, allowing PRIME Display Offload Sink to be used regardless of X server version. Default: PRIME Display Offload Sink will be allowed only for X server 1.20.7 or newer. ______________________________________________________________________________ Appendix C. Display Device Names ______________________________________________________________________________ A "display device" refers to a hardware device capable of displaying an image. Most NVIDIA GPUs can drive multiple display devices simultaneously. Many X configuration options can be used to separately configure each display device in use by the X screen. To address an individual display device, you can use one of several names that are assigned to it. For example, the "ModeValidation" X configuration option by default applies to all display devices on the X screen. E.g., Option "ModeValidation" "NoMaxPClkCheck" You can use a display device name qualifier to configure each display device's ModeValidation separately. E.g., Option "ModeValidation" "DFP-0: NoMaxPClkCheck; CRT-1: NoVesaModes" The description of each X configuration option in Appendix B provides more detail on the available syntax for each option. The available display device names vary by GPU. To find all available names for your configuration, start the X server with verbose logging enabled (e.g., `startx -- -logverbose 5`, or enable the "ModeDebug" X configuration option with `nvidia-xconfig --mode-debug` and restart the X server). The X log (normally /var/log/Xorg.0.log) will contain a list of what display devices are valid for the GPU. E.g., (--) NVIDIA(0): Valid display device(s) on Quadro 6000 at PCI:10:0:0 (--) NVIDIA(0): CRT-0 (--) NVIDIA(0): CRT-1 (--) NVIDIA(0): DELL U2410 (DFP-0) (connected) (--) NVIDIA(0): NEC LCD1980SXi (DFP-1) (connected) The X log will also contain a list of which display devices are assigned to the X screen. E.g., (II) NVIDIA(0): Display device(s) assigned to X screen 0: (II) NVIDIA(0): CRT-0 (II) NVIDIA(0): CRT-1 (II) NVIDIA(0): DELL U2410 (DFP-0) (II) NVIDIA(0): NEC LCD1980SXi (DFP-1) Note that when multiple X screens are configured on the same GPU, the NVIDIA X driver assigns different display devices to each X screen. On X servers that support RandR 1.2 or later, the NVIDIA X driver will create an RandR output for each display device assigned to an X screen. The X log will also report a list of "Name Aliases" for each display device. E.g., (--) NVIDIA(0): Name Aliases for NEC LCD1980SXi (DFP-1): (--) NVIDIA(0): DFP (--) NVIDIA(0): DFP-1 (--) NVIDIA(0): DPY-3 (--) NVIDIA(0): DVI-I-3 (--) NVIDIA(0): DPY-EDID-373091cb-5c07-6430-54d2-1112efd64b44 (--) NVIDIA(0): Connector-0 These aliases can be used interchangeably to refer to the same display device in any X configuration option, as an nvidia-settings target specification, or in NV-CONTROL protocol that uses similar strings, such as NV_CTRL_STRING_CURRENT_METAMODE_VERSION_2 (available through the nvidia-settings command line as `nvidia-settings --query CurrentMetaMode`). Each alias has different properties that may affect which alias is appropriate to use. The possible alias names are: o A "type"-based name (e.g., "DFP-1"). This name is a unique index plus a display device type name, though in actuality the "type name" is selected based on the protocol through which the X driver communicates to the display device. If the X driver communicates using VGA, then the name is "CRT"; if the driver communicates using TMDS, LVDS, or DP, then the name is "DFP". This may cause confusion in some cases (e.g., a digital flat panel connected via VGA will have the name "CRT"), but this name alias is provided for backwards compatibility with earlier NVIDIA driver releases. Also for backwards compatibility, an alias is provided that uses the "type name" without an index. This name alias will match any display device of that type: it is not unique across the X screen. Note that the index in this type-based name is based on which physical connector is used. If you reconnect a display device to a different connector on the GPU, the type-based name will be different. o A connector-based name (e.g., "DVI-I-3"). This name is a unique index plus a name that is based on the physical connector through which the display device is connected to the GPU. E.g., "VGA-1", "DVI-I-0", "DVI-D-3", "LVDS-1", "DP-2", "HDMI-3", "eDP-6", "USB-C-0". On X servers that support RandR 1.2 or later, this name is also used as the RandR output name. Note that the index in this connector-based name is based on which physical connector is used. If you reconnect a display device to a different connector on the GPU, the connector-based name will be different. When Mosaic is enabled, this name is prefixed with a GPU identifier to make it unique. For example, a Mosaic configuration with two DisplayPort devices might have two different outputs with names "GPU-0.DP-0" and "GPU-1.DP-0", respectively. See Appendix K for a description of valid GPU names. o An EDID-based name (e.g., "DPY-EDID-373091cb-5c07-6430-54d2-1112efd64b44"). This name is a SHA-1 hash, formatted in canonical UUID 8-4-4-4-12 format, of the display device's EDID. This name will be the same regardless of which physical connector on the GPU you use, but it will not be unique if you have multiple display devices with the same EDID. o An NV-CONTROL target ID-based name (e.g., "DPY-3"). The NVIDIA X driver will assign a unique ID to each display device on the entire X server. These IDs are not guaranteed to be persistent from one run of the X server to the next, so is likely not convenient for X configuration file use. It is more frequently used in communication with NV-CONTROL clients such as nvidia-settings. o A connector number-based name (e.g., "Connector-0"), where each name identifies a physical connector on the graphics card. Connector numbers for these names are guaranteed to begin at 0 and monotonically increase to the maximum number of connectors on the graphics card. Note that multiple display devices may have the same connector number-based name: DisplayPort Multi-Stream devices on the same connector, or display devices representing different protocols that can be possibly driven by the same connector (e.g., DisplayPort versus TMDS protocols driven over a DisplayPort connector). This name is most useful with the "ConnectedMonitor" X configuration option in situations where you want to emulate the presence of one or more connected monitors, but do not necessarily care which physical connectors are used. E.g., Option "ConnectedMonitor" "Connector-0, Connector-1" When DisplayPort 1.2 branch devices are present, display devices will be created with type- and connector-based names that are based on how they are connected to the branch device tree. For example, if a connector named DP-2 has a branch device attached and a DisplayPort device is connected to the branch device's first downstream port, a display device named "DP-2.1" might be created. If another branch device is connected between the first branch device and the display device, the name might be "DP-2.1.1". Any display device name can have an optional GPU qualifier prefix. E.g., "GPU-0.DVI-I-3". This is useful in Mosaic configurations: type- and connector-based display device names are only unique within a GPU, so the GPU qualifier is used to distinguish between identically named display devices on different GPUs. For example: Option "MetaModes" "GPU-0.CRT-0: 1600x1200, GPU-1.CRT-0: 1024x768" If no GPU is specified for a particular display device name, the setting will apply to any devices with that name across all GPUs. Note that the GPU UUID can also be used as the qualifier. E.g., "GPU-758a4cf7-0761-62c7-9bf7-c7d950b817c6.DVI-I-1". See Appendix K For details. ______________________________________________________________________________ Appendix D. GLX Support ______________________________________________________________________________ This release supports GLX 1.4. Additionally, the following GLX extensions are supported on appropriate GPUs: o GLX_EXT_visual_info o GLX_EXT_visual_rating o GLX_SGIX_fbconfig o GLX_SGIX_pbuffer o GLX_ARB_get_proc_address o GLX_SGI_video_sync o GLX_SGI_swap_control o GLX_ARB_multisample o GLX_NV_float_buffer o GLX_ARB_fbconfig_float o GLX_NV_swap_group o GLX_NV_video_out o GLX_EXT_texture_from_pixmap o GLX_NV_copy_image o GLX_ARB_create_context o GLX_EXT_import_context o GLX_EXT_fbconfig_packed_float o GLX_EXT_framebuffer_sRGB o GLX_NV_present_video o GLX_NV_multisample_coverage o GLX_EXT_swap_control o GLX_NV_video_capture o GLX_ARB_create_context_profile o GLX_EXT_create_context_es_profile o GLX_EXT_create_context_es2_profile o GLX_EXT_swap_control_tear o GLX_EXT_buffer_age o GLX_ARB_create_context_robustness o GLX_ARB_context_flush_control o GLX_ARB_create_context_no_error o GLX_EXT_libglvnd o GLX_EXT_stereo_tree o GLX_NV_copy_buffer o GLX_NV_delay_before_swap o GLX_NV_robustness_video_memory_purge o GLX_NV_multigpu_context For a description of these extensions, see the OpenGL extension registry at http://www.opengl.org/registry/ Some of the above extensions exist as part of core GLX 1.4 functionality, however, they are also exported as extensions for backwards compatibility. Unofficial GLX protocol support exists in NVIDIA's GLX client and GLX server implementations for the following OpenGL extensions: o GL_ARB_geometry_shader4 o GL_ARB_shader_objects o GL_ARB_texture_buffer_object o GL_ARB_vertex_buffer_object o GL_ARB_vertex_shader o GL_EXT_bindable_uniform o GL_EXT_compiled_vertex_array o GL_EXT_geometry_shader4 o GL_EXT_gpu_shader4 o GL_EXT_texture_buffer_object o GL_NV_geometry_program4 o GL_NV_vertex_program o GL_NV_parameter_buffer_object o GL_NV_vertex_program4 Until the GLX protocol for these OpenGL extensions is finalized, using these extensions through GLX indirect rendering will require the AllowUnofficialGLXProtocol X configuration option, and the __GL_ALLOW_UNOFFICIAL_PROTOCOL environment variable in the environment of the client application. Unofficial protocol requires the use of NVIDIA GLX libraries on both the client and the server. Note: GLX protocol is used when an OpenGL application indirect renders (i.e., runs on one computer, but submits protocol requests such that the rendering is performed on another computer). The above OpenGL extensions are fully supported when doing direct rendering. GLX visuals and FBConfigs are only available for X screens with depths 16, 24, or 30. ______________________________________________________________________________ Appendix E. Dots Per Inch ______________________________________________________________________________ DPI (Dots Per Inch), also known as PPI (Pixels Per Inch), is a property of an X screen that describes the physical size of pixels. Some X applications, such as xterm, can use the DPI of an X screen to determine how large (in pixels) to draw an object in order for that object to be displayed at the desired physical size on the display device. The DPI of an X screen is computed by dividing the size of the X screen in pixels by the size of the X screen in inches: DPI = SizeInPixels / SizeInInches Since the X screen stores its physical size in millimeters rather than inches (1 inch = 25.4 millimeters): DPI = (SizeInPixels * 25.4) / SizeInMillimeters The NVIDIA X driver reports the size of the X screen in pixels and in millimeters. When the XRandR extension resizes the X screen in pixels, the NVIDIA X driver computes a new size in millimeters for the X screen, to maintain a constant DPI (see the "Physical Size" column of the `xrandr -q` output as an example). This is done because a changing DPI can cause interaction problems for some applications. To disable this behavior, and instead keep the same millimeter size for the X screen (and therefore have a changing DPI), set the ConstantDPI option to FALSE. You can query the DPI of your X screen by running: % xdpyinfo | grep -B1 dot which should generate output like this: dimensions: 1280x1024 pixels (382x302 millimeters) resolution: 85x86 dots per inch The NVIDIA X driver performs several steps during X screen initialization to determine the DPI of each X screen: o If the display device provides an EDID, and the EDID contains information about the physical size of the display device, that is used to compute the DPI, along with the size in pixels of the first mode to be used on the display device. Note that in some cases, the physical size information stored in a display device's EDID may be unreliable. This could result in a display device's DPI being computed incorrectly, potentially leading to undesired consequences such as fonts that are scaled larger or smaller than expected. These issues can be worked around by manually setting a DPI using the "DPI" X configuration option, or by disabling the use of the EDID's physical size information for computing DPI by setting the UseEdidDpi X configuration option to "FALSE"'. If multiple display devices are used by this X screen, then the NVIDIA X screen will choose which display device to use. You can override this with the "UseEdidDpi" X configuration option: you can specify a particular display device to use; e.g.: Option "UseEdidDpi" "DFP-1" or disable EDID-computed DPI by setting this option to false: Option "UseEdidDpi" "FALSE" EDID-based DPI computation is enabled by default when an EDID is available. o If the "-dpi" commandline option to the X server is specified, that is used to set the DPI (see `X -h` for details). This will override the "UseEdidDpi" option. o If the DPI X configuration option is specified, that will be used to set the DPI. This will override the "UseEdidDpi" option. o If none of the above are available, then the "DisplaySize" X config file Monitor section information will be used to determine the DPI, if provided; see the xorg.conf man page for details. o If none of the above are available, the DPI defaults to 75x75. You can find how the NVIDIA X driver determined the DPI by looking in your X log file. There will be a line that looks something like the following: (--) NVIDIA(0): DPI set to (101, 101); computed from "UseEdidDpi" X config option Note that the physical size of the X screen, as reported through `xdpyinfo` is computed based on the DPI and the size of the X screen in pixels. The DPI of an X screen can be poorly defined when multiple display devices are enabled on the X screen: those display devices might have different actual DPIs, yet DPI is advertised from the X server to the X application with X screen granularity. Solutions for this include: o Use separate X screens, with one display device on each X screen; see Chapter 14 for details. o The RandR X extension version 1.2 and later reports the physical size of each RandR Output, so applications could possibly choose to render content at different sizes, depending on which portion of the X screen is displayed on which display devices. Client applications can also configure the reported per-RandR Output physical size. See, e.g., the xrandr(1) '--fbmm' command line option. o Experiment with different DPI settings to find a DPI that is suitable for all display devices on the X screen. ______________________________________________________________________________ Appendix F. i2c Bus Support ______________________________________________________________________________ The NVIDIA Linux kernel module now includes i2c (also called I-squared-c, Inter-IC Communications, or IIC) functionality that allows the NVIDIA Linux kernel module to export i2c ports found on board NVIDIA cards to the Linux kernel. This allows i2c devices on-board the NVIDIA graphics card, as well as devices connected to the VGA and/or DVI ports, to be accessed from kernel modules or userspace programs in a manner consistent with other i2c ports exported by the Linux kernel through the i2c framework. You must have i2c support compiled into the kernel, or as a module, and X must be running. The Linux kernel documentation covers the kernel and userspace /dev APIs, which you may wish to use to access NVIDIA i2c ports. For further information regarding the Linux kernel's i2c framework, refer to the documentation for your kernel, located at .../Documentation/i2c/ within the kernel source tree. The following functionality is currently supported: I2C_FUNC_I2C I2C_FUNC_SMBUS_QUICK I2C_FUNC_SMBUS_BYTE I2C_FUNC_SMBUS_BYTE_DATA I2C_FUNC_SMBUS_WORD_DATA ______________________________________________________________________________ Appendix G. VDPAU Support ______________________________________________________________________________ This release includes support for the Video Decode and Presentation API for Unix-like systems (VDPAU) on most GeForce 8 series and newer add-in cards, as well as motherboard chipsets with integrated graphics that have PureVideo support based on these GPUs. Use of VDPAU requires installation of a separate wrapper library called libvdpau. Please see your system distributor's documentation for information on how to install this library. More information can be found at http://freedesktop.org/wiki/Software/VDPAU/. VDPAU is only available for X screens with depths 16, 24, or 30. VDPAU supports Xinerama. The following restrictions apply: o Physical X screen 0 must be driven by the NVIDIA driver. o VDPAU will only display on physical X screens driven by the NVIDIA driver, and which are driven by a GPU both compatible with VDPAU, and compatible with the GPU driving physical X screen 0. Under Xinerama, VDPAU performs all operations other than display on a single GPU. By default, the GPU associated with physical X screen 0 is used. The environment variable VDPAU_NVIDIA_XINERAMA_PHYSICAL_SCREEN may be used to specify a physical screen number, and then VDPAU will operate on the GPU associated with that physical screen. This variable should be set to the integer screen number as configured in the X configuration file. The selected physical X screen must be driven by the NVIDIA driver. G1. IMPLEMENTATION LIMITS VDPAU is specified as a generic API - the choice of which features to support, and performance levels of those features, is left up to individual implementations. The details of NVIDIA's implementation are provided below. VDPVIDEOSURFACE The maximum supported resolution is 8192x8192 for GPUs with VDPAU feature sets H and I, and 4096x4096 for all other GPUs. The following surface formats and get-/put-bits combinations are supported: o VDP_CHROMA_TYPE_420 (Supported get-/put-bits formats are VDP_YCBCR_FORMAT_NV12, VDP_YCBCR_FORMAT_YV12) o VDP_CHROMA_TYPE_422 (Supported get-/put-bits formats are VDP_YCBCR_FORMAT_UYVY, VDP_YCBCR_FORMAT_YUYV) VDPBITMAPSURFACE The maximum supported resolution is 16384x16384 pixels. The following surface formats are supported: o VDP_RGBA_FORMAT_B8G8R8A8 o VDP_RGBA_FORMAT_R8G8B8A8 o VDP_RGBA_FORMAT_B10G10R10A2 o VDP_RGBA_FORMAT_R10G10B10A2 o VDP_RGBA_FORMAT_A8 Note that VdpBitmapSurfaceCreate's frequently_accessed parameter directly controls whether the bitmap data will be placed into video RAM (VDP_TRUE) or system memory (VDP_FALSE). Note that if the bitmap data cannot be placed into video RAM when requested due to resource constraints, the implementation will automatically fall back to placing the data into system RAM. VDPOUTPUTSURFACE The maximum supported resolution is 16384x16384 pixels. The following surface formats are supported: o VDP_RGBA_FORMAT_B8G8R8A8 o VDP_RGBA_FORMAT_R10G10B10A2 For all surface formats, the following get-/put-bits indexed formats are supported: o VDP_INDEXED_FORMAT_A4I4 o VDP_INDEXED_FORMAT_I4A4 o VDP_INDEXED_FORMAT_A8I8 o VDP_INDEXED_FORMAT_I8A8 For all surface formats, the following get-/put-bits YCbCr formats are supported: o VDP_YCBCR_FORMAT_Y8U8V8A8 o VDP_YCBCR_FORMAT_V8U8Y8A8 VDPDECODER In all cases, VdpDecoder objects solely support 8-bit 4:2:0 streams, and only support writing to VDP_CHROMA_TYPE_420 surfaces. The exact set of supported VdpDecoderProfile values depends on the GPU in use. Appendix A lists which GPUs support which video feature set. An explanation of each video feature set may be found below. When reading these lists, please note that VC1_SIMPLE and VC1_MAIN may be referred to as WMV, WMV3, or WMV9 in other contexts. Partial acceleration means that VLD (bitstream) decoding is performed on the CPU, with the GPU performing IDCT and motion compensation. Complete acceleration means that the GPU performs all of VLD, IDCT, and motion compensation. VDPAU FEATURE SETS A AND B GPUs with VDPAU feature sets A and B are not supported by this driver. VDPAU FEATURE SETS C, D, AND E GPUs with VDPAU feature set C, D, or E support at least the following VdpDecoderProfile values, and associated limits: o VDP_DECODER_PROFILE_MPEG1, VDP_DECODER_PROFILE_MPEG2_SIMPLE, VDP_DECODER_PROFILE_MPEG2_MAIN: o Complete acceleration. o Minimum width or height: 3 macroblocks (48 pixels). o Maximum width or height: 128 macroblocks (2048 pixels) for feature set C, 252 macroblocks (4032 pixels) wide by 253 macroblocks (4048 pixels) high for feature set D, 255 macroblocks (4080 pixels) for feature set E. o Maximum macroblocks: 8192 for feature set C, 65536 for feature sets D or E. o VDP_DECODER_PROFILE_H264_MAIN, VDP_DECODER_PROFILE_H264_HIGH, VDP_DECODER_PROFILE_H264_CONSTRAINED_BASELINE, VDP_DECODER_PROFILE_H264_PROGRESSIVE_HIGH, VDP_DECODER_PROFILE_H264_CONSTRAINED_HIGH: o Complete acceleration. o Minimum width or height: 3 macroblocks (48 pixels). o Maximum width or height: 128 macroblocks (2048 pixels) for feature set C, 252 macroblocks (4032 pixels) wide by 255 macroblocks (4080 pixels) high for feature set D, 256 macroblocks (4096 pixels) for feature set E. o Maximum macroblocks: 8192 for feature set C, 65536 for feature sets D or E. o VDP_DECODER_PROFILE_H264_BASELINE, VDP_DECODER_PROFILE_H264_EXTENDED: o Partial acceleration. The NVIDIA VDPAU implementation does not support flexible macroblock ordering, arbitrary slice ordering, redundant slices, data partitioning, SI slices, or SP slices. Content utilizing these features may decode with visible corruption. o Minimum width or height: 3 macroblocks (48 pixels). o Maximum width or height: 128 macroblocks (2048 pixels) for feature set C, 252 macroblocks (4032 pixels) wide by 255 macroblocks (4080 pixels) high for feature set D, 256 macroblocks (4096 pixels) for feature set E. o Maximum macroblocks: 8192 for feature set C, 65536 for feature sets D or E. o VDP_DECODER_PROFILE_VC1_SIMPLE, VDP_DECODER_PROFILE_VC1_MAIN, VDP_DECODER_PROFILE_VC1_ADVANCED: o Complete acceleration. o Minimum width or height: 3 macroblocks (48 pixels). o Maximum width or height: 128 macroblocks (2048 pixels). o Maximum macroblocks: 8190 o VDP_DECODER_PROFILE_MPEG4_PART2_SP, VDP_DECODER_PROFILE_MPEG4_PART2_ASP, VDP_DECODER_PROFILE_DIVX4_QMOBILE, VDP_DECODER_PROFILE_DIVX4_MOBILE, VDP_DECODER_PROFILE_DIVX4_HOME_THEATER, VDP_DECODER_PROFILE_DIVX4_HD_1080P, VDP_DECODER_PROFILE_DIVX5_QMOBILE, VDP_DECODER_PROFILE_DIVX5_MOBILE, VDP_DECODER_PROFILE_DIVX5_HOME_THEATER, VDP_DECODER_PROFILE_DIVX5_HD_1080P o Complete acceleration. o Minimum width or height: 3 macroblocks (48 pixels). o Maximum width or height: 128 macroblocks (2048 pixels). o Maximum macroblocks: 8192 The following features are currently not supported: o GMC (Global Motion Compensation) o Data partitioning o reversible VLC These GPUs also support VDP_VIDEO_MIXER_FEATURE_HIGH_QUALITY_SCALING_L1. GPUs with VDPAU feature set E support an enhanced error concealment mode which provides more robust error handling when decoding corrupted video streams. This error concealment is on by default, and may have a minor CPU performance impact in certain configurations. To disable this, set the environment variable VDPAU_NVIDIA_DISABLE_ERROR_CONCEALMENT to 1. VDPAU FEATURE SET F GPUs with VDPAU feature set F support all of the same VdpDecoderProfile values and other features as VDPAU feature set E. Feature set F adds: o VDP_DECODER_PROFILE_HEVC_MAIN, VDP_DECODER_PROFILE_HEVC_MAIN_10: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 4096 luma samples (pixels) wide by 2304 luma samples (pixels) tall. o Maximum macroblocks: not applicable. o VDP_DECODER_PROFILE_VP9_PROFILE_0: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 4096 luma samples (pixels) wide by 2304 luma samples (pixels) tall. o Maximum macroblocks: not applicable. VDPAU FEATURE SET G GPUs with VDPAU feature set G support all of the same VdpDecoderProfile values and other features as VDPAU feature set F. Feature set G adds: o VDP_DECODER_PROFILE_HEVC_MAIN_12: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 4096 luma samples (pixels) wide by 4096 luma samples (pixels) tall. o Maximum macroblocks: not applicable. VDPAU FEATURE SET H GPUs with VDPAU feature set H support all of the same VdpDecoderProfile values and other features as VDPAU feature set G. Feature set H adds: o VDP_DECODER_PROFILE_HEVC_MAIN, VDP_DECODER_PROFILE_HEVC_MAIN_10 VDP_DECODER_PROFILE_HEVC_MAIN_12: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 8192 luma samples (pixels) wide by 8192 luma samples (pixels) tall. o Maximum macroblocks: not applicable. o VDP_DECODER_PROFILE_VP9_PROFILE_0, VDP_DECODER_PROFILE_VP9_PROFILE_2: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 8192 luma samples (pixels) wide by 8192 luma samples (pixels) tall. o Maximum macroblocks: not applicable. VDPAU FEATURE SET I GPUs with VDPAU feature set I support all of the same VdpDecoderProfile values and other features as VDPAU feature set H. VDPAU FEATURE SET J GPUs with VDPAU feature set J support all of the same VdpDecoderProfile values and other features as VDPAU feature set H. Feature set J adds: o VDP_DECODER_PROFILE_HEVC_MAIN_444, VDP_DECODER_PROFILE_HEVC_MAIN_444_10 VDP_DECODER_PROFILE_HEVC_MAIN_444_12: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 8192 luma samples (pixels) wide by 8192 luma samples (pixels) tall. o Maximum macroblocks: not applicable. VDPAU FEATURE SET K GPUs with VDPAU feature set K support all of the same VdpDecoderProfile values and other features as VDPAU feature set J.Feature set K adds: o VDP_DECODER_PROFILE_AV1_MAIN: o Complete acceleration. o Minimum width or height: 128 luma samples (pixels). o Maximum width or height: 8192 luma samples (pixels) wide by 8192 luma samples (pixels) tall. o Maximum macroblocks: not applicable. VDPAU FEATURES NOTE 2 Note that all GPUs with VDPAU feature sets H and above, except GPUs with this note, support VDP_DECODER_PROFILE_VP9_PROFILE_2. Please check "VDPAU information" page in nvidia-settings for the list of supported profiles. VDPAU FEATURES NOTE 3 Note that codec support may vary by product manufacturer and region. For further details, please consult the documentation provided by the Add-In Card manufacturer or system manufacturer of your product. VDPVIDEOMIXER The maximum supported resolution is 8192x8192 for GPUs with VDPAU feature set H, and 4096x4096 for all other GPUs. The video mixer supports all video and output surface resolutions and formats that the implementation supports. The video mixer supports at most 4 auxiliary layers. The following features are supported: o VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL o VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL o VDP_VIDEO_MIXER_FEATURE_INVERSE_TELECINE o VDP_VIDEO_MIXER_FEATURE_NOISE_REDUCTION o VDP_VIDEO_MIXER_FEATURE_SHARPNESS o VDP_VIDEO_MIXER_FEATURE_LUMA_KEY In order for either VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL or VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL to operate correctly, the application must supply at least 2 past and 1 future fields to each VdpMixerRender call. If those fields are not provided, the VdpMixer will fall back to bob de-interlacing. Both regular de-interlacing and half-rate de-interlacing are supported. Both have the same requirements in terms of the number of past/future fields required. Both modes should produce equivalent results. In order for VDP_VIDEO_MIXER_FEATURE_INVERSE_TELECINE to have any effect, one of VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL or VDP_VIDEO_MIXER_FEATURE_DEINTERLACE_TEMPORAL_SPATIAL must be requested and enabled. Inverse telecine has the same requirement on the minimum number of past/future fields that must be provided. Inverse telecine will not operate when "half-rate" de-interlacing is used. While it is possible to apply de-interlacing algorithms to progressive streams using the techniques outlined in the VDPAU documentation, NVIDIA does not recommend doing so. One is likely to introduce more artifacts due to the inverse telecine process than are removed by detection of bad edits etc. VDPPRESENTATIONQUEUE The resolution of VdpTime is approximately 10 nanoseconds. At some arbitrary point during system startup, the initial value of this clock is synchronized to the system's real-time clock, as represented by nanoseconds since since Jan 1, 1970. However, no attempt is made to keep the two time-bases synchronized after this point. Divergence can and will occur. NVIDIA's VdpPresentationQueue supports two methods for displaying surfaces; overlay and blit. The overlay method will be used wherever possible, with the blit method acting as a more general fallback. Whenever a presentation queue is created, the driver determines whether the overlay method may ever be used, based on system configuration, and whether any other application already owns the overlay. If overlay usage is potentially possible, the presentation queue is marked as owning the overlay. Whenever a surface is displayed, the driver determines whether the overlay method may be used for that frame, based on both whether the presentation queue owns the overlay, and the set of overlay usage limitations below. In other words, the driver may switch back and forth between overlay and blit methods dynamically. The most likely cause for dynamic switching is when a compositing manager is enabled or disabled, and the window becomes redirected or unredirected. The following conditions or system configurations will prevent usage of the overlay path: o Overlay hardware already in use, e.g. by another VDPAU, GL, or X11 application. o Desktop rotation enabled on the given X screen. o The presentation target window is redirected, due to a compositing manager actively running. o The environment variable VDPAU_NVIDIA_NO_OVERLAY is set to a string representation of a non-zero integer. o The driver determines that the performance requirements of overlay usage cannot be met by the current hardware configuration. Both the overlay and blit methods sync to VBLANK. The overlay path is guaranteed never to tear, whereas the blit method is classed as "best effort". When TwinView is enabled, the blit method can only sync to one of the display devices; this may cause tearing corruption on the display device to which VDPAU is not syncing. You can use the environment variable VDPAU_NVIDIA_SYNC_DISPLAY_DEVICE to specify the display device to which VDPAU should sync. You should set this environment variable to the name of a display device, for example "CRT-1". Look for the line "Connected display device(s):" in your X log file for a list of the display devices present and their names. You may also find it useful to review Chapter 12 "Configuring Twinview" and the section on Ensuring Identical Mode Timings in Chapter 18. A VdpPresentationQueue allows a maximum of 8 surfaces to be QUEUED or VISIBLE at any one time. This limit is per presentation queue. If this limit is exceeded, VdpPresentationQueueDisplay blocks until an entry in the presentation queue becomes free. G2. PERFORMANCE LEVELS This documentation describes the capabilities of the NVIDIA VDPAU implementation. Hardware performance may vary significantly between cards. No guarantees are made, nor implied, that any particular combination of system configuration, GPU configuration, VDPAU feature set, VDPAU API usage, application, video stream, etc., will be able to decode streams at any particular frame rate. G3. GETTING THE BEST PERFORMANCE FROM THE API System performance (raw throughput, latency, and jitter tolerance) can be affected by a variety of factors. One of these factors is how the client application uses VDPAU; i.e. the number of surfaces allocated for buffering, order of operations, etc. NVIDIA GPUs typically contain a number of separate hardware modules that are capable of performing different parts of the video decode, post-processing, and display operations in parallel. To obtain the best performance, the client application must attempt to keep all these modules busy with work at all times. Consider the decoding process. At a bare minimum, the application must allocate one video surface for each reference frame that the stream can use (2 for MPEG or VC-1, a variable stream-dependent number for H.264) plus one surface for the picture currently being decoded. However, if this minimum number of surfaces is used, performance may be poor. This is because back-to-back decodes of non-reference frames will need to be written into the same video surface. This will require that decode of the second frame wait until decode of the first has completed; a pipeline stall. Further, if the video surfaces are being read by the video mixer for post-processing, and eventual display, this will "lock" the surfaces for even longer, since the video mixer needs to read the data from the surface, which prevents any subsequent decode operations from writing to the surface. Recall that when advanced de-interlacing techniques are used, a history of video surfaces must be provided to the video mixer, thus necessitating that even more video surfaces be allocated. For this reason, NVIDIA recommends the following number of video surfaces be allocated: o (num_ref + 3) for progressive content, and no de-interlacing. o (num_ref + 5) for interlaced content using advanced de-interlacing. Next, consider the display path via the presentation queue. This portion of the pipeline requires at least 2 output surfaces; one that is being actively displayed by the presentation queue, and one being rendered to for subsequent display. As before, using this minimum number of surfaces may not be optimal. For some video streams, the hardware may only achieve real-time decoding on average, not for each individual frame. Using compositing APIs to render on-screen displays, graphical user interfaces, etc., may introduce extra jitter and latency into the pipeline. Similarly, system level issues such as scheduler algorithms and system load may prevent the CPU portion of the driver from operating for short periods of time. All of these potential issues may be solved by allocating more output surfaces, and queuing more than one outstanding output surface into the presentation queue. The reason for using more than the minimum number of video surfaces is to ensure that the decoding and post-processing pipeline is not stalled, and hence is kept busy for the maximum amount of time possible. In contrast, the reason for using more than the minimum number of output surfaces is to hide jitter and latency in various GPU and CPU operations. The choice of exactly how many surfaces to allocate is a resource usage v.s. performance trade-off; Allocating more than the minimum number of surfaces will increase performance, but use proportionally more video RAM. This may cause allocations to fail. This could be particularly problematic on systems with a small amount of video RAM. A stellar application would automatically adjust to this by initially allocating the bare minimum number of surfaces (failures being fatal), then attempting to allocate more and more surfaces, provided the allocations kept succeeding, up to the suggested limits above. The video decoder's memory usage is also proportional to the maximum number of reference frames specified at creation time. Requesting a larger number of reference frames can significantly increase memory usage. Hence it is best for applications that decode H.264 to request only the actual number of reference frames specified in the stream, rather than e.g. hard-coding a limit of 16, or even the maximum number of surfaces allowable by some specific H.264 level at the stream's resolution. Note that the NVIDIA implementation correctly implements all required interlocks between the various pipelined hardware modules. Applications never need worry about correctness (providing their API usage is legal and sensible), but simply have to worry about performance. G4. ADDITIONAL NOTES Note that output and bitmap surfaces are not cleared to any specific value upon allocation. It is the application's responsibility to initialize all surfaces prior to using them as input to any function. Video surfaces are cleared to black upon allocation. G5. DEBUGGING AND TRACING The VDPAU wrapper library supports tracing VDPAU function calls, and their parameters. This tracing is controlled by the following environment variables: VDPAU_TRACE Enables tracing. Set to 1 to trace function calls. Set to 2 to trace all arguments passed to the function. VDPAU_TRACE_FILE Filename to write traces to. By default, traces are sent to stderr. This variable may either contain a plain filename, or a reference to an existing open file-descriptor in the format "&N" where N is the file descriptor number. The VDPAU wrapper library is responsible for determining which vendor-specific driver to load for a given X11 display/screen. At present, it hard-codes "nvidia" as the driver. The environment variable VDPAU_DRIVER may be set to override this default. The actual library loaded will be libvdpau_${VDPAU_DRIVER}.so. Setting VDPAU_DRIVER to "trace" is not advised. The NVIDIA VDPAU driver can emit some diagnostic information when an error occurs. To enable this, set the environment variable VDPAU_NVIDIA_DEBUG. A value of 1 will request a small diagnostic that will enable NVIDIA engineers to locate the source of the problem. A value of 3 will request that a complete stack backtrace be printed, which provide NVIDIA engineers with more detailed information, which may be needed to diagnose some problems. G6. MULTI-THREADING VDPAU supports multiple threads actively executing within the driver, subject to certain limitations. If any object is being created or destroyed, the VDPAU driver will become single-threaded. This includes object destruction during preemption cleanup. Otherwise, up to one thread may actively execute VdpDecoderRender per VdpDecoder object, and up to one thread may actively execute any other rendering API per VdpDevice (or child) object. Note that the driver enforces these restrictions internally; applications are not required to implement the rules outlined above. Finally, some of the "query" or "get" APIs may actively execute irrespective of the number of rendering threads currently executing. G7. XWAYLAND SUPPORT IN VDPAU The NVIDIA VDPAU driver supports running within the X Window System: either on the native X.org X server, or on the Xwayland X server within an Xwayland environment. VDPAU applications do not have to change their VDPAU API usage depending on the X server. REQUIREMENTS In order to make use of Xwayland for decode and presentation purposes, certain requirements need to be satisfied. See section "39B. REQUIREMENTS" for the list of requirements. KNOWN ISSUES o The current implementation of VDPAU with Xwayland does not honor the "Presentation Time Stamp" passed by the application to the VDPAU driver. o The NVIDIA VDPAU driver running in an Xwayland environment in some cases may lead to corruption and flickering during a video playback. This is a known issue and is tracked by: https://gitlab.freedesktop.org/xorg/xserver/-/issues/1317 ______________________________________________________________________________ Appendix H. Audio Support ______________________________________________________________________________ Many NVIDIA GPUs support embedding an audio stream in HDMI and DisplayPort signals. In most cases, the GPU contains a standard HD-Audio controller, for which there is a standard ALSA driver. For more details on how to configure and use the ALSA driver with NVIDIA hardware, please see https://download.nvidia.com/XFree86/gpu-hdmi-audio-document/. ______________________________________________________________________________ Appendix I. Tips for New Linux Users ______________________________________________________________________________ This installation guide assumes that the user has at least a basic understanding of Linux techniques and terminology. In this section we provide tips that the new user may find helpful. While the these tips are meant to clarify and assist users in installing and configuring the NVIDIA Linux Driver, it is by no means a tutorial on the use or administration of the Linux operating system. Unlike many desktop operating systems, it is relatively easy to cause irreparable damage to your Linux system. If you are unfamiliar with the use of Linux, we strongly recommend that you seek a tutorial through your distributor before proceeding. I1. THE COMMAND PROMPT While newer releases of Linux bring new desktop interfaces to the user, much of the work in Linux takes place at the command prompt. If you are familiar with the Windows operating system, the Linux command prompt is analogous to the Windows command prompt, although the syntax and use varies somewhat. All of the commands in this section are performed at the command prompt. Some systems are configured to boot into console mode, in which case the user is presented with a prompt at login. Other systems are configured to start the X window system, in which case the user must open a terminal or console window in order to get a command prompt. This can usually be done by searching the desktop menus for a terminal or console program. While it is customizable, the basic prompt usually consists of a short string of information, one of the characters '#', '$', or '%', and a cursor (possibly flashing) that indicates where the user's input will be displayed. I2. NAVIGATING THE DIRECTORY STRUCTURE Linux has a hierarchical directory structure. From anywhere in the directory structure, the 'ls' command will list the contents of that directory. The 'file' command will print the type of files in a directory. For example, % file filename will print the type of the file 'filename'. Changing directories is done with the 'cd' command. % cd dirname will change the current directory to 'dirname'. From anywhere in the directory structure, the command 'pwd' will print the name of the current directory. There are two special directories, '.' and '..', which refer to the current directory and the next directory up the hierarchy, respectively. For any commands that require a file name or directory name as an argument, you may specify the absolute or the relative paths to those elements. An absolute path begins with the "/" character, referring to the top or root of the directory structure. A relative path begins with a directory in the current working directory. The relative path may begin with '.' or '..'. Elements of a path are separated with the "/" character. As an example, if the current directory is '/home/jesse' and the user wants to change to the '/usr/local' directory, he can use either of the following commands to do so: % cd /usr/local or % cd ../../usr/local I3. FILE PERMISSIONS AND OWNERSHIP All files and directories have permissions and ownership associated with them. This is useful for preventing non-administrative users from accidentally (or maliciously) corrupting the system. The permissions and ownership for a file or directory can be determined by passing the -l option to the 'ls' command. For example: % ls -l drwxr-xr-x 2 jesse users 4096 Feb 8 09:32 bin drwxrwxrwx 10 jesse users 4096 Feb 10 12:04 pub -rw-r--r-- 1 jesse users 45 Feb 4 03:55 testfile -rwx------ 1 jesse users 93 Feb 5 06:20 myprogram -rw-rw-rw- 1 jesse users 112 Feb 5 06:20 README % The first character column in the first output field states the file type, where 'd' is a directory and '-' is a regular file. The next nine columns specify the permissions (see paragraph below) of the element. The second field indicates the number of files associated with the element, the third field indicates the owner, the fourth field indicates the group that the file is associated with, the fifth field indicates the size of the element in bytes, the sixth, seventh and eighth fields indicate the time at which the file was last modified and the ninth field is the name of the element. As stated, the last nine columns in the first field indicate the permissions of the element. These columns are grouped into threes, the first grouping indicating the permissions for the owner of the element ('jesse' in this case), the second grouping indicating the permissions for the group associated with the element, and the third grouping indicating the permissions associated with the rest of the world. The 'r', 'w', and 'x' indicate read, write and execute permissions, respectively, for each of these associations. For example, user 'jesse' has read and write permissions for 'testfile', users in the group 'users' have read permission only, and the rest of the world also has read permissions only. However, for the file 'myprogram', user 'jesse' has read, write and execute permissions (suggesting that 'myprogram' is a program that can be executed), while the group 'users' and the rest of the world have no permissions (suggesting that the owner doesn't want anyone else to run his program). The permissions, ownership and group associated with an element can be changed with the commands 'chmod', 'chown' and 'chgrp', respectively. If a user with the appropriate permissions wanted to change the user/group ownership of 'README' from jesse/users to joe/admin, he would do the following: # chown joe README # chgrp admin README The syntax for chmod is slightly more complicated and has several variations. The most concise way of setting the permissions for a single element uses a triplet of numbers, one for each of user, group and world. The value for each number in the triplet corresponds to a combination of read, write and execute permissions. Execute only is represented as 1, write only is represented as 2, and read only is represented as 4. Combinations of these permissions are represented as sums of the individual permissions. Read and execute is represented as 5, where as read, write and execute is represented as 7. No permissions is represented as 0. Thus, to give the owner read, write and execute permissions, the group read and execute permissions and the world no permissions, a user would do as follows: % chmod 750 myprogram I4. THE SHELL The shell provides an interface between the user and the operating system. It is the job of the shell to interpret the input that the user gives at the command prompt and call upon the system to do something in response. There are several different shells available, each with somewhat different syntax and capabilities. The two most common flavors of shells used on Linux stem from the Bourne shell ('sh') and the C-shell ('csh') Different users have preferences and biases towards one shell or the other, and some certainly make it easier (or at least more intuitive) to do some things than others. You can determine your current shell by printing the value of the 'SHELL' environment variable from the command prompt with % echo $SHELL You can start a new shell simply by entering the name of the shell from the command prompt: % csh or % sh and you can run a program from within a specific shell by preceding the name of the executable with the name of the shell in which it will be run: % sh myprogram The user's default shell at login is determined by whoever set up his account. While there are many syntactic differences between shells, perhaps the one that is encountered most frequently is the way in which environment variables are set. I5. SETTING ENVIRONMENT VARIABLES Every session has associated with it environment variables, which consist of name/value pairs and control the way in which the shell and programs run from the shell behave. An example of an environment variable is the 'PATH' variable, which tells the shell which directories to search when trying to locate an executable file that the user has entered at the command line. If you are certain that a command exists, but the shell complains that it cannot be found when you try to execute it, there is likely a problem with the 'PATH' variable. Environment variables are set differently depending on the shell being used. For the Bourne shell ('sh'), it is done as: % export MYVARIABLE="avalue" for the C-shell, it is done as: % setenv MYVARIABLE "avalue" In both cases the quotation marks are only necessary if the value contains spaces. The 'echo' command can be used to examine the value of an environment variable: % echo $MYVARIABLE Commands to set environment variables can also include references to other environment variables (prepended with the "$" character), including themselves. In order to add the path '/usr/local/bin' to the beginning of the search path, and the current directory '.' to the end of the search path, a user would enter % export PATH=/usr/local/bin:$PATH:. in the Bourne shell, and % setenv PATH /usr/local/bin:${PATH}:. in C-shell. Note the curly braces are required to protect the variable name in C-shell. I6. EDITING TEXT FILES There are several text editors available for the Linux operating system. Some of these editors require the X window system, while others are designed to operate in a console or terminal. It is generally a good thing to be competent with a terminal-based text editor, as there are times when the files necessary for X to run are the ones that must be edited. Three popular editors are 'vi', 'pico' and 'emacs', each of which can be started from the command line, optionally supplying the name of a file to be edited. 'vi' is arguably the most ubiquitous as well as the least intuitive of the three. 'pico' is relatively straightforward for a new user, though not as often installed on systems. If you don't have 'pico', you may have a similar editor called 'nano'. 'emacs' is highly extensible and fairly widely available, but can be somewhat unwieldy in a non-X environment. The newer versions each come with online help, and offline help can be found in the manual and info pages for each (see the section on Linux Manual and Info pages). Many programs use the 'EDITOR' environment variable to determine which text editor to start when editing is required. I7. ROOT USER Upon installation, almost all distributions set up the default administrative user with the username 'root'. There are many things on the system that only 'root' (or a similarly privileged user) can do, one of which is installing the NVIDIA Linux Driver. WE MUST EMPHASIZE THAT ASSUMING THE IDENTITY OF 'root' IS INHERENTLY RISKY AND AS 'root' IT IS RELATIVELY EASY TO CORRUPT YOUR SYSTEM OR OTHERWISE RENDER IT UNUSABLE. There are three ways to become 'root'. You may log in as 'root' as you would any other user, you may use the switch user command ('su') at the command prompt, or, on some systems, use the 'sudo' utility, which allows users to run programs as 'root' while keeping a log of their actions. This last method is useful in case a user inadvertently causes damage to the system and cannot remember what he has done (or prefers not to admit what he has done). It is generally a good practice to remain 'root' only as long as is necessary to accomplish the task requiring 'root' privileges (another useful feature of the 'sudo' utility). I8. BOOTING TO A DIFFERENT RUNLEVEL Runlevels in Linux dictate which services are started and stopped automatically when the system boots or shuts down. The runlevels typically range from 0 to 6, with runlevel 5 typically starting the X window system as part of the services (runlevel 0 is actually a system halt, and 6 is a system reboot). It is good practice to install the NVIDIA Linux Driver while X is not running, and it is a good idea to prevent X from starting on reboot in case there are problems with the installation (otherwise you may find yourself with a broken system that automatically tries to start X, but then hangs during the startup, preventing you from doing the repairs necessary to fix X). Depending on your network setup, runlevels 1, 2 or 3 should be sufficient for installing the Driver. Level 3 typically includes networking services, so if utilities used by the system during installation depend on a remote filesystem, Levels 1 and 2 will be insufficient. If your system typically boots to a console with a command prompt, you should not need to change anything. If your system typically boots to the X window system with a graphical login and desktop, you must both exit X and change your default runlevel. On most distributions, the default runlevel is stored in the file '/etc/inittab', although you may have to consult the guide for your own distribution. The line that indicates the default runlevel appears as id:n:initdefault: or similar, where "n" indicates the number of the runlevel. '/etc/inittab' must be edited as root. Please read the sections on editing files and root user if you are unfamiliar with this concept. Also, it is recommended that you create a copy of the file prior to editing it, particularly if you are new to Linux text editors, in case you accidentally corrupt the file: # cp /etc/inittab /etc/inittab.original The line should be edited such that an appropriate runlevel is the default (1, 2, or 3 on most systems): id:3:initdefault: After saving the changes, exit X. After the Driver installation is complete, you may revert the default runlevel to its original state, either by editing the '/etc/inittab' again or by moving your backup copy back to its original name. Different distributions provide different ways to exit X. On many systems, the 'init' utility will change the current runlevel. This can be used to change to a runlevel in which X is not running. # init 3 There are other methods by which to exit X. Please consult your distribution. I9. LINUX MANUAL AND INFO PAGES System manual or info pages are usually installed during installation. These pages are typically up-to-date and generally contain a comprehensive listing of the use of programs and utilities on the system. Also, many programs include the --help option, which usually prints a list of common options for that program. To view the manual page for a command, enter % man commandname at the command prompt, where commandname refers to the command in which you are interested. Similarly, entering % info commandname will bring up the info page for the command. Depending on the application, one or the other may be more up-to-date. The interface for the info system is interactive and navigable. If you are unable to locate the man page for the command you are interested in, you may need to add additional elements to your 'MANPATH' environment variable. See the section on environment variables. ______________________________________________________________________________ Appendix J. Application Profiles ______________________________________________________________________________ J1. INTRODUCTION The NVIDIA Linux driver supports configuring various driver settings on a per-process basis through the use of "application profiles": collections of settings that are only applied if the current process matches attributes detected by the driver when it is loaded into the process. This mechanism allows users to selectively override global driver settings for a particular application without the need to set environment variables on the command line prior to running the application. Application profiles consist of "rules" and "profiles". A "profile" defines what settings to use, and a "rule" identifies an application and defines what profile should be used with that application. A rule identifies an application by describing various features of the application; for example, the name of the application binary (e.g. "glxgears") or a shared library loaded into the application (e.g. "libpthread.so.0"). The particular features supported by this NVIDIA Linux implementation are listed below in the "Supported Features" section. Currently, application profiles are only supported by the NVIDIA Linux GLX implementation, but other NVIDIA driver components may use them in the future. Application profiles can be configured using the nvidia-settings control panel. To learn more, consult the online help text by clicking the "Help" button under the "Application Profiles" page in nvidia-settings. J2. ENABLING APPLICATION PROFILES IN THE OPENGL DRIVER Note: if HOME is unset, then any configuration files listed below located under $HOME will not be loaded by the driver. To enable application profile support globally on a system, edit the file $HOME/.nv/nvidia-application-profile-globals-rc to contain a JSON object with a member "enabled" set to true or false. For example, if this file contains the following string: { "enabled" : true } application profiles will be enabled globally in the driver. If this file does not exist or cannot be read by the parser, application profiles will be enabled by default. Application profile support in the driver can be toggled for an individual application by using the __GL_APPLICATION_PROFILE environment variable. Setting this to 1 enables application profile support, and setting this to 0 disables application profile support. If this environment variable is set, this overrides any setting specified in $HOME/.nv/nvidia-application-profile-globals-rc. Additionally, the application profile parser can log debugging information to stderr if the __GL_APPLICATION_PROFILE_LOG environment variable is set to 1. Conversely, setting __GL_APPLICATION_PROFILE_LOG to 0 disables logging of parse information to stderr. J3. APPLICATION PROFILE SEARCH PATH By default, when the driver component ("libGL.so.1" in the case of GLX) is loaded by a process, the driver looks for files in the following search path: o '$HOME/.nv/nvidia-application-profiles-rc' o '$HOME/.nv/nvidia-application-profiles-rc.d' o '/etc/nvidia/nvidia-application-profiles-rc' o '/etc/nvidia/nvidia-application-profiles-rc.d' o '/usr/share/nvidia/nvidia-application-profiles-570.123.06-rc' By convention, the '*-rc.d' files are directories and the '*-rc' files are regular files, but the driver places no restrictions on file type, and any of the above files can be a directory or regular file, or a symbolic link which resolves to a directory or regular file. Files of other types (e.g. character or block devices, sockets, and named pipes) will be ignored. If a file in the search path is a directory, the parser will examine all regular files (or symbolic links which resolve to regular files) in that directory in alphanumeric order, as determined by strcoll(3). Files in the directory of other types (e.g. other directories, character or block devices, sockets, and named pipes) will be ignored. J4. CONFIGURATION FILE SYNTAX When application profiles are enabled in the driver, the driver configuration is defined by a set of PROFILES and RULES. Profiles are collections of driver settings given as key/value pairs, and rules are mappings between one or more PATTERNS which match against some feature of the process and a profile. Configuration files are written in a superset of JSON (http://www.json.org/) with the following additional features: o A hash mark ('#') appearing outside of a JSON string denotes a comment, and any text appearing between the hash mark and the end of the line inclusively is ignored. o Integers can be specified in base 8 or 16, in addition to base 10. Numbers beginning with '0' and followed by a digit are interpreted to be octal, and numbers beginning with '0' and followed by 'x' or 'X' are interpreted to be hexadecimal. Each file consists of a root object with two optional members: o "rules", which contains an array of rules, and o "profiles", which contains an array of profiles. Each rule is an object with the following members: o "pattern", which contains either a string, a pattern object, or an array of zero or more pattern objects. If a string is given, it is interpreted to be a pattern object with the "feature" member set to "procname" and the "matches" member set to the value of the string. During application detection, the driver determines if each pattern in the rule matches the running process, and only applies the rule if all patterns in the rule match. If an empty array is given, the rule is unconditionally applied. o "profile", which contains either a string, array, or profile. If a string is given, it is interpreted to be the name of some profile in the configuration. If a profile is given, it is implicitly defined as part of the rule. If an array is given, the array is interpreted to be an inline profile with its "settings" member set to the contents of the array. Each profile is an object with the following members: o "name", a string which names the profile for use in a rule. This member is mandatory if the profile is specified as part of the root object's profiles array, but optional if the profile is defined inline as part of a rule. o "settings", an array of settings which can be given in two different formats: 1. As an array of keys and values, e.g. [ "key1", "value1", "key2", 3.14159, "key3", 0xF00D ] Keys must be specified as strings, while a value may be a string, number, or true/false. 2. as an array of JSON setting objects. Each setting object contains the following members: o "k" (or "key"), the key given as a string o "v" (or "value"), the value, given as a string, number, or true/false. A pattern object may consist of a pattern primitive, or a logical operation on pattern objects. A pattern primitive is an object containing the following members: o "feature", the feature to match the pattern against. Supported features are listed in the "Supported Features" section below. o "matches", the string to match. A pattern operation is an object containing the following members: o "op", a string describing the logical operation to apply to the subpatterns. Valid values are "and", "or", or "not". o "sub": a pattern object or array of one or more pattern objects, to serve as the operands. Note that the "not" operator expects exactly one object; any other number of objects will cause the pattern to fail to match. If the pattern is an operation, then the pattern matches if and only if the logical operation applied to the subpatterns is true. For example, { "op" : "or", "sub" : [ { "feature" : "procname", "matches" : "foo" }, { "feature" : "procname", "matches" : "bar" } ] } matches all processes with the name "foo" *or* "bar". Similarly, { "op" : "and", "sub" : [ { "feature" : "procname", "matches" : "foo" }, { "feature" : "dso", "matches" : "bar.so" } ] } matches all processes with the name "foo" that load DSO "bar.so", and { "op" : "not", "sub" : { "feature" : "procname", "matches" : "foo" } } matches a process which is *not* named "foo". Nested operations are possible; for example: { "op" : "and", "sub" : [ { "feature" : "dso", "matches" : "foo.so" }, { "op" : "not", "sub" : { "feature" : "procname", "matches" : "bar" } } ] } matches processes that are *not* named "bar" that load DSO "foo.so". J5. EXTENDED BACKUS-NAUR FORM (EBNF) GRAMMAR Note: this definition omits the presence of whitespace or comments, which can be inserted between any pair of symbols. This is written in an "EBNF-like" grammar based on ISO/IEC 14977, using the following (non-EBNF) extensions: o object(A, B, ...) indicates that each symbol A, B, etc. must appear exactly once in any order, delimited by commas and bracketed by curly braces, unless the given symbol expands to an empty string. For example, assuming A and B are nonempty symbols: object(A, B) ; is equivalent to: '{', (A, ',', B) | (B, ',', A), '}' ; Also, object([A], [B]); is equivalent to: '{', [ A | B | (A, ',', B) | (B, ',', A) ], '}' ; o attr(str, A) is shorthand for: ( '"str"' | "'str'" ), ':', A o array(A) is shorthand for: '[', [ array_A ], ']' where array_A is defined as: array_A = (array_A, ',' A) | A The grammar follows. config_file = object( [ attr(rules, array(rule)) ] , [ attr(profiles, array(profile)) ] ) ; rule = object(attr(pattern, pattern_object | array(pattern_object)), attr(profile, profile_ref)) ; pattern_object = pattern_op | pattern_primitive ; pattern_op = object(attr(op, string), attr(sub, pattern_object | array(pattern_object))) ; pattern_primitive = object(attr(feature, string), attr(matches, string)) ; profile_ref = string | settings | rule_profile ; profile = object(attr(name, string), attr(settings, (array(setting_kv) | array(setting_obj)))) ; rule_profile = object([ attr(name, string) ], attr(settings, (array(setting_kv) | array(setting_obj)))) ; setting_kv = string ',' value ; setting_obj = object(attr(k,string) | attr(key,string), attr(v,value) | attr(value,value)) ; string = ? any valid json string ? ; value = ? any valid json number ? | 'true' | 'false' | hex_value | oct_value ; hex_value = '0', ('x' | 'x'), hex_digit, { hex_digit } ; oct_value = '0', oct_digit, { oct_digit } ; hex_digit = ? any character in the range [0-9a-fA-F] ? ; oct_digit = ? any character in the range [0-7] ? ; J6. RULE PRECEDENCE Profiles may be specified in any order, and rules defined in files earlier in the search path may refer to profiles defined later in the search path. Rules are prioritized based on the order in which they are defined: each rule has precedence over rules defined after it in the same file, and rules defined in a file have precedence over rules defined in files that come after that file in the search path. For example, if there are two files A and B, such that A comes before B in the search path, with the following contents: # File A { "rules" : [ { "pattern" : "foo", "profile" : [ "a", 1 ] }, { "pattern" : "foo", "profile" : [ "a", 0, "b", 2 ] } ] } # File B { "rules" : [ { "pattern" : "foo", "profile" : [ "a", 0, "b", 0, "c", 3 ] } ] } and the driver is loaded into a process with the name "foo", it will apply the settings "a" = 1, "b" = 2, and "c" = 3. Settings specified via application profiles have higher precedence than global settings specified in nvidia-settings, but lower precedence than settings specified directly via environment variables. J7. CONFIGURATION FILE EXAMPLE The following is a sample configuration file which demonstrates the various ways one can specify application profiles and rules for different processes. { "rules" : [ # Define a rule with an inline profile, (implicitly) using # feature "procname". { "pattern" : "glxgears", "profile" : [ "GLSyncToVBlank", "1" ] }, # Define a rule with a named profile, (implicitly) using feature # "procname". { "pattern" : "gloss", "profile" : "p0" }, # Define a rule with a named profile, using feature "dso". { "pattern" : { "feature" : "dso", "matches" : "libpthread.so.0" }, "profile" : "p1" }, # Define a rule with a named, inline profile, using feature "true"; # patterns using this feature will always match, and can be used # to write catch-all rules. { "pattern" : { "feature" : "true", "matches" : "" }, "profile" : { "name" : "p2", "settings" : [ "GLSyncToVBlank", 1 ] } }, # Define a rule with multiple patterns. This rule will only be # applied if the current process is named "foo" and has loaded # "bar.so". { "pattern" : [ { "feature" : "procname", "matches" : "foo" }, { "feature" : "dso", "matches" : "bar.so" } ], "profile" : "p1" }, # Define a rule with no patterns. This rule will always be applied. { "pattern" : [], "profile" : "p1" } ], "profiles" : [ # define a profile with settings defined in a key/value array { "name" : "p0", "settings" : [ "GLSyncToVBlank", 0 ] }, # define a profile with settings defined in an array of setting # objects { "name" : "p1", "settings" : [ { "k" : "GLDoom3", "v" : false } ] } ] } J8. SUPPORTED FEATURES This NVIDIA Linux driver supports detection of the following features: o "true": patterns using this feature will always match, regardless of the contents of the string provided by "matches". o "procname": patterns using this feature compare the string provided by "matches" against the pathname of the current process with the leading directory components removed and match if they are equal. o "commname": patterns using this feature compare the string provided by "matches" against the commandname of the current process, as stored in /proc/self/comm. o "cmdline": patterns using this feature compare the string provided by "matches" against the argument 0 of the command line for the current process, as stored in /proc/self/cmdline (with leading directory components removed, separated by slashes and backslashes). This is useful for applications and games running through Wine. For example "glxgears" would match a process started with "/usr/bin/glxgears" or "notepad.exe" would match a process started with "wine C:\Windows\notepad.exe". o "dso": patterns using this feature compare the string provided by "matches" against the list of currently loaded libraries in the current process and match if the string matches one of the entries in the list (with leading directory components removed). Please note that application detection occurs when the driver component ("libGL.so.1" in the case of GLX) is first loaded by a process, so a pattern using this feature may fail to match if the library specified by the pattern is loaded after the component is loaded. A potential workaround for this on Linux is to set the LD_PRELOAD environment variable (see ld-linux(8)) to include the component, as in the following example: LD_PRELOAD="libGL.so.1" glxgears Note this defeats one of the objectives of application detection (namely the need to set environment variables on the command line before running the application), but this may be useful when there is a need to frequently change driver settings for a particular application: one can write a wrapper script to set LD_PRELOAD once, then modify the JSON configuration repeatedly without needing to edit the wrapper script later on. Also note that the pattern matches against library names as they appear in the maps file of that process (see proc(5)), and not the names of symbolic links to these libraries. o "findfile": patterns using this feature should provide a colon-separated list of filenames in the "matches" argument. At runtime, the driver scans the directory of the process executable and matches the pattern if every file specified in this list is present in the same directory. Please note there is currently no support for matching against files in other paths than the process executable directory. J9. LIST OF SUPPORTED APPLICATION PROFILE SETTINGS The list of supported application profile settings and their equivalent environment variable names (if any) is as follows: o "GLFSAAMode" : see __GL_FSAA_MODE o "GLLogMaxAniso" : see __GL_LOG_MAX_ANISO o "GLNoDsoFinalizer" : see __GL_NO_DSO_FINALIZER o "GLSingleThreaded" : see __GL_SINGLE_THREADED o "GLSyncDisplayDevice" : see __GL_SYNC_DISPLAY_DEVICE o "GLSyncToVblank" : see __GL_SYNC_TO_VBLANK o "GLSortFbconfigs" : see __GL_SORT_FBCONFIGS o "GLAllowUnofficialProtocol" : see __GL_ALLOW_UNOFFICIAL_PROTOCOL o "GLSELinuxBooleans" : see __GL_SELINUX_BOOLEANS o "GLShaderDiskCache" : see __GL_SHADER_DISK_CACHE o "GLShaderDiskCachePath" : see __GL_SHADER_DISK_CACHE_PATH o "GLYield" : see __GL_YIELD o "GLThreadedOptimizations" : see __GL_THREADED_OPTIMIZATIONS o "GLDoom3" : see __GL_DOOM3 o "GLExtensionStringVersion" : see __GL_ExtensionStringVersion o "GLConformantBlitFramebufferScissor" : see __GL_ConformantBlitFramebufferScissor o "GLAllowFXAAUsage" : see __GL_ALLOW_FXAA_USAGE o "GLVRRAllowed" : see __GL_VRR_ALLOWED o "GLWriteTextSection" : see __GL_WRITE_TEXT_SECTION o "GLIgnoreGLSLExtReqs" : see __GL_IGNORE_GLSL_EXT_REQS o "EGLVisibleDGPUDevices" On a multi-device system, this option can be used to make EGL ignore certain devices. This setting takes a 32-bit mask encoding a whitelist of visible devices. The bit position matches the minor number of the device to make visible. For instance, the profile { "rules" : [ { "pattern" : [], "profile" : [ "EGLVisibleDGPUDevices", 0x00000002 ] } ] } would only make the device with minor number 1 visible (i.e. /dev/nvidia1). o "EGLVisibleTegraDevices" : Same semantics as EGLVisibleDGPUDevices o "GLShowGraphicsOSD" : see __GL_SHOW_GRAPHICS_OSD o "GLSharpenEnable" : see __GL_SHARPEN_ENABLE o "GLSharpenValue" : see __GL_SHARPEN_VALUE o "GLSharpenIgnoreFilmGrain" : see __GL_SHARPEN_IGNORE_FILM_GRAIN ______________________________________________________________________________ Appendix K. GPU Names ______________________________________________________________________________ Many X configuration options that take a display device name can also be qualified with a GPU. This is done by prepending one of the GPU's names to the display device name. For example, the "MetaModes" X configuration option can be used to enable display devices from multiple GPUs in SLI Mosaic or Base Mosaic configurations: Option "MetaModes" "1280x1024 +0+0, 1280x1024 +1280+0" You can use a display device name qualifier along with a GPU qualifier to configure which display should be picked and from which GPU. E.g., Option "MetaModes" "GPU-1.DFP-0:1280x1024 +0+0, GPU-0.DFP-0:1280x1024 +1280+0" Other X configuration options that support GPU names include: o ColorRange o ColorSpace o ConnectedMonitor o CustomEDID o FlatPanelProperties o IgnoreEDIDChecksum o ModeValidation o nvidiaXineramaInfoOrder o UseDisplayDevice o UseEdidFreqs The description of each X configuration option in Appendix B provides more detail on the available syntax for each option. To find all available names for your configuration, start the X server with verbose logging enabled (e.g., `startx -- -logverbose 5`, or enable the "ModeDebug" X configuration option with `nvidia-xconfig --mode-debug` and restart the X server). The X log (normally /var/log/Xorg.0.log) will contain information regarding each GPU. E.g., (II) NVIDIA(0): NVIDIA GPU NVS 510 (GK107) at PCI:15:0:0 (GPU-0) (II) NVIDIA(0): VERBOSE: GPU UUID: GPU-758a4cf7-0761-62c7-9bf7-c7d950b817c6 Alternatively, nvidia-xconfig can be used to query the GPU names: `nvidia-xconfig --query-gpu-info` GPU #0: Name : NVS 510 UUID : GPU-758a4cf7-0761-62c7-9bf7-c7d950b817c6 PCI BusID : PCI:3:0:0 Each name has different properties that may affect which name is appropriate to use. The possible names are: o An NV-CONTROL target ID-based name (e.g., "GPU-0"). The NVIDIA X driver will assign a unique ID to each GPU on the entire X server. These IDs are not guaranteed to be persistent from one run of the X server to the next, so is likely not convenient for X configuration file use. It is more frequently used in communication with NV-CONTROL clients such as nvidia-settings. o An UUID-based name (e.g., "GPU-758a4cf7-0761-62c7-9bf7-c7d950b817c6"). This name is a SHA-1 hash, formatted in canonical UUID 8-4-4-4-12 format. This UUID is unique for each physical GPU, and will be the same regardless of where the GPU is connected. ______________________________________________________________________________ Appendix L. Wayland Known Issues ______________________________________________________________________________ L1. OVERVIEW There are several areas in which the NVIDIA driver lacks feature parity between X11 and Wayland. These may be due to limitations of the driver itself, the Wayland protocol, or the specific Wayland compositor in use. Over time this list is expected to shorten as missing functionality is implemented both in the driver and in upstream components, but the following captures the situation as of the release of this driver version. Note that this assumes a compositor with reasonably complete support for graphics-related Wayland protocol extensions. L2. NVIDIA DRIVER LIMITATIONS o Wayland compositors may not be designed to handle the case where two DRM drivers instantiate independent device nodes capable of driving the same display. When the nvidia-drm driver does not remove conflicting framebuffers and instantiate its own DRM fbdev support, simpledrm will remain present for driving the initial framebuffer console if initially loaded. In this situation, both simpledrm and nvidia-drm have DRI device nodes driving the same display. This can lead to issues such as a black screen, corruption, or flickering. A workaround for this is to set the fbdev=1 module parameter for the nvidia-drm kernel module. o The NvFBC desktop capture library does not have native Wayland support and does not work with Xwayland. o The nvidia-drm module does support the DEGAMMA_LUT, CTM, and GAMMA_LUT CRTC properties. However the COLOR_ENCODING and COLOR_RANGE plane properties are unsupported. This may impact some advanced compositor features related to color correction. o The nvidia-settings configuration tool has limited functionality on Wayland. o Front-buffer rendering in GLX does not work with Xwayland. o Explicit Sync refers to using DRM Syncobj timeline synchronization points to explicitly signal ownership transfers of shared buffers to and from the compositor. Explicit Sync support requires a kernel that contains the following two commits, which fix bugs preventing Explicit Sync from working correctly. These are present in 6.8.x or newer kernels. o "drm/syncobj: fix DRM_SYNCOBJ_WAIT_FLAGS_WAIT_AVAILABLE" with commit hash 101c9f637efa1655f55876644d4439e552267527. o "drm/syncobj: handle NULL fence in syncobj_eventfd_entry_func" with commit hash 2aa6f5b0fd052e363bb9d4b547189f0bf6b3d6d3. L3. WAYLAND PROTOCOL OR COMPOSITOR LIMITATIONS o The following workstation features are not supported by any Wayland compositors or the Wayland protocol. They will also likely require new EGL extensions or other means to expose the related hardware functionality. Features followed by an asterisk (*) are supported with VK_KHR_display. o SLI Mosaic (Chapter 31) o Frame Lock and Genlock (Chapter 32)* o Swap Groups* o Advanced display pipeline features including warp and blend, pixel shift, and emulated YUV420. o Stereo rendering* o There is no established public API through which Wayland compositors can power off video memory via RTD3 (Chapter 22). o Display multiplexers (muxes) are typically used in laptops with both integrated and discrete GPUs to provide a direct connection between the discrete GPU and the built-in display (internal mux) or an external display (external mux). On X11, the display mux can be automatically switched when a full-screen application is running on the discrete GPU, enabling enhanced display features and improved performance, but no Wayland compositors currently support this functionality. o Indirect GLX is not supported by Xwayland. o Hardware overlays cannot be used by GLX applications with Xwayland.