Nvidia Linux Display Driver 32-bit

Nvidia Linux Display Driver is a proprietary OpenGL video driver that tries to bring support for the graphics cards produced by Nvidia on Linux-based operating systems. This version supports only the 32-bit architecture. Even if it provides a settings panel, the Nvidia Linux Display Driver is not an application. It is a hardware driver, which provides some instructions that tell an operating system how to use a piece of hardware.

Installing Nvidia Linux Display Driver

In order to install the Nvidia Linux Display Driver on your Linux machine, one needs to download the binary package using the “Download” button above, save the file on their Home folder and execute the sh ./NVIDIA-Linux-x86-xxx.xx.run command, as root, in a Linux Terminal (where xxx.xx is the version number).

During the installation, users will be asked if they want to edit the X configuration file manually or let the installer do all the work. Alternatively, after installation, you can run the nvidia-xconfig command via a Linux Terminal to set the new driver as the default one.

Drop-in replacement for Nouveau

These days, many Linux flavors are bundled with an open source driver for Nvidia graphics cards, called Nouveau, created by a group of developers who work for the freedesktop.org project.

The good thing about the Nouveau driver is that it works pretty well by default and it is integrated into the upstream Linux kernel, so you don’t have to download or install anything on your Linux operating system.

Supports a wide range of Nvidia GPUs

In our tests, we’ve discovered that the Nvidia Linux Display Driver supports much more Nvidia graphics cards than the open source Nouveau driver. It plays well with most of the ION, ION LE, Quadro Plex, Quadro Sync, GRID, NVS, nForce Professional 3000, Quadro FX, Quadro Blade, as well as all the GeForce series.

All in all, if you have a PC or laptop equipped with an Nvidia GPU, and you have issues with the default Nouveau open source driver, you may want to install the Nvidia Linux Display Driver. It’s also the number one choice for gaming users.

What is new in this release:

• Improved compatibility with recent Linux kernels.
• Fixed an intermittent hang of Vulkan applications running fullscreen when flipping is allowed.
• Removed informational messages that were printed by nvidia-modeset.ko whenever a GPU device was allocated or freed.
• Fixed a bug that caused kwin OpenGL compositing to crash when launching certain OpenGL applications.

What is new in version 390.67:

• Fixed a bug that could cause kernel panics when using Quadro SDI Capture hardware.
• Fixed a bug that caused kwin OpenGL compositing to crash when launching certain OpenGL applications.
• Fixed an intermittent crash when launching Vulkan applications.
• Fixed an intermittent crash when launching applications through Wine.
• Fixed a bug that caused the driver, in some low bandwidth DisplayPort configurations, to not implicitly enable display dithering. This resulted in visible banding.
• Added support for X.Org xserver ABI 24 (xorg-server 1.20).
• Improved nvidia-bug-report.sh to check for kern.log which is the default kernel log-file location for many Debian-based Linux distributions.
• Fixed a bug which could cause X servers that export a Video Driver ABI earlier than 0.8 to crash when running X11 applications which call XRenderAddTraps().

What is new in version :

• Added support for the following GPU:
• TITAN V
• GTX 1070 Ti
• Fixed a bug that caused Vulkan X11 swapchains to fail on GPUs without a display engine, such as some Tesla-branded graphics cards and some Optimus laptops.
• Fixed a bug that caused fullscreen Vulkan applications to hang on some
• Kepler GPUs, such as the GeForce GTX 680.
• Fixed a bug where the G-SYNC indicator was reporting "normal" instead
• of "G-SYNC" on Vulkan applications when G-SYNC was enabled.

What is new in version 387.22:

• Added support for the following GPU:
• GTX 1070Ti
• Fixed a regression that could cause driver errors when setting modes that include DisplayPort Multi-Stream Transport devices.
• Added an nvidia.ko kernel module parameter, NVreg_EnableBacklightHandler, which can be used to enable experimental handling of laptop backlight brightness through /sys/class/backlight/. This handler overrides the ACPI-based one provided by the video.ko kernel module.
• NVreg_EnableBacklightHandler is disabled by default.
• Added G-SYNC to all supported Vulkan swapchains for Maxwell and up. G-SYNC is enabled by default when using G-SYNC-ready monitors. For direct-to-display swapchains, an application profile with "GLGSYNCAllowed" setting set to 'false' can be used to disable this feature:
• { "rules" : [
• { "pattern" : [], "profile" : [ "GLGSYNCAllowed", false ] }
• Fixed a regression that caused some display connectors on some GPUs to not report a connected HDMI or DisplayPort audio device even if the connected monitor supports audio.
• Fixed a bug that sometimes prevented the "Reset Default Configuration" button in the nvidia-settings "ECC Settings" page from being available when the ECC configuration is set to a non-default state.
• Fixed a bug that caused nvidia-settings to enforce overly aggressive limits on display positions in the "X Server Display Configuration" page under some circumstances.
• Fixed a bug that could cause the "Enable Base Mosaic (Surround)" checkbox in nvidia-settings to disappear when an X screen, rather than a display, is selected in the "X Server Display Configuration" page.
• Fixed a race condition that could lead to crashes when OpenGL programs manipulated vertex buffer objects from multiple threads simultaneously.
• Fixed a bug that caused the nvidia-settings control panel to retain some settings that had been applied, but not confirmed. This resulted in unwanted settings being applied to subsequent settings changes.
• Improved performance of fullscreen Vulkan applications using X11 swapchains. This optimization will cause more events that trigger an out-of-date swapchain, such as when 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. See section 30.8 of the Vulkan specification.
• Added support for YUV 4:2:0 compression for monitors connected via DisplayPort in configurations where either the display or GPU is incapable of driving the current mode in RGB 4:4:4. See the description in the "Programming Modes" appendix for details.
• Added framebuffer console hot plug handling to nvidia-modeset.
• Note that hot plugging is only handled when nvidia-modeset is initialized; for example, when Xorg or nvidia-persistenced is running or when nvidia-drm is loaded with the "modeset=1" parameter.
• Added an "AllowGSYNC" MetaMode attribute that can be used to disable G-SYNC completely. This can be use to allow enabling features that are incompatible with G-SYNC, such as Ultra Low Motion Blur or Frame Lock.
• Fixed several problems that prevented the "cc_version_check" sanity test from running correctly when building the NVIDIA kernel modules. As these problems would have masked mismatches between the compiler versions used to build the kernel and the NVIDIA kernel modules for an extended period of time, nvidia-installer has been updated to ignore CC version mismatches by default when they are detected.
• Updated nvidia-installer to use the init_module syscall directly to test load kernel modules instead of libkmod/insmod. This removes libkmod as a build dependency of nvidia-installer, and insmod as a run-time dependency.
• Tiled monitors formerly resulted in a separate Xinerama screen being reported for each tile. They will now, by default, be combined into a single large Xinerama screen.
• The individual panels in a tiled monitor will now be arranged based on the layout information provided in the monitor's EDID. This can be overridden by either manually specifying offsets or using the "MetaModeOrientation" option.
• Disabled interlaced modes over DisplayPort by default due to incomplete support in the GPU. Added "AllowDpInterlaced" mode validation token to override this default behavior and allow interlaced modes over DisplayPort protocol anyway.
• Fixed a bug that could cause some eDP G-SYNC displays to flicker at low refresh rates.
• Fixed a bug that could cause OpenGL applications to crash after a prolonged DPMS sleep state on a monitor driven with PRIME Sync.
• Fixed a bug that artificially limited the maximum pixel clock to 300 MHz when using certain more capable DisplayPort to HDMI adapters.
• Fixed a bug that prevented the NVIDIA kernel modules from building for non-SMP Linux kernels.
• Updated the output of `nvidia-smi nvlink --status` to include reporting NVLink speed.
• Fixed a bug that caused incorrect PCI topology reporting in nvidia-smi on Intel Skylake systems.

What is new in version 384.59:

• Added support for the following GPUs:
• GeForce GT 1030
• GeForce MX150
• Fixed a bug that prevented changes to stereo eye assignment from getting applied from the nvidia-settings control panel.
• Fixed a bug that caused S4 suspend (also known as "hibernate") to hang during suspend when SLI was enabled.
• Fixed a bug where devices would not be properly enumerated by the Vulkan driver. See the "Known Issues" section of the README for more details.
• Fixed a bug that could cause OpenGL applications to crash after calling glDeleteBuffers on VBOs that have been remapped with ARB_vertex_attrib_binding.
• Fixed a bug that could sometimes cause OpenGL applications to lock up until the X server receives input.
• Fixed a bug that caused nvidia-settings to drop device BusID values when making changes to an existing X configuration file.
• Fixed a regression that prevented nvidia-xconfig from querying some GPUs, e.g. when running `nvidia-xconfig -a`.
• Fixed a bug that caused VDPAU playback to be corrupted when extended to coordinates beyond 16384 pixels.
• Updated nvidia-persistenced to initialize nvidia-modeset.ko in addition to nvidia.ko when setting persistence mode.
• Restored several sanity checks that were inadvertently removed from the kernel module build process in the 355.06 driver.
• Added support for a "Nearest" transform filter in the X driver. This causes the X driver to use nearest neighbor filtering when performing screen transformations. The filter can be requested through the 'filter' argument to RandR's RRSetCrtcTransform, or through the 'ResamplingMethod' MetaMode token.
• Fixed a bug where nvidia-installer would install a wrong version of the Vulkan ICD configuration file when GLVND is disabled for the GLX client.
• Fix a bug with GLX_EXT_buffer_age where incorrect buffer age values would be reported for SLI AFR configurations. In such configurations buffer age may now be greater than 3, the previous maximum buffer age.
• Fixed a bug that could cause hanging and Xids when performing RandR transforms with Overlay and SLI enabled.
• Extended the information reported by the NVIDIA Xinerama X extension to report PRIME displays in addition to directly-connected displays.
• Fixed a bug that caused HDMI audio devices to appear or disappear inconsistently when HDMI devices were hotplugged or unplugged.
• Fixed a bug that could cause driver errors when setting modes on X screens running at Depth 8 or Depth 15.
• Fixed a bug that could cause intermittent kernel panics when running with PRIME Sync.
• Fixed a bug that caused a kernel panic when hotplugging HDMI displays on some Zotac mini PCs.
• Updated nvidia-installer to label kernel modules with SELinux file type 'modules_object_t'. Some system SELinux policies only permit loading of kernel modules with this SELinux file type.
• Removed support for checking for and downloading updated driver packages and precompiled kernel interfaces from nvidia-installer. This functionality was limited to unencrypted ftp and http, and was implemented using code that is no longer actively maintained.

What is new in version 381.22:

• Fixed a bug that caused the GLX_EXT_buffer_age buffer age extension to return incorrect values for stereo drawables when using passive stereo modes 5 through 9.
• Fixed a bug in Vulkan direct to display where DP1.2 monitors were not being enumerated.
• Improved compatibility with recent kernels.
• Fixed a bug that caused "nvidia-settings --query all" to print many duplicate entries.
• Fixed a bug that caused applications to crash in some situations when calling glXMakeCurrent while OpenGL threaded optimizations were enabled. This frequently occurred when Steam was attempting to make a video appear full-screen.
• Fixed a bug that caused VDPAU applications to use the blit presentation queue when a previous VDPAU application didn't shut down cleanly.
• Fixed hangs and crashes that could occur when an OpenGL context is created while the system is out of available memory.
• Fixed a bug that caused corruption when OpenGL windows were moved or resized.
• Fixed a bug that caused X screens that use Option "UseDisplayDevice" "none" to be resized to 640x480 when using "xrandr -s" to change the screen configuration.
• Fixed a kernel crash that occurred when attempting to map large user memory allocations into CUDA.
• Disabled OpenGL threaded optimizations by default, initially enabled in 378.09, due to various reports of instability.
• Added support for the following Vulkan extensions:
• VK_EXT_acquire_xlib_display
• VK_EXT_display_control
• VK_EXT_display_surface_counter
• VK_EXT_direct_mode_display
• VK_KHX_external_memory
• VK_KHX_external_memory_fd
• VK_KHX_external_semaphore
• VK_KHX_external_semaphore_fd
• These extensions require a Vulkan loader version >= 1.0.42.
• Removed the X driver's logo splash screen and the corresponding NoLogoand LogoPath xorg.conf options.
• Added the "ResamplingMethod" MetaMode option, adding support for bicubic resampling methods when scaling screen transformations are in use. See the README for more details.

What is new in version 378.13:

• Added support for the following GPU(s):
• Quadro P3000
• Quadro GP100
• Quadro P4000
• Quadro P2000
• Quadro P1000
• Quadro P600
• Quadro P400
• Quadro M1200
• Quadro M2200
• Disabled OpenGL threaded optimizations by default under Xinerama.
• Added support for the ARB_parallel_shader_compile extension to allow multi-threaded compilation of GLSL shaders.
• Updated the X driver to ignore any Virtual Reality Head Mounted Displays (HMDs). See the "AllowHMD" X configuration option in the README for details.
• The driver will now advertise GLX FBConfigs with no depth bits on depth 30 X screens.
• Added support in nvidia-settings to view configured PRIME displays. To enable PRIME displays, see "Offloading Graphics Display with RandR 1.4" in the README.
• Added infrastructure which enables the NVIDIA EGL driver to load EGL external platform libraries that add client-side support for new window systems, beyond the existing libnvidia-egl-wayland.so.1. For more details, see:
• - https://github.com/NVIDIA/eglexternalplatform
• - https://github.com/NVIDIA/egl-wayland
• Added support for the following Vulkan extensions:
• VK_KHR_display
• VK_KHR_display_swapchain
• Enabled OpenGL threaded optimizations by default in the driver. Refer to the "Threaded Optimizations" section in the "Specifying OpenGL Environment Variable Settings" chapter of the README for details. These optimizations will self-disable when they are degrading performance. As a result, performance should be unchanged for many applications, and increased for those that benefit from threaded optimizations and were not already forcing them enabled.
• Fixed a bug that prevented PRIME Sync from working on notebooks with GeForce GTX 4xx and 5xx series GPUs.
• Fixed a bug that caused system hangs when resuming from suspend with some GPUs.
• Fixed a regression that could cause corruption when hot-plugging displays.
• Fixed a regression that prevented systems with multiple DisplayPort monitors from resuming correctly from suspend.
• Added support for the screen_info.ext_lfb_base field, on kernels that have it, in order to properly handle UEFI framebuffer consoles with physical addresses above 4GB.
• Added support for X.Org xserver ABI 23 (xorg-server 1.19)
• Fixed a bug that allowed nvidia-installer to attempt loading kernel modules that were built against non-running kernels.

What is new in version 375.26:

• Added support for the following GPU(s):
• Quadro M3000 SE
• Fixed a regression that prevented the backlight on some notebook computer displays from turning back on after being shut off.
• Fixed a regression that prevented the console from being restored properly when using SLI.
• Fixed a bug that caused installer manifest entries for some 32-bit compatibility libraries to be duplicated in the installer package. These entries led nvidia-installer to issue warning messages when installing the 32-bit compatibility libraries, for example: 'Cannot create symlink /usr/lib/libEGL.so (File exists)'.
• Fixed a regression that could cause the nvidia-settings control panel to crash on startup with certain GPU configurations.
• Fixed a regression that prevented the GLX_NV_swap_group extension from working when using SLI or Quadro Sync.
• Fixed a long-standing bug that caused the NVIDIA X driver to not correctly detect AC/battery hotplug events on notebooks.
• Fixed an interaction problem between the Linux kernel's CONFIG_HARDENED_USERCOPY feature (new in Linux kernel 4.8) and nvidia-drm.ko with modeset=1 enabled. This problem caused CONFIG_HARDENED_USERCOPY to erroneously halt the system, issuing reports of the form:
• usercopy: kernel memory overwrite attempt detected to ffff8803e8ec7ce0 () (8 bytes)
• Fixed a bug that prevented nvidia-bug-report.sh from finding relevant messages in kernel log files.
• Added support for the screen_info.ext_lfb_base field, on kernels that have it, in order to properly handle UEFI framebuffer consoles with physical addresses above 4GB.

What is new in version 375.20:

• Added support for the following GPUs:
• GeForce GTX 1050
• GeForce GTX 1050 Ti
• Quadro M620
• Quadro M5000 SE
• Quadro M3000 SE
• Fixed a bug that could cause displays in SLI Mosaic to blank when applying a transformation matrix.
• Added EGL support to the GL Vendor Neutral Dispatch Library (libGLVND). The installer package now includes both GLVND and non-GLVND versions of the EGL libraries: the --glvnd-egl-client and --no-glvnd-egl-client options in nvidia-installer can be used to select which to install.
• Fixed a bug that prevented G-SYNC from working on notebook displays.
• Fixed a bug that caused DisplayPort multistream devices to go blank after the console is restored.
• Fixed a bug that caused a kernel crash when starting X for the second time when more than one DisplayPort multistream device is connected to a single DisplayPort connector via a branch device.
• Updated nvidia-settings to not have a build-time dependence on an external nvml development package (a regression introduced in 375.10).
• Fixed a crash in nvidia-settings when adding Application Profile Rule and Profile entries.
• Removed the "Enable Tooltip" option in nvidia-settings for the GTK 2 interface.
• Added new X configuration options:
• ForceCompositionPipeline
• ForceFullCompositionPipeline
• which override the MetaMode tokens with the same names.
• Fixed a bug that caused issues with panning and cursor constraining when mixing PRIME-driven displays with natively driven displays.
• Fixed a bug that caused long delays when leaving the VT or disabling a display device while an OpenGL application is running.
• Improved console restore behavior on systems that use the UEFI Graphics Output Protocol, and most vesafb modes.
• Added support for the RandR TILE property added in RandR 1.5.
• Raised the on-disk OpenGL shader cache size from 64MB to 128MB. Refer to the "Specifying OpenGL Environment Variable Settings" section of the driver README for details on how to select its location or disable it.
• Fixed a bug in synchronization primitives shared between X and OpenGL.
• Added support for X.Org xserver ABI 23 (xorg-server 1.19)
• Fixed a bug that allowed nvidia-installer to attempt loading kernel modules that were built against non-running kernels.

What is new in version 370.28:

• Fixed undefined symbol references in libnvidia-glcore.so.
• Fixed a potential hang in Vulkan applications such as vkQuake and DOTA2 that use vkFences for present synchronization.
• Fixed a bug that caused the driver to not set HDMI YUV 420 4K modes correctly on some TVs, such as the VIZIO D40u-D1.
• Updated CurrentMetaMode reporting via the NV-CONTROL API to reflect whether ForceCompositionPipeline or ForceFullCompositionPipeline are set in the current MetaMode.
• Added the ability to over- and under-clock certain GeForce GPUs in the GeForce GTX 1000 series and later. For GPUs that allow it, an offset can be applied to clock values in some clock domains of all performance levels. This clock manipulation is done at the user's own risk. See the README documentation of the "CoolBits" X configuration option for more details.
• Implemented initial support for PRIME Synchronization to synchronize flipping between an NVIDIA PRIME display source and a separate PRIME display sink. This support is experimental and currently requires a prerelease X.org X server.
• Fixed a bug that prevented Vulkan applications from presenting from multiple queues to the same X11 swapchain.
• Added the "PixelShiftMode" MetaMode option, enabling support for 4K and 8K pixel shift displays. See the README for details.

What is new in version 367.44:

• Added support for the following GPUs:
• TITAN X (Pascal)
• GeForce GTX 1060 6GB
• GeForce GTX 1060 3GB
• Fixed a regression that caused applications using indirect GLX to crash.
• Fixed a regression introduced in 367.35 that caused the first modeset of the X server to display blank if the features requested in the X configuration file enabled the X driver's composition pipeline. This would be triggered, e.g., by MetaMode tokens such as ForceCompositionPipeline, ForceFullCompositionPipeline, Rotation, Reflection, and Transform.

What is new in version 367.35:

• Fixed a regression that could cause console corruption when resuming from suspend.
• Improved buffer write performance of the nvidia-drm DRM KMS driver by using write-combined DRM Dumb Buffers where available.
• Fixed a bug that caused X to crash when applying changes to the RandR CscMatrix property while VT-switched away from X.
• Fixed a bug that caused a crash when exiting nvidia-settings on displays with 8 or 15 bit color depths.
• Added support for VDPAU Feature Set H to the NVIDIA VDPAU driver. GPUs with VDPAU Feature Set H are capable of hardware-accelerated decoding of 8192x8192 (8k) H.265/HEVC video streams.
• Fixed a bug that caused the X server to sometimes skip displaying Vulkan frames when the Composite extension is enabled.
• Added support for querying clock values on Pascal GPUs.
• Removed the Base Mosaic configuration option from nvidia-settings on systems where the feature is not actually supported.
• Fixed a bug that caused nvidia-smi to report an inaccurate version number.
• Fixed a bug that could lead to a system crash if there was a peer-to-peer mapping still active during CUDA context teardown.
• Fixed a bug that prevented nvidia-bug-report.sh from finding relevant messages in kernel log files.

What is new in version 364.19:

• Added support for the following GPUs:
• Quadro M6000 24GB
• Quadro M5500
• Fixed some locking issues in libnvidia-egl-wayland.so.
• Fixed several libnvidia-egl-wayland.so interaction problems with buffer management Wayland protocols.
• Fixed several EGL driver interaction problems with resizing windows in Wayland.
• Fixed a bug that caused some DRM PRIME configurations to fail when DRM KMS was enabled (i.e., nvidia-drm.ko's modeset kernel module parameter was set to "1").
• Fixed a bug that could cause a kernel crash when VDPAU was used while DRM KMS was enabled.
• Changed the default OpenGL installation type to use the GLVND GLX client libraries rather than the legacy non-GLVND libraries.
• Added initial support for Direct Rendering Manager Kernel Modesetting (DRM KMS).See the DRM KMS section of the README for details.
• Added a new kernel module, nvidia-drm.ko, which registers as a DRM driver with both PRIME and DRM KMS support.
• Added support for the following EGL extensions:
• EGL_EXT_platform_wayland
• to enable Wayland applications to run on NVIDIA's EGL implementation,
• EGL_WL_bind_wayland_display
• to enable Wayland compositors to run on NVIDIA's EGL implementation, and
• EGL_EXT_device_drm
• EGL_EXT_output_drm
• EGL_EXT_stream_consumer_egloutput
• to enable Mir and Wayland compositors to display their content through EGLDevice, EGLOutput, and EGLstreams.
• Added a Wayland platform library, libnvidia-egl-wayland.so, to allow Wayland compositors that support EGLDevice, EGLOutput, and EGLstreams to share EGL buffers with Wayland applications.
• Fixed a bug that could cause incorrect frame rate reporting on Quadro Sync configurations with multiple GPUs.
• Added support for the Vulkan API version 1.0.
• Improved X colormap precision from 8 significant bits to 11 on GeForce GPUs.Quadro GPUs already used 11 bits of precision.
• Added a new RandR property, CscMatrix, which specifies a 3x4 color-space conversion matrix.The matrix is applied after the X colormap and before the gamma ramp.This property is available on GF119 and newer GPUs.
• Improved handling of the X gamma ramp on GF119 and newer GPUs.On these GPUs, the RandR gamma ramp is always 1024 entries and now applies to the cursor and VDPAU or workstation overlays in addition to the X root window.
• Reworked how the NVIDIA driver registers with the Linux kernel's DRM subsystem for PRIME support.As a result, PRIME support requires Linux kernel version 3.13 or newer (previously, PRIME support required Linux kernel version 3.10 or newer).
• Improved the interactivity of applications that use a hardware cursor while G-SYNC is active.

What is new in version 361.42:

• Added support for the following GPUs:
• GeForce 920MX
• GeForce 930MX
• Quadro M6000 24GB
• Limited the default concurrency level in nvidia-settings to 32, to avoid hitting the maximum tasks limit on systems with many CPUs.
• Fixed a regression that caused GPU exceptions and incorrect rendering when using pbuffers with EGLDevice.
• Fixed a bug that caused nvidia-settings to crash when pairing glasses with the 3D Vision Pro transmitter on some systems.
• Fixed a bug that caused installer manifest entries for some 32-bit compatibility libraries to be duplicated in the installer package.
• Fixed a regression that caused the HSync and VSync mode timing polarity flags to be inverted.

What is new in version 361.28:

• Added support for the following GPU: GeForce 945A
• Added a legacy, non-GLVND libGL.so GLX client library to the NVIDIA Linux driver installer package, and the ability to select between a GLVND or non-GLVND GLX client library at installation time. This allows users to install the legacy non-GLVND GLX client library in order to work around compatibility issues which may arise due to GLX applications which depend upon behaviors of the NVIDIA GLX client driver which are not defined by the Linux OpenGL ABI version 1.0.
• By default, nvidia-installer will install the legacy, non-GLVND GLX client libraries. The --glvnd-glx-client command line option can be used to override the default, and install the GLVND GLX client libraries instead. Please contact the vendors of any applications that are not compatible with GLVND to ensure that their applications be updated for compatibility with GLVND.
• The presence of multiple GLX client libraries in the package has implications for repackagers of the NVIDIA driver; see the libGL.so entry in the "Installed Components" chapter of the README for details.
• The OpenGL Vendor-Neutral Driver (GLVND) infrastructure is now included and supported by the NVIDIA GLX and OpenGL drivers. This should not cause any visible changes in behavior for end users, but some internal driver component libraries have been renamed and/or moved as a result. These changes may affect scripts that rely on the presence of NVIDIA OpenGL driver components other than those specified in the Linux OpenGL ABI version 1.0, maintainers of alternative NVIDIA driver installation packages, and applications which rely on the presence of any non- OpenGL/GLX symbols in the libGL.so.1 library and its dependencies in any way. Please see https://github.com/NVIDIA/libglvnd. For more information on the GLVND project. The Linux OpenGL ABI version 1.0 specification is available at https://www.opengl.org/registry/ABI
• Updated nvidia-installer to run ldconfig(8) and depmod(8) after uninstalling an existing driver, instead of only running these two commands when installing a new driver.
• Removed the VDPAU wrapper and trace libraries libvdpau.so.1 and libvdpau_trace.so.1 from the driver package. These libraries may be available via your distribution's package management system, or can be compiled from the sources available at http://people.freedesktop.org/~aplattner/vdpau/. Please see http://freedesktop.org/wiki/Software/VDPAU/ for more information.
• Fixed a bug in the EGL driver where a mutex was unlocked more than once. This triggers undefined behavior, and in particular, if lock elision is enabled in glibc, may result in a segmentation fault.

What is new in version 358.16:

• Fixed a regression, introduced in 358.09, that caused incorrect rendering of OpenGL stereo applications in combination with Quadro Sync on certain GPUs.
• Fixed a regression, introduced in 358.09, that caused OpenGL stereo applications to hang when SLI AFR, SFR, or SLIAA modes were enabled.
• Fixed a regression, introduced in 358.09, that caused the system to hang while launching OpenGL applications when some SLI Mosaic topologies were in use.
• Fixed a bug that could cause some G-SYNC monitors to flicker on Kepler-based GPUs when running at low frame rates.
• Fixed a regression that could cause OpenGL or VDPAU applications to hang in Xinerama layouts which include multiple X screens that are driven by the same GPU.
• Added a fallback to allow GLX to function when UNIX domain socket communication is blocked between NVIDIA client-side GLX and NVIDIA server-side GLX. For example, the default SELinux policy on Fedora 22 prevents the NVIDIA 358.09 GLX driver from working in GDM3:
• https://bugzilla.redhat.com/show_bug.cgi?id=1271401
• Fixed a bug that could cause nvidia-installer to crash when installing from a driver package that has had precompiled kernel interfaces added to it (e.g., by using the "--add-this-kernel" option).
• Fixed a regression from 358.09 that caused G-SYNC monitors configured to use 3D Vision stereo to go blank and report an unsupported signal.
• Fixed a regression from 358.09 that caused OpenGL applications that failed to negotiate permission to perform glXSwapBuffers via flipping to stop updating the screen. This was triggered on Fedora 22 where the SELinux system policy prevents gnome-shell from communicating with the X server.
• Now, OpenGL applications that fail to negotiate flipping will fall back to glXSwapBuffers via blitting.
• See https://bugzilla.redhat.com/show_bug.cgi?id=1271401 for more details.
• Fixed a regression that reduced OpenGL performance on headless X server configurations.
• Fixed a memory leak that occurred after destroying a GLXWindow which still has the current context attached to it.
• Fixed a bug which caused EGL pbuffers to be created with both a front and back buffer, instead of a back buffer only, as is required for EGL.
• Added a new kernel module, nvidia-modeset.ko. This new driver component works in conjunction with the nvidia.ko kernel module to program the display engine of the GPU.
• nvidia-modeset.ko does not provide any new user-visible functionality or interfaces to third party applications. However, in a later release, nvidia-modeset.ko will be used as a basis for the modesetting interface provided by the kernel's direct rendering manager (DRM).
• Reduced flickering and delays when transitioning into or out of G-SYNC mode. As part of this change, monitors that have G-SYNC indicators in their on-screen displays will now always report that they are in G-SYNC mode. The OpenGL G-SYNC visual indicator can be enabled in nvidia-settings to determine when G-SYNC is actually being used.
• GLX protocol for the following OpenGL extension from OpenGL 3.0 has been promoted from unofficial to ARB approved official status:
• GL_EXT_draw_buffers2
• GLX protocol for the following OpenGL 3.0 commands:
• BindBufferRangeNV
• BindBufferBaseNV
• BeginTransformFeedbackNV
• EndTransformFeedbackNV
• GetTransformFeedbackVaryingEXT
• TransformFeedbackVaryingsEXT
• which are part of the following extensions:
• GL_NV_transform_feedback
• GL_EXT_transform_feedback
• has been promoted from unofficial to ARB approved official status.
• With the above changes, GLX protocol for OpenGL 3.0 has been promoted from unofficial to ARB approved official status.
• Added a new system memory allocation mechanism for large allocations in the OpenGL driver. This mechanism allows unmapping the allocation from the process when it is not in use, making more virtual address space available to the application. It is enabled by default on 32 bit OpenGL applications with Linux 3.11+ and glibc 2.19+. Memory allocated this way will consume space in /dev/shm. Setting the environment variable __GL_DevShmPageableAllocations to 2 will disable this feature
• Added support for OpenGL 4.3.
• Added support for X.Org xserver ABI 20 (xorg-server 1.18).

What is new in version 355.11:

• Added support for the following GPUs:
• GeForce GTX 950
• Quadro M4000
• Quadro M5000
• Fixed a bug that caused applications using OpenCL or CUDA with OpenGL interopability to crash when using GLX indirect rendering.
• Fixed a bug that caused VDPAU to only display the top half of a video frame when decoding and displaying H.265/HEVC encoded video streams.
• Fixed a bug that could cause GPU exceptions when using glReadPixels() on the depth or stencil buffers of multisample windows while resizing or moving the windows.
• Fixed a bug that could cause data from one texture level to overwrite data from the next lowest level, when creating a texture view that did not include the higher of the two levels.
• Fixed a bug that could cause the nvidia-settings control panel to crash when updating the display layout.
• Corrected some erroneous reporting of support for GLX extensions: several extensions were being reported as supported for indirect GLX, which were in fact only supported under direct rendering.
• Added support for the following EGL extensions:
• EGL_KHR_swap_buffers_with_damage
• EGL_NV_stream_consumer_gltexture_yuv
• Replaced the build system for the NVIDIA kernel modules and updated the installer package and nvidia-installer to use the new build system and kernel module source code layout. For more information about the new build system and layout, see the README document at: ftp://download.nvidia.com/XFree86/packaging/linux/new-kbuild-for-355/
• Added experimental full OpenGL support to EGL.
• Marked the DeleteUnusedDP12Displays option as deprecated.
• Version 1.5.0 of the X Resize and Rotate specification added a note that dynamically-created outputs will not be destroyed, so this option is deprecated and will be removed in a future driver release.
• Added support for VDPAU profiles added in VDPAU 0.9:
• VDP_DECODER_PROFILE_H264_BASELINE
• VDP_DECODER_PROFILE_H264_CONSTRAINED_BASELINE
• VDP_DECODER_PROFILE_H264_EXTENDED
• VDP_DECODER_PROFILE_H264_PROGRESSIVE_HIGH
• VDP_DECODER_PROFILE_H264_CONSTRAINED_HIGH
• Fixed a bug that prevented more than one RandR output from sharing user-added modes.
• Fixed a bug that caused application-specified swap intervals to be ignored on some screens when using Xinerama.
• Fixed a bug that caused user-supplied RandR modes with nonsensical combinations of the +HSync, -HSync, +VSync, and -VSync flags to corrupt the mode list.
• Added support to make an OpenGL 3.0 and above context current without making current to any drawable.

What is new in version 352.21:

• Added support for the following GPUs:
• Tesla K40t
• GeForce 720A
• GeForce 920A
• GeForce 930A
• GeForce 940A
• GeForce GTX 950A
• GeForce GTX 980 Ti
• Fixed a bug in nvidia-settings that caused the application to crash when saving the EDID to a file.
• Fixed a bug that prevented the "mkprecompiled" utility included in the driver package from reading files correctly.
• Fixed a bug that could cause an Xid error when terminating a video playback application using the overlay presentation queue in VDPAU.
• Updated nvidia-installer to avoid recursing too deeply into kernel source trees under /usr/lib/modules, mirroring an existing restriction on recursion under /lib/modules.
• Updated nvidia-installer to use modprobe(8) when leaving the NVIDIA kernel module loaded after installation, instead of insmod(8) or libkmod. This allows the kernel module to honor any configuration directives that apply to it in /etc/modprobe.d when it is loaded.
• Fixed a bug that allowed console messages from the Linux kernel to be drawn over the user interface of nvidia-installer.

What is new in version 346.72:

• Fixed a bug causing loss of stereo synchronization in certain Quadro Sync framelock configurations.
• Fixed Pixel Buffer Object operations when row length is less than width. GL_[UN]PACK_ROW_LENGTH can be set to a value lower than the width of the operation being carried out. The OpenGL specification allows for this (the source or destination lines will be overlapping). Previously, our implementation of Pixel Buffer Objects did not support this case and would throw an error.
• Fixed a rare deadlock condition when running applications that use OpenGL in multiple threads on a Quadro GPU.