|
Size: 2497
Comment:
|
Size: 5818
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 1: | Line 1: |
| ## page was renamed from Howto/NVIDIA_Optimus ## page was renamed from Howto/nVidia_Optimus |
|
| Line 2: | Line 4: |
| There are additional steps needed to integrate the package with optimus. | With Fedora 25 and later, NVIDIA Optimus devices are automatically detected with GDM/Gnome. They will run with output source and not with offload sync by default. |
| Line 4: | Line 7: |
| This Howto is a subset of the main documentation, please read the [[Howto/nVidia|Howto NVIDIA]] first. | With Fedora 29 and later, along NVIDIA driver 435.17+ and xorg-x11-server 1.20 (add additional patches), it's now possible to have full Optimus support (with offload sync only with glx). Starting with NVIDIA driver 440.26+ render offload works with EGL. |
| Line 6: | Line 10: |
| Also note that NVIDIA currently only support "outputsource" and not "offloadsink". It means that you cannot disable the dGPU (nvidia). The current workaround is to reboot onto the free Software version using an alternative grub2 boot menu. |
|
| Line 9: | Line 11: |
| This is only tested using the main driver (367.xx and later). | This Howto is a subset of the main documentation, please read the [[Howto/NVIDIA|NVIDIA]] Howto first. |
| Line 11: | Line 13: |
| !!! BIG FAT WARNING - This is still experimental documentation, only experimented users should follow. I highly recommends to have another PC for remote debug. |
|
| Line 14: | Line 14: |
| == GLVND enabled mesa == This is optional for optimus, and will provide a glvnd enabled mesa that will replace the fedora version. Once installed, it will be easier to switch between FOSS stack (nouveau) and the binary driver (nvidia). The long term plan is to have the switch enabled in the fedora mesa package (see rhbz#1388810 ). |
== Known limitation == Full Optimus Support can be achieved with xorg-x11-server 1.20.5 with patches as provided in a COPR repository. For information, the support as landed officially in Xorg server 1.21 that might be provided by default for Fedora 32. I Highly recommends Optimus users to enable this COPR repository to benefit from Full Optimus Support ahead of Xorg 1.21. |
| Line 19: | Line 19: |
| Note: Copr is not multilibs compliant, you need to copy the x86_64 repo as a new file and replace "arch" to i386, if using 32bit programs. | == NVIDIA PRIME Support == |
| Line 21: | Line 22: |
| # dnf copr enable kwizart/glvnd # dnf update |
sudo -s dnf copr enable kwizart/xorg-x11-server_nvidia dnf update xorg-x11-server\* dnf install rpmfusion-nonfree-release-rawhide dnf update --enablerepo=rpmfusion-nonfree-rawhide xorg-x11-drv-nvidia\* cat > /etc/modprobe.d/nvidia.conf <<EOF # Enable DynamicPwerManagement # http://download.nvidia.com/XFree86/Linux-x86_64/435.17/README/dynamicpowermanagement.html options nvidia NVreg_DynamicPowerManagement=0x02 EOF }}} == PRIME Render Offload == As documented at http://download.nvidia.com/XFree86/Linux-x86_64/435.17/README/primerenderoffload.html === 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, 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. NVIDIA's EGL implementation does not yet support PRIME render offload. Examples: {{{ __NV_PRIME_RENDER_OFFLOAD=1 vkcube __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo | grep vendor }}} === 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. The VK_LAYER_NV_optimus layer causes the GPUs to be sorted such that the NVIDIA GPUs are enumerated first. 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 only report NVIDIA GPUs to the Vulkan application. The value non_NVIDIA_only causes VK_LAYER_NV_optimus to only report non-NVIDIA GPUs to the Vulkan application. 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 }}} === Finer-Grained Control of GLX + OpenGL === For GLX + OpenGL, the environment variable __NV_PRIME_RENDER_OFFLOAD_PROVIDER provides finer-grained control. While __NV_PRIME_RENDER_OFFLOAD=1 tells GLX 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 }}} == NVIDIA PrimaryGPU Support == Before the Full optimus support, the only way to enable the NVIDIA driver was to set the NVIDIA GPU to be used by default. To recover this previous behaviour, you can use: {{{ cp -p /usr/share/X11/xorg.conf.d/nvidia.conf /etc/X11/xorg.conf.d/nvidia.conf }}} And edit the file to use: Option "PrimaryGPU" "yes" == PRIME Synchronization == With Xorg server 1.19 (Fedora 25 and later), this feature allows buffer sharing between the Intel and the NVIDIA card when using Xorg (not relevant for Wayland). It should be enabled bytThen you can enable Prime Synchronisation for your session: {{{ xrandr --output <output> --set "PRIME Synchronization" 1 |
| Line 26: | Line 85: |
| == Proprietary/FLOSS switch == At this time, this can be done by manually editing "rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1" from the grub2 cmdline. And replace by "rd.driver.blacklist=nvidia,nvidia_drm,nvidia_modeset modprobe.blacklist=nvidia,nvidia_drm,nvidia_modeset" The plan is to have a grub2 menu to have the choice. Please see [[https://bugzilla.rpmfusion.org/show_bug.cgi?id=4315|Bugzilla [RFE] Switcher for Xorg nvidia/FOSS config]] |
|
| Line 27: | Line 89: |
== Additional Configuration == * For Fedora 25 and later. Please verify to have the modified xorg-x11-server package currently in updates-testing repository. It contains some not yet upstream [[https://lists.x.org/archives/xorg-devel/2016-December/052066.html|patches]] by Hans De Goede (Fedora/RPM Fusion contributor). {{{ # dnf update xorg-x11-server\* --enablerepo=updates-testing # rpm -q xorg-x11-server-Xorg xorg-x11-server-Xorg-1.19.1-1.fc25 }}} * Up to Fedora 24 , you still need a xorg.conf and adapted display manager configuration (GDM, KDM, LightDM, etc). Best is to upgrade to Fedora 25. In case you previously had a working f24 setup. You need to remove the previous configuration files. |
|
| Line 41: | Line 92: |
| *A: nvidia-prime is not something from NVIDIA despite the name. It's a collection of integration scripts made by canonical for Ubuntu. Best would be to avoid using custom scripts and to have the driver to setup appropriately if on Optimus hardware or single GPU setup. | *A: nvidia-prime is not something from NVIDIA despite the name. It's a collection of integration scripts made by canonical for Ubuntu. Better to avoid using custom scripts and to have the driver to setup appropriately if on Optimus hardware or single GPU setup. With Fedora 25 and later, everything is automatically setup. |
| Line 44: | Line 95: |
| * http://download.nvidia.com/XFree86/Linux-x86_64/435.17/README/primerenderoffload.html * http://download.nvidia.com/XFree86/Linux-x86_64/435.17/README/dynamicpowermanagement.html |
Introduction
With Fedora 25 and later, NVIDIA Optimus devices are automatically detected with GDM/Gnome. They will run with output source and not with offload sync by default.
With Fedora 29 and later, along NVIDIA driver 435.17+ and xorg-x11-server 1.20 (add additional patches), it's now possible to have full Optimus support (with offload sync only with glx). Starting with NVIDIA driver 440.26+ render offload works with EGL.
This Howto is a subset of the main documentation, please read the NVIDIA Howto first.
Known limitation
Full Optimus Support can be achieved with xorg-x11-server 1.20.5 with patches as provided in a COPR repository. For information, the support as landed officially in Xorg server 1.21 that might be provided by default for Fedora 32. I Highly recommends Optimus users to enable this COPR repository to benefit from Full Optimus Support ahead of Xorg 1.21.
NVIDIA PRIME Support
sudo -s dnf copr enable kwizart/xorg-x11-server_nvidia dnf update xorg-x11-server\* dnf install rpmfusion-nonfree-release-rawhide dnf update --enablerepo=rpmfusion-nonfree-rawhide xorg-x11-drv-nvidia\* cat > /etc/modprobe.d/nvidia.conf <<EOF # Enable DynamicPwerManagement # http://download.nvidia.com/XFree86/Linux-x86_64/435.17/README/dynamicpowermanagement.html options nvidia NVreg_DynamicPowerManagement=0x02 EOF
PRIME Render Offload
As documented at http://download.nvidia.com/XFree86/Linux-x86_64/435.17/README/primerenderoffload.html
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, 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. NVIDIA's EGL implementation does not yet support PRIME render offload.
Examples:
__NV_PRIME_RENDER_OFFLOAD=1 vkcube __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxinfo | grep vendor
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. The VK_LAYER_NV_optimus layer causes the GPUs to be sorted such that the NVIDIA GPUs are enumerated first. 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 only report NVIDIA GPUs to the Vulkan application. The value non_NVIDIA_only causes VK_LAYER_NV_optimus to only report non-NVIDIA GPUs to the Vulkan application.
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
Finer-Grained Control of GLX + OpenGL
For GLX + OpenGL, the environment variable NV_PRIME_RENDER_OFFLOAD_PROVIDER provides finer-grained control. While NV_PRIME_RENDER_OFFLOAD=1 tells GLX 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:
Before the Full optimus support, the only way to enable the NVIDIA driver was to set the NVIDIA GPU to be used by default. To recover this previous behaviour, you can use: And edit the file to use: Option "PrimaryGPU" "yes"
With Xorg server 1.19 (Fedora 25 and later), this feature allows buffer sharing between the Intel and the NVIDIA card when using Xorg (not relevant for Wayland). It should be enabled bytThen you can enable Prime Synchronisation for your session:
At this time, this can be done by manually editing "rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1" from the grub2 cmdline. And replace by "rd.driver.blacklist=nvidia,nvidia_drm,nvidia_modeset modprobe.blacklist=nvidia,nvidia_drm,nvidia_modeset" The plan is to have a grub2 menu to have the choice. Please see Bugzilla [RFE] Switcher for Xorg nvidia/FOSS config
__NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears
__NV_PRIME_RENDER_OFFLOAD_PROVIDER=NVIDIA-G0 __GLX_VENDOR_LIBRARY_NAME=nvidia glxgears
NVIDIA PrimaryGPU Support
cp -p /usr/share/X11/xorg.conf.d/nvidia.conf /etc/X11/xorg.conf.d/nvidia.conf
PRIME Synchronization
xrandr --output <output> --set "PRIME Synchronization" 1
Proprietary/FLOSS switch
FAQ
References