Anyone have definitive guidance on getting Nvidia Prime offloading to work in Fedora 35

I’ve spent all day messing with it.

Can anyone that has it working say if they are using Wayland or Xorg?

I followed the core steps from rpmfusion:
https://rpmfusion.org/Howto/NVIDIA
and the linked Optimus Page there, which also suggests the following
https://download.nvidia.com/XFree86/Linux-x86_64/495.46/README/primerenderoffload.html

Here is my xorg.conf file:
Section "OutputClass"
        Identifier "nvidia"
        MatchDriver "nvidia-drm"
        Driver "nvidia"
        Option "AllowEmptyInitialConfiguration"
        Option "SLI" "Auto"
        Option "BaseMosaic" "on"
EndSection

Section "ServerLayout"
        Identifier "layout"
        Option "AllowNVIDIAGPUScreens"
        Screen 0 "igpu"
        Inactive "nvidia"
EndSection

Section "Device"
    Identifier "nvidia"
    Driver "nvidia"
    BusID "PCI:1:0:0"
EndSection

Section "Device"
    Identifier "igpu"
    Driver "modesetting"
    BusID  "PCI:6:0:0"
EndSection

Section "Screen"
    Identifier "igpu"
    Device "igpu"
EndSection

Section "Screen"
    Identifier "nvidia"
    Device "nvidia"
    Option "AllowEmptyInitialConfiguration" "True"
#   Option "PrimaryGPU" "yes"
EndSection

I’ve tried disabling Wayland, but I’m not sure I did with the right xorg.conf at the time.

With this xorg.conf, under wayland i get:

xrandr --listproviders
Providers: number : 0

With an earlier simpler xorg.conf, I had two providers, but right clicking steam to launch with discrete gpu wouldn’t do anything and xorg or gnome-shell always seemed like it was running from nvidia-smi. Ideally I’d like only heavy applications I tell to use the nvidia gpu to use it.

1 Like

We have this:

https://docs.fedoraproject.org/en-US/quick-docs/how-to-set-nvidia-as-primary-gpu-on-optimus-based-laptops/

Worth a look?

Thanks Ankur, I did see that and was able to make the nvidia as primary accidentially.

I’m actually wanting to go the reverse direction and have it work in hybrid mode so i can use prime render offload. I tried pop_os briefly cause it pitches that it’s easy to do there. Whats interesting is I had the same result on the default fedora/rpmfusion nvidia install… so I may just assume it’s working as expected and hope someone figures out how to powerdown the dgpu when not in use instead of having it attach to idling xorg. Unless anyone has suggestions to try or confirm that I have prime offload hybrid working.

1 Like