[FEAT] <title>gpu passthrough

[nsomniac281]

Is this a new feature request?

• I have searched the existing issues

Wanted change

would it be possible to add the nvidia gpu drivers and the NVIDIA_VISIBLE_DEVICES to the container to allow for GPU passthrough?

Reason for change

this would enable vulcan and probably greatly improve performance

Proposed code change

No response

[github-actions]

Thanks for opening your first issue here! Be sure to follow the relevant issue templates, or risk having this issue marked as invalid.

[j0nnymoe]

https://github.com/linuxserver/docker-xemu#nvidia-gpu-support

[nsomniac281]

ive tried to add the variables to the container and it still tells me that it cant find a compatible gpu.

[thelamer]

"ive tried to add the variables to the container" be specific.

Basic smoke test.

docker run --rm -it -p 3001:3001 -e PIXELFLUX_WAYLAND=true -e DRINODE=/dev/dri/renderD129 -e DRI_NODE=/dev/dri/renderD129 --runtime nvidia --gpus all --shm-size=1gb linuxserver/xemu bash

(I am on 129)

Also read through Steam, PIXELFLUX_WAYLAND=true is going to become default on gaming images soon:

https://github.com/linuxserver/docker-steam/blob/master/README.md?plain=1#L66-L78

And this will be in all readmes soon:

https://github.com/linuxserver/docker-jenkins-builder/pull/374/files#diff-af4b710b8e5fc9a40817a1cd8d0b115151552b58796d34f16429dd910638e628R54-R83

Basically you pass PIXELFLUX_WAYLAND, the dri nodes, and make sure the runtime is setup properly.

This is on my 3060 at 4x res.

halo.mp4

[nsomniac281]

"ive tried to add the variables to the container" be specific.

Basic smoke test.

docker run --rm -it -p 3001:3001 -e PIXELFLUX_WAYLAND=true -e DRINODE=/dev/dri/renderD129 -e DRI_NODE=/dev/dri/renderD129 --runtime nvidia --gpus all --shm-size=1gb linuxserver/xemu bash

(I am on 129)

Also read through Steam, PIXELFLUX_WAYLAND=true is going to become default on gaming images soon:

https://github.com/linuxserver/docker-steam/blob/master/README.md?plain=1#L66-L78

And this will be in all readmes soon:

https://github.com/linuxserver/docker-jenkins-builder/pull/374/files#diff-af4b710b8e5fc9a40817a1cd8d0b115151552b58796d34f16429dd910638e628R54-R83

Basically you pass PIXELFLUX_WAYLAND, the dri nodes, and make sure the runtime is setup properly.

This is on my 3060 at 4x res.
halo.mp4

I tried to add the variables linked in the reply to my first request, though i don't believe that i did it right, i have the container running in docker on unraid 6.11.5 and honestly have no idea how to edit the container itself to add this stuff. i tried to go into the settings and add the variables, because usually that's all i have to do, this all may be slightly out of the scope of my ability.

[thelamer]

I dunno I run Docker on just a normal Linux host and (this goes for basically every dev) do not test UnRaid, so if you figure something out post it here so other people can know.

As far as I understand their setup though you can just make a compose file, but the proprietary drivers and such no clue, I just run the bin from nvidias website now. Using MIT or in repo drivers is pointless with a lot of this stuff especially with the latest nvidia container toolkit requiring CDI.

[nsomniac281]

I absolutely will, i'm trying to set up an image to use as a dedicated servers for insignia, we have a growing community and having preset servers that dont have to run natively on a physical xbox would be game changing.

[nsomniac281]

alright, so i got most of it working except for the PIXELFLUX_WAYLAND=true any time i get the image to successfully dl and build with that line added the emu never actually starts. I'm relatively certain that it's an issue on the unraid side of things. I will try and figure it and when i do ill post a full guide on how to get it all working for everyone else. It could also be that my server just isn't quite powerful/new enough also.

Specs:
Dell R720xd with UnRaid 6.11.5
2x Intel® Xeon® CPU E5-2670 v2 @ 2.50GHz base
64gb samsung b-die ddr3-1800mhz
nVidia Quadro P2200 5gb w/ nVidia Driver Version: 565.57.01
128TB WD Red Pro HDD
2x 1TB Sandisk SSD

[thelamer]

Try different clients too, Firefox works best now because it usually uses libav for decoding.

[LinuxServer-CI]

This issue has been automatically marked as stale because it has not had recent activity. This might be due to missing feedback from OP. It will be closed if no further activity occurs. Thank you for your contributions.