Podman + PRIME (dual gpu intel + nvidia) not working

netbrain commented 10 months ago

When using PRIME the following command works for offloading zwift onto the nvidia gpu in docker.

docker run  -d \
  -e __NV_PRIME_RENDER_OFFLOAD=1 \
  -e __GLX_VENDOR_LIBRARY_NAME=nvidia \
  --rm \
  --privileged \
  -e DISPLAY=:0 \
  -v /tmp/.X11-unix:/tmp/.X11-unix \
  -v /run/user/1000/pulse:/run/user/1000/pulse \
  -v zwift-netbrain:/home/user/Zwift \
  --gpus device=GPU-f9b76734-5873-5957-26f3-99604ac5f15a docker.io/netbrain/zwift:latest

However the equivalent command for podman launches the zwift instance on the intel graphics card.

Should probably look into the following resources in order to determine how this can be solved for podman.

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/cdi-support.html

gregh3285 commented 5 months ago

For what it's worth, the following breadcrumb helped me get past the above: https://github.com/NVIDIA/nvidia-container-toolkit/issues/154#issuecomment-1876686322

sHedC commented 3 days ago

Did you look at #147 wonder if this is related?

netbrain / zwift

Podman + PRIME (dual gpu intel + nvidia) not working #21