Hi everyone, I have a problem with my AMD Radeon RX580 GPU, which won’t run on Zima OS. Is there anyone who can explain how to get my GPU to work? Thank you very much.
The RX 580 should normally use the Linux amdgpu driver, but ZimaOS is not a normal desktop Linux distro, so we need to first check if the card is detected and if the driver loads.
Can you SSH into ZimaOS and post the output of:
lspci -nn | grep -Ei 'vga|3d|display|amd|ati|radeon'
dmesg | grep -Ei 'amdgpu|radeon|drm|firmware' | tail -80
ls -lah /dev/dri
That will show whether the issue is GPU detection, missing AMD driver/firmware, or the app/container not getting GPU access.
Hi Saputra,
I had a look, but it is hard to read the command results clearly from the video.
From the screenshot, ZimaOS does appear to detect the RX 580 at hardware level, which is a good sign.
Can you please copy/paste the text output from these commands instead of video?
ls -lah /dev/dri
dmesg | grep -Ei 'amdgpu|radeon|drm|firmware' | tail -80
This will show whether the AMD driver is loading properly and whether the GPU is available to apps/containers.
Hi Saputra,
Thanks, that output helps.
Your RX 580 is detected and the AMD driver is loading correctly.
I can see:
/dev/dri/card0
/dev/dri/renderD128
And the logs show:
amdgpu kernel modesetting enabled
UVD and UVD ENC initialized successfully
VCE initialized successfully
So this does not look like a basic driver detection problem.
The message:
Cannot find any crtc or sizes
usually points more toward display/output detection, for example no monitor/EDID/display mode being detected, rather than the GPU not loading.
What are you trying to use the RX 580 for?
Display output, Jellyfin/Plex/Emby transcoding, or VM passthrough?
If it is for a Docker app like Jellyfin, then the next thing to check is whether the container has access to:
/dev/dri
The host side looks mostly OK from your screenshots.
Good evening, how can I direct VGA to Ollama? I don’t know how.
For Ollama, it is not really “VGA output” that gets passed through.
Ollama needs GPU compute support, usually through ROCm or Vulkan.
Your RX 580 is detected by ZimaOS and /dev/dri exists, but RX 580 is an older Polaris card and is not listed in Ollama’s current supported AMD ROCm GPU list.
So it may work for display or some media tasks, but Ollama GPU acceleration may not work on this card.
You can still run Ollama on CPU, but for GPU acceleration an NVIDIA card or a newer supported AMD GPU is usually the safer path.
Excuse me, may I ask, if one day I buy a NVIDIA 30 series GPU, will it be detected and visible in the ZimaOS dashboard?
You can also check the official ZimaCube GPU expansion guide here:
It lists supported GPU families, including NVIDIA RTX 30 series cards.
Just note, there are still two separate things:
Detected by ZimaOS / shown in dashboard
and
Usable by Ollama for GPU acceleration
For Ollama, the important host check is:
nvidia-smi
If nvidia-smi works on ZimaOS, then GPU use with Docker/Ollama is much more realistic.