Hello,

I installed the Ollama container a while ago and have been using it from time to time for a few questions, and it works very well.

However, I wanted to use image analysis (with qwen2.5vl:32b) and there was a problem: my CPU usage went up to 100%.

(EDIT: in fact even with normal text requests with deepseek-r1 I have the same problem).

However, in unRAID dashboard, in the GPU section, I can see the Ollama logo in the active apps...

If I run nvidia-smi in the Ollama container, it recognizes the card but says "No running processes found."

Here are some screenshots:

I tested everything: I added the environment variables "NVIDIA_DRIVER_CAPABILITIES: all," "NVIDIA_VISIBLE_DEVICES: GPU-afa2e9e4-37ae-beec-51f3-6655e5ff3698", I added "--gpus device=GPU-afa2e9e4-37ae-beec-51f3-6655e5ff3698" to the Extra parameters, etc.

Here is my config:

When I start Ollama and read logs, it seems to see the GPU:

time=2025-10-09T09:42:59.736Z level=INFO source=gpu.go:217 msg="looking for compatible GPUs"
time=2025-10-09T09:42:59.909Z level=INFO source=types.go:131 msg="inference compute" id=GPU-afa2e9e4-37ae-beec-51f3-6655e5ff3698 library=cuda variant=v13 compute=8.9 driver=13.0 name="NVIDIA GeForce RTX 4090" total="22.0 GiB" available="18.5 GiB"

Did I make a mistake somewhere? How can I make the GPU handle everything?

Edited October 9, 2025Oct 9 by Benji7790

Hi, did you ever found a solution? I have exactly same problem, with 96%/4% CPU/GPU usage and i feel like i already tried everything.

Thanks