Have you ever tried doing a GPU passthrough on Linux (host can either be Windows or Linux)?
Assuming that you blocked the nvidia drivers from installing, is there a way to set up a GPU passthrough on an EndeavourOS VM?
Also can you use NVIDIA as a GPU/CUDA tool and virtio as a display or should it be removed?
It’s actually very easy, if you have a 2nd GPU to dedicate to the VM. I ran this on multiple systems for a few years, but decided it wasn’t worth it in my use-case.
I would take a look at the VM Curator project, as well as Virt-Manager. VM Curator has explicit instructions on GPU-passthrough, with both single and dual GPU systems.
Already dedicated the 3080 (PCIe passthrough), then nvidia-smi ended up not seeing the 3080 (lspci did see the 3080 though). Been trying to install llama.cpp and ollama in the VM (qwen3:9b and some other model).
Anyway I’ll check VM Curator for the Endeavour VM