Hello,

I'm having a persistent issue passing my NVIDIA GTX 1080 Ti to a dockurr/windows container. The passthrough fails and Windows only sees a virtual display adapter. I have spent a long time troubleshooting this and have tried all the standard fixes.

Symptoms:

• The container starts successfully.

• Running nvidia-smi inside the container works, but it shows "No running processes."

• The Windows guest OS only sees a "Microsoft Basic Display Adapter" in Device Manager.

What I Have Already Tried and Confirmed:

• Hardware Virtualization: BIOS VT-x and VT-d are both enabled; KVM is active (/dev/kvm exists).

• Unraid Host: The Nvidia-Driver plugin is installed and nvidia-smi works correctly on the Unraid host.

• Docker Config: The container is configured with the --runtime=nvidia and --gpus all extra parameters.

• VBIOS: I successfully dumped a VBIOS file from the card using a temporary VM. The file is correctly mapped to the container's /vbios path, and the MANUAL_VBIOS environment variable is set to point to it.

• Headless Operation: The server is running "headless." The monitor is physically disconnected from the GPU, and the video=efifb:off kernel flag has been added to the boot configuration.

• Server Firewall: We have forcibly disabled the Unraid firewall by adding USE_NF="no" to the network.cfg file and rebooting.

Despite all of these steps, the GPU is still not being recognized by the Windows guest. I have attached my server's diagnostic file. Any insight would be greatly appreciated.

thinkstation-diagnostics-20250827-1002.zip

1 hour ago, josh951 said:

Despite all of these steps, the GPU is still not being recognized by the Windows guest. I have attached my server's diagnostic file. Any insight would be greatly appreciated.

i cant imagine this will work properly ...

the docker access is way different from a passed through dgpu, and as you want a passed though GPU in your VM which runs in a docker kvm enviroment ...

if you really want a win vm with a passed dGPU, may rather consider setting up a regular win VM with passed dGPU, uninstall the nvidia driver, remove all your extra's and simple vfio bind your dGPU, pass it to your VM, and you done ...