microsoft
Nvidia GPU passthrough fails with GPUs not consistently in WDDM mode
Windows Version
Microsoft Windows [Version 10.0.26100.4061]
WSL Version
2.4.13.0
Are you using WSL 1 or WSL 2?
- [x] WSL 2
- [ ] WSL 1
Kernel Version
Linux version 5.15.167.4-microsoft-standard-WSL2
Distro Version
Ubuntu 22.04
Other Software
Nvidia GPU driver version is 572.61 and 576.52 tested
Repro Steps
take a sytem with at least two Nvidia CPUs, one in TCC mode and the other in WDDM mode run nvidia-smi inside WSL ->
Unable to determine the device handle for GPU0: 0000:C1:00.0: Unknown Error
Unable to determine the device handle for GPU1: 0000:E1:00.0: Unknown Error
No devices were found
(more detailed descriptions already posted in Nvidia Developer Forum: https://forums.developer.nvidia.com/t/wsl2-nvidia-smi-unable-to-determine-the-device-handle-for-gpu/334254 and https://forums.developer.nvidia.com/t/tesla-gpu-on-windows-server-2022-not-detected-in-wsl-and-docker-containers/333910 )
Expected Behavior
the GPUs in TCC mode should be ignored as TCC driver mode is currently not supported with WSL (https://github.com/microsoft/WSL/issues/9952) but the GPUs in WDDM mode should work normally
Actual Behavior
no GPUs can be used in WSL
Diagnostic Logs
No response
Logs are required for review from WSL team
If this a feature request, please reply with '/feature'. If this is a question, reply with '/question'. Otherwise please attach logs by following the instructions below, your issue will not be reviewed unless they are added. These logs will help us understand what is going on in your machine.
How to collect WSL logs
Download and execute collect-wsl-logs.ps1 in an administrative powershell prompt:
Invoke-WebRequest -UseBasicParsing "https://raw.githubusercontent.com/microsoft/WSL/master/diagnostics/collect-wsl-logs.ps1" -OutFile collect-wsl-logs.ps1
Set-ExecutionPolicy Bypass -Scope Process -Force
.\collect-wsl-logs.ps1
The script will output the path of the log file once done.
If this is a networking issue, please use collect-networking-logs.ps1, following the instructions here
Once completed please upload the output files to this Github issue.
Click here for more info on logging If you choose to email these logs instead of attaching to the bug, please send them to [email protected] with the number of the github issue in the subject, and in the message a link to your comment in the github issue and reply with '/emailed-logs'.
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
The log file doesn't contain any WSL traces. Please make sure that you reproduced the issue while the log collection was running.
Diagnostic information
Detected appx version: 2.4.13.0
Found no WSL traces in the logs
The log collection was running. The issue is difficult to capture using the provided log collection script because it’s about something not happening — specifically, the GPU isn’t available in WSL when it should be. I included a command in the issue description that shows an error on the command line, but that output doesn’t show up in the collected logs as the Nvidia tool terminates properly...
@HMedbian: Can you capture a full WSL boot and a repro of the error ? You can do that by running wsl --shutdown before running the script.
/logs
@OneBlue : Thank you for taking a look. I have captured the logs again with shutting down before: WslLogs-2025-06-03_01-04-10.zip
FYSA - on my system, "nvidia-smi -i 0" works on wsl
0 is my WDDM card - 4090 1 is my TCC card - P40
"nvidia-smi -i 1" returns: Unable to determine the device handle for GPU0000:05:00.0: Unknown Error
kernel is 5.15.167.4-microsoft-standard-WSL2
Seems like both of the failure examples linked have the TCC card as card 0.
@chlimouj Thank you for your comment — that's indeed very interesting. I believe the NVIDIA driver orders GPUs based on their PCI bus IDs, so simply swapping the GPUs’ PCI slots might serve as a workaround for some users. Unfortunately, I can't try this myself due to space constraints in my PC case, but I’ll forward this information to the NVIDIA Developer Forum. Perhaps someone there can try it out.