When will WSL2 support Tesla GPU?

[mszhanyi]

Is your feature request related to a problem? Please describe.
More and more people are using Azure machine to do AI development and test. So far, only Tesla GPUs (M50, k80, P100, V100) are supported in Azure.
From https://developer.nvidia.com/cuda/wsl/download, Nvidia only provides Geforce and Quardro driver for WSL.
It means we couldn't do any GPU-related job with WSL in Azure.

Describe the solution you'd like
Support Telsa driver for WSL.

[luiset83]

The Quadro driver installs on the supported insider build with a Tesla GPU, however presumably because there is no WDDM driver active on a Tesla card, WSL2 does not support running CUDA apps -- there is no /dev/dxg device exposed since Tesla cards use the TCC driver.

When attempting to run deviceQuery on WSL2 with a Tesla GPU one gets:
cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version

A better title should be support TCC (Tesla Compute Cluster) driver for CUDA applications on Tesla GPUs. I see no other hits on Google regarding this, so it's possible it was not an entirely popular use-case :(

[Hassanfarooq92]

Found a work around.
Download quadro drivers for wsl2, extract them by running installer but don't quit the installer.
Go to device manager, right click and select update the driver, check have disk and un-tick show compatible models only.
Now you have to select the similar quadro model which have the same chipset as ur tcc. For my case m40 is same as quadro m6000. Update the driver and now device manager should show it as quadro gpu (this means we were able to load the required driver for wsl2).
Now go to cmd/terminal, use nvidia smi to change the driver mode to wddm by using the command (Google it), after changing the mode from 1 to 0, it will appear as TCC* instead of TCC. You're good to go now.

[microsoft-github-policy-service]

This issue has been automatically closed since it has not had any activity for the past year. If you're still experiencing this issue please re-file this as a new issue or feature request.

Thank you!

[littleyeson]

Found a work around. Download quadro drivers for wsl2, extract them by running installer but don't quit the installer. Go to device manager, right click and select update the driver, check have disk and un-tick show compatible models only. Now you have to select the similar quadro model which have the same chipset as ur tcc. For my case m40 is same as quadro m6000. Update the driver and now device manager should show it as quadro gpu (this means we were able to load the required driver for wsl2). Now go to cmd/terminal, use nvidia smi to change the driver mode to wddm by using the command (Google it), after changing the mode from 1 to 0, it will appear as TCC* instead of TCC. You're good to go now.

Is Tesla V100 use this way to work into WSL2?