HARDWARE-ACCELERATED WSL

WSL supports running with a virtual GPU, so that Linux programs can take advantage of high-powered graphics cards and run compute jobs (not games at this stage) at near-native speeds. There’s also a DirectML backend for TensorFlow (see https:// github.com/microsoft/DirectML), enabling speedy machine learning experiments on Direct3D 12-supported hardware.

Further to this, in spring 2021 the d3d12 driver became an official part of the open source Mesa 21.0 stack.

To make use of this, and speed up your Blender renders or whatever, you might need to update your Windows GPU driver to one supporting

WDDM 3.0.

Nvidia’s WSL CUDA drivers are considered public preview, and should be delivered to Windows Insiders through the normal Windows Update channel. Otherwise they’re available at https://developer.nvidia.com/ cuda/wsl. Owners of AMD graphics hardware can get the preview drivers at www.amd.com/en/support/kb/ release-notes/rn-rad-win-wslsupport. Intel’s can be found at https://bit.ly/lxf282-intel-drivers.

Since only Ubuntu LTSes are available on WSL, they miss out on all these exciting developments. So you could install Arch, or some other bleeding-edge distro. Or you could avail yourself of Canonical’s Ubuntu on Windows Community Preview. Read more and follow the download links at https://ubuntu.com/blog/ announcing-ubuntu-on-windowscommunity-preview-wsl-2.