I am using it for ML workloads at the moment, so CLI is the only way, anyways. Right now, I have a separate linux box, with a cobbled together VFIO VM. Nvidia doesn't want to support it (for GeForce - or lower end Quadros without infrastructure and licensing costs), AMD just says "you do all of the development to make this work!" and Intel's works, but doesn't have anywhere near the performance necessary to matter. This hasn't even covered the difficulties of getting Optimus gpus working well in Linux. Meanwhile, MSFT is presenting a first party solution (basically the inverse of my current setup) with the support of Nvidia.

MSFT did say they were working on graphical interface/application acceleration of WSL2 applications, but I'm not interested in that.

Quote from: Sam Medley on June 18, 2020, 03:40:44
@jeremy:
From what I can tell, WSL only runs in the command line and should only engage the GPU when the program or app is specifically told to. That would be an app-specific command. Optimus shouldn't interfere with an app making direct calls to the GPU, but that would be interesting to test out.

@jeremy:
From what I can tell, WSL only runs in the command line and should only engage the GPU when the program or app is specifically told to. That would be an app-specific command. Optimus shouldn't interfere with an app making direct calls to the GPU, but that would be interesting to test out.

Pretty cool. If it works with Optimus, then I can simplify my environment and workflow a bit.