It's advertised as a "clean room" re-implementation. What part would be illegal?

If they had to reverse engineer any compiled code to do this, I think that would be against licenses they had to agree to?

At least grounds for suing and starting an extensive discovery process and possibly a costly injunction...

We have not reverse engineered any compiled code in the process of developing SCALE.

It was clean-room implemented purely from the API surface and by trial-and-error with open CUDA code.

Isn't that exactly what a "clean room" approach avoids?

oh definitely. But if I was NVIDIA I'd want to verify that in court after discovery rather than relying on their claim on a website.

good point

FWIW, I think this is really great work and I wish only the best for scale. Super impressed.

Can't run useful shit on it: https://docs.nvidia.com/deeplearning/cudnn/latest/reference/...

Namely:

"4.1 License Scope. The SDK is licensed for you to develop applications only for use in systems with NVIDIA GPUs."

So add a cheap NVidia card alongside grunty AMD ones, and check for its existence. It doesn't seem to say it needs to run on NVidia GPUs.

Heh, true. On the other hand, I bet companies are eager to challenge the wrath of a $3T company for a promise of "maybe it'll work, not all of it but at least it'll run worse, at least for now".

I don't think the terms of the Nvidia SDK can restrict running software without said SDK. Nvidia's libraries don't seem to be involved here. Their hardware isn't involved either. It's just some ascii in a bunch of text files being hacked around with before running on someone else's hardware.