AMD Makes A New Open-Source AI Software Acquisition

[phoronix]

Phoronix: AMD Makes A New Open-Source AI Software Acquisition

It was just a few weeks ago that AMD acquired AI software company Mipsology to help their AI software efforts on FPGAs. Today AMD announced another notable AI software acquisition: open-source AI software vendor Nod.ai...

https://www.phoronix.com/news/AMD-Acquires-Nod-AI

[unwind-protect]

That would be nicer if they allowed arbitrary cards - including gaming cards - to the used for ROCm, like NVidia does.

[Retramon]

That would be nicer if they allowed arbitrary cards - including gaming cards - to the used for ROCm, like NVidia does.

Yeah that's a real bummer, especially because it makes sense try to compete with CUDA in every way but I think they will hopefully get a whip of wisdom from their own OpenSIL not letting out Ryzen Desktop. I think ROCm simply isn't built-up yet and not ready for primetime practical use so hence limited support, perhaps this is a good thing for the time being.

[skeevy420]

Yeah that's a real bummer, especially because it makes sense try to compete with CUDA in every way but I think they will hopefully get a whip of wisdom from their own OpenSIL not letting out Ryzen Desktop. I think ROCm simply isn't built-up yet and not ready for primetime practical use so hence limited support, perhaps this is a good thing for the time being.

Yep and, to be fair, CUDA has almost 10 years on ROCm.

[geerge]

That would be nicer if they allowed arbitrary cards - including gaming cards - to the used for ROCm, like NVidia does.

Get that possibility out of your head, it would be nice but rocm is not for those with hope.

[edwaleni]

Maybe I should announce an AI startup, make the rounds of various conferences, get one round of angel funding and then sell myself off. It seems to be the "in" thing to do right now.

[brucethemoose]

Maybe I should announce an AI startup, make the rounds of various conferences, get one round of angel funding and then sell myself off. It seems to be the "in" thing to do right now.

Nod.ai's SHARK framework is, as far as I know, still the fastest way to run Stable Diffusion on AMD GPUs. And its free.

I dunno if they're directly making contributions to IREE/MLIR as well, but if they are, its probably one of the best acquisitions AMD could make... As long as AMD doesn't disrupt that work. They are sustainably chipping away at the CUDA moat.