The apple silicon architecture is really interesting, but its wasted on the company managing and producing the products built on it.
The biggest issue facing all silicon right now is CUDA. If you can’t do ML acceleration, in a meaningful way that connects with existing architectures, its a nonstarter. I’m not sure what the solution is, but it isn’t metal. We need a FOSS drop in replacement for CUDA similar to x86
The apple silicon architecture is really interesting, but its wasted on the company managing and producing the products built on it.
The biggest issue facing all silicon right now is CUDA. If you can’t do ML acceleration, in a meaningful way that connects with existing architectures, its a nonstarter. I’m not sure what the solution is, but it isn’t metal. We need a FOSS drop in replacement for CUDA similar to x86
Thoughts and prayers to Vosen, AMD, and ZLUDA.
I mean, there is OpenCL and AMD is actually compatible with lots of frameworks these days. It’s not impossible, people just don’t want to