It's not really 100k years. A ROCm subset good enough to run an LLM can be done ...

		cyberax on Nov 7, 2024 \| parent \| context \| favorite \| on: Nvidia Rides AI Wave to Pass Apple as Largest Comp... It's not really 100k years. A ROCm subset good enough to run an LLM can be done within a year or so. You can _technicall_ run it now with tinygrad. The silence from AMD has been deafening, though. I can't fathom why they're just ignoring the AI market.