Comment by pmarreck

Comment by pmarreck 2 days ago

0 replies

In the arguably-unique case of Apple Silicon, I'm not sure about that. The SoC-integrated GPU and unified RAM ends up being extremely good for running LLM's locally and at low energy cost.

Of course, there's the upfront cost of Apple hardware... and the lack of server hardware per se... and Apple's seeming jekyll/hyde treatment of any use-case of their GPU's that doesn't involve their own direct business...