tartavull on MTIA v1: Meta’s first-generation AI inference accelerator · Hackerly

Comment on MTIA v1: Meta’s first-generation AI inference accelerator

How do they compare to TPUs?