Local LLM use isn't really continuous though. You're comparing Apples to oranges. For training especially and for the datacenter, NVIDIA cards still make absol

Local LLM use isn't really continuous though. You're comparing Apples to oranges. For training especially and for the datacenter, NVIDIA cards still make absolute sense. But for consumer use cases, and even for developers, it's not so cut and dry. Very comparable performance for short-term bursts of compute.