I don’t understand. The GPU was created as a dedicated module. It just happened that the parallel tasks needed for rendering games could be used first for crypto and later LLM’s.
The problem is factory capacity. It doesn’t matter how you divide up the components based on purpose if one company prepays all factories to only make components for themselves.
That’s even worse. It means there will be no left over GPUs for gaming. Factories will produce AI specific chips and nothing else because gamers won’t pay $10k for a GPU card.
it’s almost like there should be dedicated modules for each purpose, rather than the GPU being the one-stop-shop for high-demand computing.
I don’t understand. The GPU was created as a dedicated module. It just happened that the parallel tasks needed for rendering games could be used first for crypto and later LLM’s.
The problem is factory capacity. It doesn’t matter how you divide up the components based on purpose if one company prepays all factories to only make components for themselves.
Specialized AI chips are coming… AIUs lol or SIUs (is you’re stupid as F)
That’s even worse. It means there will be no left over GPUs for gaming. Factories will produce AI specific chips and nothing else because gamers won’t pay $10k for a GPU card.
NPU are kinda useless