Ep 11: Future of AI Computing and How to Build & Nurture Hardware Teams with Jim Keller, Tenstorrent
Computer Architecture Podcast
00:00
Nvidia's CUDA GPUs Aren't as Fast as GPUs, Right?
There's a really interesting dynamic about this. It sort of feels like when we first started building out of order processors back in, you know, I guess I started working on it in 95. And then there's ways to optimize further by breaking operations into the right side and local memory access. So they're big enough to have a good amount of compute power but small enough for local data access. But people are taking different approaches at this because people can't do everything.
Play episode from 25:30
Transcript
