Love the lectures, especially on the newer architectures. What do you think of the GroqChip for LLMs, not sure if it’s all scale, but they seem to have made some real advances in inference times. Is it SIMD or more VLIW or a mix of both? Seems very different from the original TPU, but by the same designer. Your thoughts? Can you cover in the next update to this lecture? Thanks
love your lecture！
Love the lectures, especially on the newer architectures. What do you think of the GroqChip for LLMs, not sure if it’s all scale, but they seem to have made some real advances in inference times. Is it SIMD or more VLIW or a mix of both? Seems very different from the original TPU, but by the same designer. Your thoughts? Can you cover in the next update to this lecture? Thanks