4 ms·
AMD and Cerebras Launch AI Inference Solution
- ckrapu 3mo agoThey're a bit late to the party.
- rbanffy 3mo agoWhy would you say that?
- techgnosis 3mo agoI'm not sure I understand. It's Helios but with WSE attached? And it uses one or the other depending on some criteria?
- wtallis 3mo agoLLM inference is a two-phase process. The first phase is prompt processing aka prefill. It's compute-heavy but requires relatively low memory bandwidth. The second phase is token generation aka decode, which doesn't require much in the way of FLOPs but wants as much memory bandwidth as possible. This announcement is for a system to do the first phase on Helios and the second phase on WSE.
- sccvcxv 3mo agolol when is the music gonna stop?
- rbanffy 3mo agoNobody knows the answer to that, but if it’s efficiency they are aiming for, which results in lower inference costs, it’ll last for a while. The inference to training ratio will only get higher and the margins will only get lower.
- throw1234567891 3mo agoWhen Etched becomes GA.
- JSR_FDED 3mo agoIt’s disaggregated so you Know it’s good
- rbanffy 3mo agoThe two major sides of the system excel at different things that happen to be complementary in AI inference.