100x defect tolerance: How we solved the yield problemcerebras.ai331 points·by jwan584·anno scorso·179 comments
Cerebras CS-3: the fastest and most scalable AI acceleratorcerebras.net2 points·by jwan584·2 anni fa·0 comments
jwan584·anno scorso·discussThe point about using FP32 for training is wrong. Mixed precision (FP16 multiplies, FP32 accumulates) has been use for years – the original paper came out in 2017.
jwan584·anno scorso·discussA good talk on how Cerebras does power & cooling (8min) https://www.youtube.com/watch?v=wSptSOcO6Vw&ab_channel=Appli...