Btw UCIe was a hint @ EMIB-T & Intel but I don’t think Google is an immediate customer but definitely interested. My sources tell me that the main customer for HBF is Chinese (Huawei?) which makes sense since they are completely blacklisted from HBM… The success of HBF will hinder on reliability and my view is it’s 2 years at best and although I know several techniques to fix this it’s an engineering challenge that would eat up silicon area that should be spent on compute. My guess is things will be dynamic and that disaggregated memory will be a better solution. Don’t worry guys I’ll be in the room when they share the results and I’ll make sure to ask questions. Also P.S. Jukan please follow my back ily and the capacitor thing was just a joke ❤️ will send you some intel swag to make amends.
