A deep dive into the architecture of Nvidia's Rubin CPX chip, which is optimized for long-context AI tasks and the prefill phase of inference (SemiAnalysis)

SemiAnalysis:
A deep dive into the architecture of Nvidia's Rubin CPX chip, which is optimized for long-context AI tasks and the prefill phase of inference  —  New Prefill Specialized GPU, Rack Architecture, BOM, Disaggregated PD, Higher Perf per TCO, Lower TCO, GDDR7 & HBM Market Trends

from Techmeme https://ift.tt/xVpBA2s