At the SC 19 conference, Intel officially announced the next generation of Xe architecture GPU and 10nm technology,coded Sapphire Rapids, and the Aurora Aurora Billion Supercomputing, which will be available in 2021, will host Intel’s dream of high-performance computing at HPC.
Among the many HPC technologies, a technology called Rambo Cache (Rambo Cache) attracted attention, taken by Raja Koduri, Intel’s senior vice president and leader of the stand-alone GPU, and which is not serious compared to Intel’s previous technology naming. Showing Raja Koduri’s little humor – Rambo is arguably the representative of american hard-core movies.
Rambo Cache is essentially an intermediary layer for connecting CPU, GPU, and HBM caches, based on Intel’s EMIB encapsulation technology, providing extreme memory bandwidth and FP64 floating-point performance, and supporting memory/cache ECC correction, and the strongest RAS.
According to the editor-in-chief of Theantech’s website, Intel’s choice of the name Rambo Cache is temporary, with no specific meaning or acronym, and the final market name may be completely different from Rambo Cache.
Judging from the performance of the network, it is thought that Raja Koduri chose the name Rambo Cache, is a flash, with the intention of attacking the opponent, telling friends that Intel is still very strong.