The chance for Raja Koduri to publish some Tweets.
Raja Koduri, who heads the GPU department of Intel, commonly raves on Twitter concerning the options he’s growing for the corporate. In the previous couple of hours, he has as soon as once more touted the Ponte Vecchio GPU primarily based on the Xe-HPC structure, in response to the announcement of the Leibniz Supercomputing Center to combine Intel Sapphire Rapids processors (4th technology Xeon Scalable) and Ponte Vecchio GPUs from subsequent 12 months into its SuperMUC-NG supercomputer, throughout part 2. On this regard, Raja Koduri has writes the next message, which wants no translation:

“Sapphire Rapids, Ponte Vecchio, Optane and One API
An AI dream machine😍
Plenty of CPU + GPU computation, plenty of reminiscence and open software program
Thanks @LRZ_DE and @Lenovodc to your collaboration and your belief ”.
Within the course of, Raja Koduri printed an animated picture exhibiting how the Ponte Vecchio GPU is structure.
Raja Koduri checks an Intel Xe HPG graphics card on 3DMark Mesh Shader
A supercomputer presently fifteenth within the TOP500
In reality, one can perceive the delight that Raja Koduri feels; Undoubtedly, Ponte Vecchio is a chip of nice complexity. In response to info relationship from the top of March, it carries greater than 100 billion transistors and 47 “Magical Tiles”. The promise: FP16 computing energy of 1 PFLOPS (1000 TFLOPS). For comparability, NVIDIA’s A100 gives 312 TFLOPS in FP16.
We have no idea the anticipated efficiency for the SuperMUC-NG part 2. At the moment in its part 1, it has 12,960 Intel Xeon Platinum 8174 processors. Every has 24 cores. Subsequently, the entire variety of CPU cores is 311,040 as proven on the official site. The supercomputer has a Linpack Rmax of 19.5 PFLOPS and a Rpeak of 26.9 PFLOPS, which places it within the fifteenth place of the TOP500.
Leave a Reply
You must be logged in to post a comment.