Systems like the DOE’s El Capitan are ushering in a new class of supercomputing with exascale-class systems that are 1,000X faster than the previous generation petascale systems first introduced 12 years ago.
The new performance record of 2 exaflops (2,000 petaflops) will be more powerful than the Top 200 fastest supercomputers in the world combined and is an increase of more than 30% from initially projected estimates calculated seven months ago. This was made possible by a new partnership between HPE, AMD and the U.S. DOE to combine HPE’s Cray Shasta system and Slingshot interconnect, a specialized HPC networking solution, with next-generation AMD EPYC processors and next-generation AMD Radeon Instinct GPUs.
“We are pleased to continue our longstanding journey with HPE in co-developing innovative technologies for a range of solutions, and now, for the world’s most powerful, exascale-class supercomputer,” said Forrest Norrod, senior vice president and general manager of the Datacenter and Embedded Solutions Business Group, AMD.
“The exceptional computing power promised by El Capitan, based on HPE’s Cray Shasta architecture, will ensure the NNSA laboratories can continue to excel at their national security missions and make it possible for the U.S. to remain competitive on the global stage in high-performance computing for many years to come,” said Bill Goldstein, director, Lawrence Livermore National Laboratory.
Table of Contents
Enhanced Performance for DOE’s El Capitan
Streamlined communication between HPE’s Cray Slingshot interconnects a specialized HPC networking solution and new next-generation AMD Radeon Instinct GPUs that are based on a new compute-optimized architecture for workloads including HPC and AI.
High density compute blades powered by next-generation AMD EPYC processors, codenamed “Genoa” featuring the “Zen 4” processor core
A new approach using accelerator-centric compute blades (in a 4:1 GPU to CPU ratio, connected by the 3rd Gen AMD Infinity Architecture for high-bandwidth, low latency connections) to increase performance for data-intensive AI, machine learning, and analytics needs by offloading processing from the CPU to the GPU.
