AMD announces a partnership with HPE (HP Enterprise) and the National Nuclear Security Administration US (NNSA) which will lead to the construction of the new one El Capitan supercomputer, a system that once completed (2023) will be the most powerful in the world with a calculation capacity of no less than 2 ExaFLOPS (2×1018) in double precision floating point operations.
EL Capitan will be based exclusively on AMD technology, will use AMD EPYC "Genoa" server processors is Next Generation Radeon Instinct GPU in tandem with the heterogeneous calculation software AMD ROCm (open source). EL Capitan will be the second US-based exascale system based on the AMD platform afterwards Frontier, 1.5 ExaFLOPS supercomputer created by Cray (acquired in May 2019 from HPE) for the Oak Ridge National Laboratory.
The new AMD platform will be optimized for areas such as artificial intelligence and machine learning, especially for greater integration of these into NNSA missions in areas such as research, computational techniques and data analysis.
At the base of the new AMD node there will be an EPYC "Genoa" processor, a solution that will use cores with Zen 4 architecture and a 7nm ++ production process. Each node then provides 4 Radeon Instinct cards, GPUs optimized for AI and ML that will take advantage of a renewed architecture and new generation HBM (High Bandwidth Memory) memories. CPU and GPU interface using the architecture AMD Third Generation Infinity, a solution that will allow you to reduce latencies while ensuring high bandwidth.
El Capitan is preparing to set a new record and represents a significant step forward for AMD in the HPC sector, a goal particularly underlined by Forrest Norrod, vice president and General Manager Datacenter and Embedded Systems Group for AMD:
Powered by next-generation AMD EPYC CPUs and Radeon Instinct GPUs, El Capitan will bring unprecedented progress in areas such as HPC and AI; AMD is enabling the NNSA Tri-Lab community to achieve their mission-critical goals and contribute to new advances in artificial intelligence in the industry. We are extremely proud to continue our exascale work with HPE and NNSA and look forward to delivering the most powerful supercomputer in the world scheduled for early 2023.
Related posts:
7-nanometer Nvidia GPU, TSMC will handle most of the production 
ASRock X299, a new BIOS allows you to install 2 TB of memory 
MSI Prestige X570 Creation Review: Test | Specs | Hashrate 
Radeon RX 5500 XT, PCI Express 3.0 castrates performance? 
An overclocker ran 1TB of RAM on an X299 motherboard limited to 256GB 
Best Review 2021: MSI MPG X570 Gaming Edge WiFi Under $250 ($200)
