El Capitan Supercomputer

El Capitan is the supercomputer of an eksaflopsny class.

2020: Appointment and description of El Capitan

As reported on March 4, 2020 in Hewlett Packard Enterprise (HPE) company, the El Capitan supercomputer created for National administration of nuclear security (NNSA) at the U.S. Department of Energy will have performance in 2 Eksaflops. In more detail about the project of creation of the supercomputer read here.

Purpose of the supercomputer

A system will work based on Livermore National Laboratory of E. Lawrence (LLNL) and to be used by three national laboratories NNSA: LLNL, Sandiysky national laboratories and Los Alamos national laboratory.

It is expected that a system will allow to carry out difficult simulations and to make modeling for ensuring nuclear security of the USA. According to the company, HPE optimizes El Capitan under difficult and costly on time research 3D - modeling for tasks of NNSA with which modern hi-tech supercomputers do not cope. El Capitan will give to scientists an opportunity to use applications using modern workloads with hard work with data, such as modeling, simulation, analytics and work with the artificial intelligence (AI).

Researchers of LLNL are going to use a system for studying of applications which integrate artificial intelligence and machine learning into workloads of high-performance computing.

At the expense of what performance in 2 eksaflops is reached

As explained in HPE, systems similar to El Capitan represent the class Eksaflopsnykh supercomputers which are one thousand times faster than the systems of the PetaFLOPS class of last generation which for the first time were provided 12 years ago. Performance in 2 Eksaflops (2,000 PetaFLOPS) was will reach thanks to partnership between HPE, AMD and the U.S. Department of Energy which allows to integrate systems capabilities of HPE Cray Shasta and interconnect of Slingshot with opportunities processors AMD EPYC and the graphic processors AMD Radeon Instinct of the next generation.

Joint efforts of HPE and AMD created the technologies designed to support workloads of high-performance computing and AI due to the following improvements:

• The optimized interaction between interconnect of HPE Cray Slingshot and the graphic processors AMD Radeon Instinct of the next generation which architecture was optimized for NRS-and AI-workloads;
• High density computing "edges" (blades) based on powerful processors of the next generation of AMD EPYC code-named "Genoa" with cores of Zen 4;
• The approach using "the accelerator in the center" (accelerator-centric) computing "edges" (with a ratio of number of graphic processors to central 4:1 and AMD Infinity Architecture of third generation which will provide between them high-speed communication with low delays) allowing to increase performance for accomplishment it is information the capacious processes connected with AI, machine learning and analytics thanks to unloading of central processors using graphic.

Improvements also concerned the systems capabilities of data storage and software integrated into architecture of HPE Cray Shasta. Additional use of the local flash drives developed especially for ensuring necessary performance for the updated systems will provide the buffer for balancing of work of the memory set on a payment and multi-layer systems of data storage which will be monitored by the intelligent software solution Cray Shasta. It will allow to automate data movement for ensuring optimal storage and timely access.

In addition to El Capitan, HPE will create two more Eksaflopsny systems announced in 2019 for the Ministry of Energy: Aurora and Frontier.