El Capitan (supercomputer)

El Capitan (supercomputer) is a pre-exascale supercomputer developed for the United States Department of Energy (DOE) and installed at Lawrence Livermore National Laboratory (LLNL) in Livermore, California. It is part of the DOE's CORAL-2 initiative and is designed to be one of the world's most powerful systems for open science and national security applications. The system is built by Hewlett Packard Enterprise (HPE) and utilizes advanced AMD processors and GPU accelerators.

Overview

El Capitan is a key component of the National Nuclear Security Administration's (NNSA) Advanced Simulation and Computing (ASC) program, which supports the Stockpile Stewardship Program. Its development was formally announced in 2019 as the successor to the Sierra (supercomputer) and Summit (supercomputer) systems. The supercomputer is named after the iconic El Capitan granite monolith in Yosemite National Park, continuing a tradition of naming Lawrence Livermore National Laboratory systems after California geographical features. Upon its completion, it is expected to significantly advance capabilities in computational physics and weapons science.

Development and construction

The contract for El Capitan was awarded to Hewlett Packard Enterprise and its partner AMD in 2020, following a competitive procurement process under the CORAL-2 collaboration, which also included Oak Ridge National Laboratory and Argonne National Laboratory. The system's design leverages the HPE Cray EX supercomputer architecture, a platform also used for the Frontier (supercomputer) and Aurora (supercomputer) systems. Major construction and integration phases took place at the LLNL facility throughout 2023 and 2024, with the goal of achieving full operational capability. The project involved extensive collaboration between computer scientists at HPE, AMD, and researchers from the NNSA laboratories.

Technical specifications

El Capitan is based on a heterogeneous CPU-GPU architecture, combining optimized AMD EPYC central processing units with next-generation AMD Instinct accelerator units. The system utilizes the Slingshot (interconnect) high-performance networking fabric developed by HPE Cray to connect its numerous nodes. It features a state-of-the-art liquid cooling system to manage the immense thermal load generated by its dense compute components. The overall design prioritizes energy efficiency and computational density to achieve its target performance within power constraints.

Performance and capabilities

El Capitan is designed to deliver sustained performance in the exascale computing realm, capable of executing over a quintillion calculations per second. In early benchmarks, its precursor nodes demonstrated exceptional performance on key applications from the HPCG Benchmark and LINPACK suites. The system's architecture is optimized for complex, multi-physics simulations involving hydrodynamics, radiation transport, and materials science. Its massive memory hierarchy and fast interconnect are intended to enable unprecedented fidelity in modeling physical phenomena.

Intended applications

The primary mission for El Capitan is to support the National Nuclear Security Administration in ensuring the safety, security, and reliability of the United States' nuclear weapons stockpile without underground testing. This involves running sophisticated 3D simulation codes like xRage and Kull for weapon physics. Furthermore, the system will be used for unclassified open science research in areas such as climate modeling, astrophysics, and bioinformatics, benefiting the broader scientific community. Its computational power is also expected to accelerate discovery in energy security and critical materials research. Category:Supercomputers Category:Lawrence Livermore National Laboratory Category:Computers introduced in 2024