El Capitan: world's fastest exascale supercomputer debuts

Hewlett Packard Enterprise has announced the delivery of El Capitan, the world's fastest direct liquid-cooled exascale supercomputer, to the United States Department of Energy's Lawrence Livermore National Laboratory.

The El Capitan supercomputer, achieving 1.742 exaflops, is now recognised as the most powerful supercomputer globally. It is part of the only three exascale systems ever built, all by Hewlett Packard Enterprise.

El Capitan is also notable for being energy efficient, ranking among the top 20 in the Green500 list of the world's most energy-efficient supercomputers, with a performance of 58.89 gigaflops per watt.

Trish Damkroger, Senior Vice President and General Manager of HPC & AI Infrastructure Solutions at Hewlett Packard Enterprise, stated, "El Capitan marks another significant milestone in exascale supercomputing, bringing monumental performance, energy efficiency and the capabilities to accelerate AI-driven scientific discovery and make incredible breakthroughs to strengthen national security and unlock new opportunities in renewable energy." She added, "We are proud of this achievement, which resulted from a strong partnership and years-long research and development with the U.S. Department of Energy, National Nuclear Security Administration, Lawrence Livermore National Laboratory, and AMD. We look forward to future discoveries and engineering breakthroughs that El Capitan will enable."

El Capitan is set to bolster the United States' capabilities in maintaining national security and will assist the National Nuclear Security Administration's Tri-Labs – Lawrence Livermore National Laboratory, Sandia National Laboratories, and Los Alamos National Laboratory.

These enhancements will support efforts to ensure the safety and reliability of the nation's nuclear arsenal and aid in both current and future stockpile modernisation efforts. Lawrence Livermore National Laboratory plans to utilise AI models on El Capitan for a range of classified and unclassified needs.

Rob Neely, Weapon Simulation and Computing Program Director at Lawrence Livermore National Laboratory, commented, "We are pleased to have worked so successfully alongside our partners HPE and AMD over these last five years to deploy El Capitan. This long-anticipated resource will allow us to perform the high-fidelity 3D modelling and simulation we need to effectively carry out our national security mission. We have a large team of people who have been getting applications ready to unleash the power of this system – many of them with the help of HPE in our Centre of Excellence – and I can't wait to demonstrate the capabilities to our sponsors and supporters."

"We are also committed to investing in AI on El Capitan and will be using the system for large-scale AI training and inference to make our calculations faster, more efficient, and potentially more accurate. Having a capability that is dialed in for both modelling and simulation and AI workloads, all in one system, is very exciting."

In addition to its primary mission, El Capitan will support secondary missions aimed at enhancing national security, such as nuclear nonproliferation and counterterrorism. It will also be used in material discovery, nuclear data research, and high energy density science investigations, including research on inertial confinement fusion at the National Ignition Facility. The supercomputer's insights will also contribute to research in energy security, climate change, power grid modernisation, and drug discovery.

The architecture enabling El Capitan's performance is built on Hewlett Packard Enterprise's supercomputing solution, HPE Cray EX, which features AMD Instinct MI300A APUs. These integrate both CPU and GPU cores with high-bandwidth memory, supported by the HPE Slingshot interconnect and a custom storage solution. HPE Slingshot, an Ethernet-based high-speed fabric, supports large-scale calculations across the system's more than 11,000 nodes. As part of a public-private partnership, Lawrence Livermore National Laboratory and Hewlett Packard Enterprise have developed a near-node local storage solution to reduce latency, which is dynamically configurable and tiered to a global Lustre-based file system.

El Capitan's design also emphasises energy efficiency through the use of a high-density, 100% fanless direct liquid cooling system architecture.