El Capitan remains world’s fastest as HPE leads supercomputer race
El Capitan has retained its position as the fastest supercomputer in the world according to the latest TOP500 list, marking the eighth consecutive time a system built by Hewlett Packard Enterprise (HPE) has held the top spot. The list, updated twice yearly, also recorded that HPE-built exascale supercomputers occupy the first, second and third places for the third time in succession, underscoring the firm's influence in the high-performance computing (HPC) sector.
Fastest systems
El Capitan, located at Lawrence Livermore National Labouratory, achieved 1.809 exaflops, surpassing its previous benchmarks by 4%. The system also leads the High-Performance Conjugate Gradient and HPL-MxP benchmarks, measuring performance in scientific applications and mixed-precision AI computations. El Capitan is ranked 23rd on the Green500 list for energy efficiency, and utilises AMD Instinct MI300A APUs.
Frontier, based at Oak Ridge National Labouratory, maintains its status as the world's second fastest system with 1.353 exaflops. It also ranks third for mixed-precision calculations crucial for AI and HPC workloads, with a performance level of 11.4 exaflops. Powered by AMD Instinct MI250 GPUs and AMD EPYC CPUs, Frontier continues to contribute to research in fields including quantum physics and renewable energy.
Aurora, built for Argonne National Labouratory in partnership with Intel, delivers 1.021 exaflops. It placed second in the HPL-MxP mixed-precision benchmark with 11.6 exaflops, advancing AI and data-driven research capabilities.
Broader performance
Other HPE-built systems in the global top ten include HPC6, Alps, and LUMI. HPC6, developed for Eni, provides 477.9 petaflops to support industrial processes and decarbonisation research. Alps, operated by the Swiss National Supercomputing Centre, delivers 434.9 petaflops for research in medicine, quantum chemistry, and climatology. LUMI, serving the EuroHPC Joint Undertaking, provides 379.7 petaflops for climate modelling and neural network training across European languages.
Energy efficiency
HPE's Cray systems have a long tradition of focusing on energy efficiency using direct liquid cooling. The Green500 list, which measures power efficiency, now includes ten HPE systems in its top 20. These include Isambard-AI at the University of Bristol, the SSC-24 Energy Module for Samsung Electronics, and Helios GPU in Poland, among others. HPE's 100% fanless direct liquid cooling architecture is central to these developments.
Multiple HPE systems have set regional records. Isambard-AI is part of the UK Government's AI Research Resource, supporting research into robotics, big data, and climate science. Helios GPU is the fastest system in Poland, and Shaheen III - GPU at King Abdullah University of Science and Technology marks the most powerful system in the Middle East.
Regional advances
Among systems new to the TOP500, TELUS's sovereign AI factory is now the most powerful supercomputer in Canada, delivering 22.74 petaflops. In Kazakhstan, Alem.Cloud's system claims the top regional position, ranking at number 86 on the global list and offering 20.48 petaflops for government, academic and commercial users as well as supporting the development of domestic AI tools.
Scientific impact
HPE systems are used in several research projects nominated for the ACM Gordon Bell Prize. El Capitan supports high-fidelity simulations for earthquake and tsunami modelling. The Frontier and Alps systems were utilised by ETH Zurich to simulate nanoribbon transistor behaviour at the quantum level, while Alps also powered global climate simulations at sub-2 kilometre resolutions. Aurora enabled simulations of light interactions in quantum materials, handling systems of over one trillion atoms. LUMI was integral to Climate DT's multi-decade global simulations, which scaled to one-kilometre resolutions for various climate system components.
Other research supported by HPE systems includes studies of protein folding with El Capitan, earthquake simulation for infrastructure improvement on Frontier, and battery technology innovation with Aurora.
Open source software
The High Performance Software Foundation (HPSF) has welcomed the Chapel project as its newest member. Originally created by Cray and now managed by HPE, Chapel is an open-source initiative aimed at simplifying programming for parallel computers. Through its inclusion in the Linux Foundation's broader governance model, Chapel is expected to attract new users and contributors from the global HPC community.
"Chapel's entry into the High Performance Software Foundation represents an important step in fostering community-driven open source development for high-performance computing applications and making parallel programming more accessible," said Jim Ang, Corporate Vice President, HPC and AI, Hewlett Packard Enterprise.
