Lawrence Livermore National Laboratory's (LLNL) El Capitan once again claimed the top spot on the Top500 List of the world's most powerful supercomputers, announced today at the 2025 International Conference for High Performance Computing, Networking, Storage and Analysis (SC25) conference in St. Louis.
The National Nuclear Security Administration's (NNSA's) first exascale supercomputer, El Capitan - developed in partnership with HPE and AMD - retained its No. 1 ranking with a verified 1.809 exaFLOPs (quintillion calculations per second) on the organization's High Performance Linpack (HPL) benchmark. The result marks a modest gain from the previous list released in June, reflecting newly identified optimization opportunities, and reaffirms El Capitan's standing as the fastest computer ever verified and the pinnacle of exascale computing.
The machine is funded by NNSA's Advanced Simulation and Computing (ASC) program and is dedicated to national security, performing critical calculations to ensure the safety, security and reliability of the nation's nuclear deterrent.
El Capitan continued to lead other major industry metrics, again earning the "triple crown" of performance with No. 1 rankings on the High Performance Conjugate Gradients (HPCG) benchmark at 17.41 petaFLOPs and the HPL Mixed Precision (MxP) benchmark at 16.7 exaFLOPs. The machine also placed an impressive 23rd on the Green500 list at 60.94 gigaFLOPs per watt, demonstrating its ability to combine unmatched performance with excellent energy efficiency across a range of high-end simulation and AI workloads. El Capitan's theoretical peak of 2.88 exaFLOPs cements the Department of Energy's leadership in exascale computing - with Oak Ridge National Laboratory's Frontier and Argonne National Laboratory's Aurora again rounding out the top three - and efficient performance.
"Since debuting atop the Top500 list a year ago, El Capitan has consistently pushed the boundaries of scientific discovery, excelling in both advanced modeling and simulation and AI workloads," said Rob Neely, associate director for Weapon Simulation & Computing at LLNL. "This year, it has elevated our national security mission, serving as a capstone achievement for the ASC program's exascale goals. El Capitan's success is a direct result of the dedication and expertise of our facilities, operations and code development teams, working seamlessly alongside our valued partners at HPE and AMD. Together, we have set a new standard for what's possible in high-performance computing."
Built to serve NNSA's stockpile stewardship mission as a shared resource for LLNL and the Los Alamos and Sandia national laboratories, El Capitan delivers more than 20 times the speed of LLNL's recently retired Sierra system. It combines HPE Cray EX architecture with over 46,000 AMD Instinct Accelerated Processing Units (APUs), linked by HPE Slingshot interconnect, to achieve exceptional scalability and throughput. Its unprecedented speed and precision enable scientists to perform full-scale, 3D simulations faster and at higher resolution, accelerating time-to-solution for critical national security applications.
In addition to El Capitan, LLNL's Tuolumne - a sibling system to El Capitan sharing the same architecture - repeated at No. 12 on the Top500 at 208.1 petaFLOPs on the HPL, highlighting the Lab's leadership in deploying next-generation architectures. Tuolumne brings El Capitan-class technology into the Lab's unclassified environment, providing a peak of 288 petaFLOPs of performance for open science research and enabling researchers to port, optimize and validate applications for El Capitan.
LLNL placed 11 additional systems on the Top500 - with the Lab's 13 total machines representing the most of any supercomputing facility in the world - including long-standing platforms Sierra (23rd) and Lassen (99th), as well as RZAdams (64), Bengal (218) and Dane (220), and El Capitan early access systems RZVernal (265), Tioga (290) and Tenaya (368). Several other advanced computing clusters supporting the Lab's AI, materials science, bioscience and energy research portfolios also made the list, including the debut of CTS-2 Matrix (408). Additionally, RZAdams placed 16th on the Green500, which ranks supercomputers by computational performance on the HPL benchmark per watt of power consumed. Tuolumne was 19th on the list.
Taken together, this fleet forms one of the most capable and varied supercomputing ecosystems in the world, allowing LLNL scientists and engineers to match workloads to the right scale of resources and move innovations rapidly from concept to full-scale simulation.
