El Capitan is based on HPE’s Cray EX supercomputer platform and powered by AMD’s Instinct MI300A APUs. It achieved a maximum performance of 1.7 exaflops or 1.7 million teraflops on the latest biannual Top500 list, which ranks the world’s supercomputers.
Having narrowly avoided being overtaken by Aurora in the last Top500 list, Frontier fell to second spot, achieving a maximum performance of just 1.3 exaflops.
Subscribe today for free
El Capitan is being funded by the National Nuclear Security Administration’s Advanced Simulation and Computing (ASC) programme and is being used by the US Department of Energy to check the safety and reliability of the US’ nuclear stockpile.
It’s projected to eventually break the 2 exaflop barrier, and requires around 30 megawatts of energy to run at peak — enough power to run a mid-size city.
El Capitan is also being used to power AI model training capabilities, including classified AI projects.
The supercomputer was greenlit back in 2019, when the Department of Energy signed off on the $600 million project.
Prior to it coming online this year, three prototypes of the supercomputer — rzVernal, Tioga, and Tenaya — were powerful enough that they ranked on the Top500 list in June 2023.
“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," said Rob Neely, weapon simulation and computing programme director at the at Lawrence Livermore National Laboratory (LLNL).
Despite El Capitan displacing Frontier, the latest list proved a win for HPE and AMD as both systems leverage hardware from the vendors.
“Leveraging the AMD Instinct MI300A APUs, we've built a system that was once unimaginable, pushing the absolute boundaries of computational performance while maintaining exceptional energy efficiency,” said Bronis R. de Supinski, chief technology officer for Livermore Computing at Lawrence Livermore National Laboratory.
Frontier had held the top spot since June 2022 and narrowly avoided being overtaken by Aurora back in May.
The latest list sees a major shakeup of the top 10. With El Capitan shunting to the front, Aurora, which is also based on HPE’s Cray EX but uses Intel’s Max Series GPUs, was pushed down to third.
Notable top changes saw Fugaku pushed out of the top five for the first time, with Microsoft’s Eagle, the highest-ranked cloud-based system and the new HPC6 from energy provider Eni making up the new top five.
Fujitsu’s Fugaku does, however, retain its position as the most powerful supercomputer in Asia Pacific, being the only representative from the region in the top 10. The Japanese supercomputer did, however, maintain its grip as the leader on the HPCG benchmark with 16 PFlop/s, a position it has held since June 2020.
The rest of the top supercomputers are European-based, including the Swiss system Alps, Lumi from Finland, and Italy’s Leonardo.
Rounding out the top 10 is another HPE and AMD-based system, Tuolumne, a sister system to El Capitan, which is being used to power unclassified projects in energy security, climate change, and cancer drug discovery.
The latest Top500 list was dominated by supercomputing units housing processors from AMD
Of the top 10, five supercomputers use AMD processors, up from two (El Capitan, Frontier, HPC6, LUMI, Tuolumne), while three use Intel Xeons, down from five (Aurora, Eagle, Leonardo).
Fugaku uses custom hardware from Fujitsu — the Arm-based A64FX, while Alps was the only top 10 system powered by an Nvidia Grace 72C processor.
“At AMD, we are driving the future of computing with leadership performance and capabilities that will continue to define the convergence of HPC and AI for years to come,” said Forrest Norrod, EVP and general manager at AMD.
Other new entrants to the Top500 include CHIE-3 and CHIE-2, two SoftBank-owned supercomputers running on Nvidia and Intel hardware which entered at 18th and 17th, respectively.
Also new to the list was JUPITER, the supercomputer procured by EuroHPC, the EU’s supercomputing initiative, which came in at 18th place.
Summit, the highest-ranked IBM-based system dropped from 12th to 14th on this latest list, with it being the sole IBM system in the entire top 30.
China and the US dominated the list, however, the number of entries from China dropped again from 80 to 63 and failed to add any new machines to the TOP500 list.
The US added two systems to the list, bringing its total number of systems to 173. Meanwhile, Germany appears to be quickly catching up to China, with 41 supercomputers on the latest list.
The Green500: Performance vs. Power Consumption
Despite taking the top two spots, both El Capitan and Frontier came just numbers 18 and 22, respectively on the Green500 list, which ranks supercomputers based on their energy efficiency.
JEDI once again took the gong for the greenest supercomputer in the world, while a fellow BullSequana XH3000-based system came second, ROMEO-2025.
Isambard-AI, the UK-government-backed supercomputer had been the second-greenest last time out but fell to fourth.
Adastra 2, the HPE and AMD-based unit operated by GENCI, France’s high-performance computing agency, rounded out the top three.
In addition to its fall on the green list, Isambard-AI fell from position 128 to 155 on the power rankings. However, only Phase One was tested, with the system still being expanded.
Of the top 10 most powerful, it was Alps which had the highest Green500 ranking, coming in at 14th place.
RELATED STORIES
HPE’s new AI servers tap Nvidia, AMD, Intel GPUs for supercomputer flexibility
Swiss supercomputer Alps officially unveiled to power European AI research
Musk to double xAI’s Colossus cluster to 200K GPUs for Grok 3 training
