According to the November 2024 TOP500 rankings, Lawrence Livermore’s El Capitan is the most powerful supercomputer. In this list, derived from the TOP500 supercomputer ranking, Frontier and Aurora round out the top three slots.
These systems range from newly installed exascale powerhouses to long-established machines that are continuously evolving. All performance results come from the Linpack (HPL) benchmark, which measures performance in petaflops (PF/s) or exaflops (EF/s). A future addition, which could surpass El Capitan for first place, is the Colossus, a supercomputer currently under development by Elon Musk’s OpenAI competitor, xAI. At present, the Colossus cluster, which xAI uses to train its Grok AI models, has 100,000 Nvidia GPUs. The company plans to double that amount and eventually expand to one million or more GPUs.
1. El Capitan
Overview: El Capitan is installed at Lawrence Livermore National Laboratory (LLNL), became the third exascale system to reach over 1.7 exaflops on the Linpack benchmark.
Key Specs: Features AMD 4th Gen EPYC (24C @ 1.8 GHz) CPUs and AMD Instinct MI300A GPUs, totaling 11,039,616 cores and about 29,581 kW power consumption.
Highlights: Ranked #1 on the November 2024 TOP500 list and #18 on the GREEN500, achieving an energy efficiency of 58.89 GF/watt. The system was deployed in 2024.
2. Frontier
Overview: Frontier, operated by Oak Ridge National Laboratory, was the world’s first publicly disclosed exascale system and held the top spot on previous lists.
Key Specs: Uses AMD Optimized 3rd Gen EPYC (64C @ 2 GHz) and AMD Instinct MI250X accelerators, delivering a Linpack performance of 1.353 exaflops with 9,066,176 total cores.
Notes: Former TOP500 leader that increased its performance from 1.206 EF/s to 1.353 EF/s since the last ranking.
3. Aurora
Overview: Developed by Intel Corporation, Aurora is operated by Argonne National Laboratory. It is a next-generation exascale system designed for scientific research, engineering simulations, and data-intensive applications.
Key Specs: Uses Intel Xeon CPU Max 9470 (52C @ 2.4GHz) and Intel Data Center GPU Max, totaling 9,264,128 cores and about 38,698.36 kW power consumption.
Notes: Aurora was delayed for several years owing to changes in processor designs, but it now showcases cutting-edge Intel technology. Its architecture specializes in machine learning, making it a hybrid system for traditional and AI workloads.
4. Eagle
Overview: Located at the National Renewable Energy Laboratory (NREL), Eagle is the first cloud-based supercomputer to enter the top five. It facilitates large-scale AI model training and supports commercial research needs.
Key Specs: Features Xeon Platinum 8480C (48C @ 2GHz) CPUs and NVIDIA H100 GPUs, totaling 2,073,600 cores. (No public power data provided.)
5. HPC6
HPC6
Overview: HPC6 is designed for industrial applications, particularly oil and gas exploration, at Eni S.p.A. in Italy. It processes seismic data to locate energy resources precisely.
Key Specs: Features AMD Optimized 3rd Gen EPYC (64C @ 2GHz) CPUs and AMD Instinct MI250X GPUs, totaling 3,143,520 cores and about 8,460.90 kW power consumption.
6. Fugaku
Overview: Fugaku is Japan’s fastest supercomputer. It was built by Riken in partnership with Fujitsu and is housed at the Riken Center for Computational Science in Japan. It’s based on Fujitsu’s custom Arm A64FX processor and doesn’t use GPUs.
Key Specs: Uses A64FX (48C @ 2.2GHz) CPUs (no GPUs) with Tofu interconnect D, totaling 7,630,848 cores and about 29,899.23 kW power consumption.
Notes: Fugaku, named after Mount Fuji, is powered by energy-efficient ARM processors.
7. Alps
Overview: The Alps, located at Switzerland’s Swiss National Supercomputing Centre (CSCS), specializes in high-performance data analysis and scientific simulations. It is used in climate research, physics, and biology.
Key Specs: It uses NVIDIA Grace (72C @ 3.1GHz) CPUs and the NVIDIA GH200 Superchip, totaling 2,121,600 cores and consuming about 7,124.00 kW of power.
Notes: Alps is among the first to integrate NVIDIA Grace Hopper Superchips.
8. LUMI
Overview: Located in EuroHPC/CSC in Finland, Lumi is one of Europe’s most powerful supercomputers. It runs entirely on hydropower. Lumi supports climate research, genomics, and AI.
Key Specs: Uses AMD Optimized 3rd Gen EPYC (64C @ 2GHz) CPUs and AMD Instinct MI250X GPUs, totaling 2,752,704 cores and about 7,106.82 kW power consumption.
Notes: LUMI’s excess heat is used to warm nearby homes.
9. Leonardo
Overview: Located at EuroHPC/CINECA in Italy, Leonardo is focused on large-scale simulations in astrophysics and material science. It combines AI capabilities with traditional computing for hybrid workflows.
Key Specs: Uses Xeon Platinum 8358 (32C @ 2.6GHz) CPUs and NVIDIA A100 SXM4 64 GB GPUs, totaling 1,824,768 cores and about 7,493.74 kW power consumption.
Notes: Named after Leonardo da Vinci.
10. Tuolumne
Courtesy of LLNL
Overview: Housed at Lawrence Livermore National Laboratory (LLNL) in Livermore, California. Tuolumne supports classified and unclassified research, focusing on multiphysics simulations. It complements the more powerful El Capitan in enhancing national security and scientific discovery.
Tell Us What You Think!