The new El Capitan system at the Lawrence Livermore National Laboratory in California made its debut at No. 1 on the latest TOP500 list of the world’s most powerful supercomputers, released today. The Frontier system at Oak Ridge National Laboratory in Tennessee moved down to the No. 2 spot, ceding the top spot it held for the last five rankings.
With more than 11 million CPU and GPU cores, El Capitan achieved 1.742 Exaflop/s on the HPL benchmark. It’s based on AMD 4th generation EPYC processors with 24 cores at 1.8GHz and AMD Instinct MI300A accelerators.
Frontier achieved an HPL score of 1.353 Exaflop/s. It’s equipped with AMD 3rd generation EPYC 64C 2GHz processors.
There are now three Exascale systems leading the TOP500: El Capitan, Frontier, and Aurora, the third-fastest system with a preliminary HPL score of 1.012 Exaflop/s. All three Exascale systems are installed at Department of Energy laboratories in the US.
The Eagle system installed on the Microsoft Azure cloud in the US claimed the No. 4 spot and remains the highest-ranked cloud-based system on the TOP500 with an HPL score of 561.2 PFlop/s. At No. 5 is the HPC6 system in Ferrera Erbognone, Italy – it’s now the fastest system in Europe.
In this 64th edition of the TOP500, six of the top 10 systems are manufactured by HPE. There’s one Intel system, one Microsoft Azure, one Fujitsu and one Eviden. Seven of the computers on the top 10 use the Slingshot-11 interconnect, while two use Infiniband. One has its own proprietary interconnect.
Here’s a breakdown of specific details for the 10 fastest supercomputer systems on the TOP500 list for November 2024:
No. 1: El Capitan
- Location: Lawrence Livermore National Laboratory, Livermore, Calif.
- HPL score: 1.742 Exaflop/s
- System model: HPE Cray EX255a
- Processors: AMD 4th generation EPYC processors with 24 cores at 1.8 GHz and AMD Instinct MI300A accelerators
- Total core count: 11,039,616
- Interconnect: Slingshot 11
- Energy efficiency: 58.89 Gigaflops/watt
No. 2: Frontier
- Location: Oak Ridge National Laboratory in Oak Ridge, Tenn.
- HPL score: 1.353 Exaflop/s
- System model: HPE Cray EX235a
- Processors: 3rd Gen AMD EPYC CPUs optimized for HPC and AI with AMD Instinct 250X accelerators
- Total core count: 9,066,176
- Interconnect: Slingshot 11
- Energy efficiency: 54.98 Gigaflops/watt
No. 3: Aurora
- Location: Argonne Leadership Computing Facility, Argonne, Ill.
- HPL score: 1.012 Exaflop/s
- System model: HPE Cray EX – Intel Exascale Compute Blades
- Processors: Intel Xeon CPU Max Series processors and Intel Data Center GPU Max Series accelerators
- Total core count: 9,264,128
- Interconnect: Slingshot 11
- Energy efficiency: 26.15 Gigaflops/watt
No. 4: Eagle
- Location: Microsoft Azure cluster
- HPL score: 561 Petaflop/s
- System model: Microsoft NDv5
- Processors: Xeon Platinum 8480C processors and Nvidia H100 accelerators
- Total core count: 2,073,600
- Interconnect: Nvidia Infiniband NDR
- Energy efficiency: unknown
No. 5: HPC6
- Location: Eni S.p.A center in Ferrera Erbognone in Italy
- HPL score: 477.9 Petaflop/s
- System model: HPE Cray EX235a
- Processors: 3rd Gen AMD EPYC CPUs optimized for HPC and AI with AMD Instinct MI250X accelerators
- Total core count: 3,143,520
- Interconnect: Slingshot 11
- Energy efficiency: 56.48 Gigaflops/watt
No. 6: Fugaku
- Location: RIKEN Center for Computational Science (R-CCS) in Kobe, Japan
- HPL score: 442 Petaflop/s
- System model: Supercomputer Fugaku, Fujitsu cluster
- Processors: A64FX 48C 2.2GHz
- Total core count: 7,630,848
- Interconnect: proprietary Tofu interconnect D
- Energy efficiency: 15.42 Gigaflops/watt
No. 7: Alps
- Location: Swiss National Supercomputing Centre (CSCS) in Switzerland
- HPL score: 434.9 Petaflop/s
- System model: HPE Cray EX254n
- Processors: Nvidia Grace 72C 3.1GHz and Nvidia GH200 Superchip
- Total core count: 2,121,600
- Interconnect: Slingshot 11
- Energy efficiency: 61.05 Gigaflops/watt
No. 8: LUMI
- Location: EuroHPC center at CSC in Finland
- HPL score: 380 Petaflop/s
- System model: HPE Cray EX235a
- Processors: AMD Optimized 3rd Generation EPYC 64C 2GHz and AMD Instinct MI250X accelerators
- Total core count: 2,752,704
- Interconnect: Slingshot 11
- Energy efficiency: 53.43 Gigaflops/watt
No. 9: Leonardo
- Location: EuroHPC site in CINECA, Italy
- HPL score: 241.2 Petaflop/s
- System model: Atos BullSequana XH2000
- Processors: Xeon Platinum 8358 32C 2.6GHz processors and Nvidia A100 SXM4 64 GB accelerators
- Total core count: 1,824,768
- Interconnect: Quad-rail Nvidia HDR100 Infiniband
- Energy efficiency: 32.19 Gigaflops/watt
No. 10: Tuolumne
- Location: Lawrence Livermore National Laboratory, Calif.
- HPL score: 208.1 Petaflop/s
- System model: HPE Cray EX255a
- Processors: AMD 4th Gen EPYC 24C 1.8GHz and AMD Instinct MI300A accelerators
- Total core count: 1,161,216
- Interconnect: Slingshot 11
- Energy efficiency: 61.45 Gigaflops/watt
TOP500 history and highlights
The first iteration of the TOP500 list was created for a small conference in Germany in June 1993. The authors decided to continue compiling the list, which is now a twice-yearly event. Some highlights from the November 2024 TOP500 list include:
Nvidia accelerators: A total of 211 systems on the list are using accelerator/co-processor technology, up from 194 six months ago. Of these, 72 use Nvidia Ampere chips, 60 use Nvidia Hopper, and 33 systems use Nvidia Volta.
Intel vs. AMD: Intel continues to provide the processors for the largest share (61.8%) of TOP500 systems, down from 63% six months ago. Meanwhile, 162 (32.4%) of the systems in the current list use AMD processors, up from 31.2% six months ago.
Geographic distribution: The US is home to 173 of the top 500 systems. China has 63, Germany has 41, Japan has 34, and France has 24 systems. Rounding out the countries with the highest numbers of TOP500 systems are the UK (14), South Korea (13), Italy (13), Netherlands (10), Canada (9) and Brazil (9). The TOP500 authors made this observation:
“While China and the United States were once again the countries that earned the most entries on the entire TOP500 list, it would appear that China is not participating to the extent that it once did. The United States added two systems to the list, bringing its total number of systems to 173. China once again dropped its number of representative machines on the list from 80 to 63 systems. Like last list, China did not introduce any new machines to the TOP500 list. Germany is quickly catching up to China, with 41 machines on the list.”
HPC manufacturers: While six of the top10 systems are manufactured by HPE, Lenovo dominates the full TOP500 list. Lenovo is the manufacturer for 162 systems, followed by HPE with 115. Rounding out the top 10 HPC manufacturers are Eviden (52), Dell (37), Nvidia (26), Fujitsu (15), NEC (14), Inspur (11), Microsoft Azure (8) and Penguin Computing (7).
Source:: Network World