Local time on June 10, the international TOP500 organization released the 65th edition of the TOP500 list of supercomputers. The supercomputer systems based on the American AMD architecture have taken the dominant position in the top ten list, helping the supercomputer system El Capitan of Lawrence Livermore National Laboratory (LNNL) in California maintain its top spot in global supercomputer performance with a peak computing power of 1.742 EFlop/s.

Chinese supercomputers did not enter the top ten due to no longer updating data for the HPL benchmark test. Currently, the Sunway TaihuLight has dropped to 21st place, and Tianhe-2A has fallen to 31st place.

The following are the top ten lists of global supercomputer performance:

1. El Capitan ranks first in the HPL benchmark test with a peak computing power of 1.742 EFlop/s. Located at the Lawrence Livermore National Laboratory (LNNL) in California, USA, this supercomputer is based on the AMD fourth-generation EPYC processor (24 cores, 1.8 GHz) and AMD Instinct MI300A accelerator, with 11,039,616 computing cores, running under the HPE Cray EX255a architecture, using HPE Slingshot network interconnection technology, with an efficiency of 58.9 Gigaflop/ watt. Its calculation speed in the HPCG benchmark test (aimed at assessing the performance of high-performance computing systems in solving linear equation systems) also reached 17.41 Petaflop/s.

2. Frontier ranks second with an HPL performance of 1.353 EFlop/s. This supercomputer is located at Oak Ridge National Laboratory (ORNL) in Tennessee, USA, operated by the US Department of Energy (DOE). It adopts the HPE Cray EX235a architecture, 2GHz AMD EPYC 64C CPU, and AMD Instinct 250X GPU, with a total of 8,699,904 CPU and GPU cores, and uses HPE Slingshot 11 network for data transmission.

3. Aurora ranks third with an HPL performance of 1.012 EFlop/s. This supercomputer is installed at Argonne National Laboratory in Illinois, USA, and is also operated by the US Department of Energy. Aurora is based on the HPE Cray EX network and Slingshot interconnection technology, integrating 21,248 Intel Xeon Max series CPUs, 63,744 Intel Max series GPUs, and 20.42 PB memory.

4. JUPITER Booster ranks fourth with an HPL performance of 793.4 PFlop/s. This system is installed at the EuroPHC/FZJ in Jülich, Germany, operated by the Jülich Supercomputing Center. It is based on the Eviden BullSequana XH3000 direct liquid cooling architecture and adopts NVIDIA Grace Hopper superchips. Currently, the system is only partially operational, so this performance is not the full performance.

5. Eagle ranks fifth with an HPL performance of 561 PFlop/s. This supercomputer is installed on Microsoft Azure cloud platform, and it remains the highest-ranked cloud computing system in the Top500 supercomputers. The powerful performance of this Microsoft NDv5 system comes from the combination of Intel Xeon Platinum 8480C processors and NVIDIA H100 GPU accelerators.

6. HPC6 ranks sixth with an HPL performance of 477.9 PFlop/s. This supercomputer is installed at Eni SpA's center in Ferrara Erbonio, Italy, based on the HPE Cray EX235a system, integrating third-generation AMD EPYC CPUs optimized for HPC and AI, equipped with AMD Instinct 250X accelerators and HPE Slingshot interconnection technology.

7. Supercomputer Fugaku ranks seventh with an HPL performance of 442 PFlop/s. This supercomputer is installed at the RIKEN Center for Computational Science (R-CCS) in Kobe, Japan, based on the 2.2GHz Fujitsu A64FX 48C processor, Tofu interconnect D interconnect, with a total of 7,630,848 cores. Additionally, this supercomputer also ranks second fastest in the HPCG benchmark test with a performance of 16 Teraflop/s.

8. Alps ranks eighth with an HPL performance of 434.9 PFlop/s. This supercomputer is located at the Swiss National Supercomputing Center (CSCS), using the HPE Cray EX254n system, equipped with NVIDIA Grace 72C and NVIDIA GH200 superchips as well as Slingshot interconnection technology.

9. LUMI ranks ninth with an HPL performance of 380 PFlop/s. This supercomputer is installed at the EuroHPC center of CSC in Finland, based on the HPE Cray EX235a system, AMD third-generation EPYC 64C 2GHz processor, AMD Instinct MI250X accelerator, and Slingshot-11 network.

10. Leonardo ranks tenth with an HPL performance of 241.2 Petaflop/s. This supercomputer is installed at CINECA in Italy. It is a set of Atos BullSequana XH2000 systems, with Intel Xeon Platinum 8358 32C 2.6GHz processors, and equipped with NVIDIA A100 SXM4 40 GB, with an interconnection interface of four-track NVIDIA HDR100 InfiniBand.

The TOP500 ranking shows that AMD and Intel processors are the preferred choice for the top ten systems. Among them, four systems use AMD processors (El Capitan, Frontier, HPC6, and LUMI), three systems use Intel processors (Aurora, Eagle, and Leonardo). Alps and JUPITER Booster rely on NVIDIA Grace Hopper superchips, while Fugaku continues to use the proprietary Fujitsu A64FX ARM-based processor. Seven of the top ten computers (El Capitan, Frontier, Aurora, HPC6, Alps, LUMI, and JUPITER Booster) use Slingshot interconnection technology, while two (Eagle and Leonardo) use InfiniBand interconnection technology. Fugaku retains its proprietary Tofu interconnection technology.

Although China and the United States remain the countries with the most participants in the TOP500 list, the number of Chinese participants continues to decline. The United States added two new systems, bringing its total number of participating systems to 173. The number of Chinese participating systems decreased from 63 to 46, and like the previous list, no new systems were added. Germany continues to narrow the gap, currently with 43 systems listed. By continent, North America leads with 187 systems, followed by Europe (163 systems) and Asia (135 systems).

This article is an exclusive contribution from Guancha Observer and cannot be reprinted without permission.

Source: https://www.toutiao.com/article/7514935221150679571/

Disclaimer: The views expressed in this article are those of the author and welcome your opinions to be expressed in the buttons below.