Fugaku Is Still the Fastest Supercomputer in the World

Fugaku Is Still the Fastest Supercomputer in the World

Fugaku was ranked number one on the June 2021 TOP500 supercomputers list. This is Fugaku’s third consecutive top spot and three times faster than any other.

Fugaku, a supercomputer created by Fujitsu remains number one on the TOP500 list of the fastest supercomputers worldwide, where it is still three-times faster than its nearest competitor.

The race for fastest is still tight. Perlmutter from the National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory (part of the US Department of Energy) has been added to the top 10. It is now at number 5 and moves up to numbers 6 through 10, which was published on November 2020. Dammam-7 was dropped from the top 10.

This list tracks the evolution of supercomputer architecture and performance since 1993. It is updated each June and November. These speeds are calculated using the High Performance Linpack benchmark which measures how systems solve dense linear equations.

The top 10 ranked countries in the most recent ranking include the United States, China, and Germany.

TOP500’s Top 10 Fastest Supercomputers, June 2021

Number 1: Fugaku

Fugaku – The system reached the same speed in November at 442 Pflop/s. It was constructed by Fujitsu and has 7,630 848 cores.

Number 2: Summit

Summit – Summit is the fastest system in America, scoring 148.8 Pflop/s. This was the same score as November last year. There are 4,356 nodes with two Power9 CPUs with 22 cores and six Nvidia TeslaV100 GPUs. Each GPU has 80 streaming multiprocessors. Each node is linked to each other by a Mellanox dual rail EDR InfiniBand network. The system was built by IBM and is located at Oak Ridge National Laboratory (ORNL), Tennessee.

Number 3: Sierra

Sierra – This system, which has an architecture similar to Summit, achieved 94.6 Pflop/s. It is the same result as the last time. . It is equipped with 4,320 nodes and two Power9 CPUs as well as four Nvidia Tesla V100 GPUs. It is located at the Lawrence Livermore National Laboratory, California.

Number 4: Sunway TaihuLight

Sunway TaihuLight-This system maintains its ranking with the exact score it achieved with its last ranking: 93 Pflop/s. Sunway TaihuLight uses Sunway SW26010 processors that have 10,649,600 cores. It was created by the National Research Center of Parallel Computer Engineering & Technology in China. It is currently installed at the National Supercomputing Center, Wuxi.

Number 5: Perlmutter

Perlmutter –  This is the lone new entry in the top 10. It’s based on the HPE Cray Shasta platform and is a heterogeneous system that includes both AMD EPYC-based and 1536 Nvidia-accelerated nodes. It reached 64.6 Pflop/s.

Number 6: Selene

Selene – This system, which was ranked five last time, is now a Nvidia DGX A100 SUPERPOD based upon an AMD EPYC CPU with Nvidia A100 acceleration and a Mellanox HDR InfiniBand network. It reached 63.4 Pflop/s, and was installed at Nvidia’s US facilities.

Number 7: Tianhe-2A

Tianhe-2A (Milky Way-2A), powered by Intel Xeon CPUs. It also has NUDT’s Matrix 2000 DSP accelerators. The system can achieve 61.4 Pflop/s. It was created by China’s National University of Defense Technology and deployed at Guangzhou’s National Supercomputer Centre.

Number 8: JUWELS booster module

BullSequana System Booster Module – This BullSequana system, which is 44.1 Pflop/s at its peak, is Europe’s most powerful. It uses AMD EPYC processors and Nvidia A100s acceleration. The network is similar to the Selene system. It was built by Atos and is located at the Forschungszentrum Juelich in Germany.

Number 9: HPC5

HPC5 – is a PowerEdge-system built by Dell. It achieves 35.5 Pflop/s using Nvidia Tesla V100s accelerators and Mellanox HDR InfiniBand network. It was installed by Eni S.p.A. in Italy.

Number 10: Frontera

Frontera – This Dell C6420 reached 23.5 Pflop/s with 448,448 Intel Platinum Xeon Cores. It resides at Texas Advanced Computing Center of University of Texas.

Leave a Reply

Your email address will not be published. Required fields are marked *