For the third time in a row, Frontier is ranked number one among the world’s fastest supercomputers, and it remains the only whose fastest speed exceeds one exaFLOPS.
At 1.194 quintillion floating point operations per second (FLOPS), Frontier kept its ranking with more than double the top speed of its nearest competitor, according to the list compiled by TOP500, which issues the rankings twice a year. A quintillion is 1018 or one exaFLOPS (EFLOPS).
The number two machine, Fugaku, maxed out at 442.01petaFLOPS. A petaFLOPS is 1015 FLOPS.
Two competitors in the top 10 improved their speeds since the last ranking period that ended in November 2022, but not nearly enough to even draw close. Those two—LUMI and Leonardo—rank third and fourth, respectively.
The 10 fastest on the list are the same ones as last time and rank in the same order.
The speed measurement used in evaluating the computers is known as the High Performance Linpack (HPL) benchmark, which measures how well systems solve a dense system of linear equations.
Half of the top 10 are deployed in the US. Two are in China, and there's one each in Finald, Italy, and Japan.
In addition to ranking number one for flat-out speed, Frontier also ranked number one in suitability for doing artificial intelligence (AI) functions, a classification known as HPL-MxP.
Frontier and LUMI were also among the top 10 for power efficiency as measured by GFLOPS per Watt. Frontier came in at number six with 52.592GFLOPS/W, and LUMI came in number seven with 51.382GFLOPS/W.
The most power efficient overall was Henri, operated by the Flatiorn Institute of New York City, with 65.396 GFLOPS/W.
Here are the 10 overall fastest on the TOP500 list for June 2023 and where they are located:
#1—Frontier
An HPE Cray EX system run by the US Department of Energy, Frontier incorporates 3rd Gen AMD EPYC™ CPUs representing 8,730,112 cores that have been optimized for high-performance computing (HPC) and AI with AMD Instinct™ 250X accelerators and Slingshot-11 interconnects. Its HPL benchmark was 1.194EFLOPS.
#2—Fugaku
Supercomputer Fugaku, housed at the RIKEN Center for Computational Science in Kobe, Japan, scored 442.01PFLOPS in the HPL test. It is built on the Fujitsu A64FX microprocessor and has 7,630,848 cores.
#3—LUMI
LUMI is an HPE Cray EX system at the EuroHPC center at CSC in Kajaani, Finland, with a performance of 309.1 PFLOPS. It relies on AMD processors and boasts 2,220,288 cores.
#4—Leonardo
Leonardo, which resides in Bolbogna, Italy, is an Intel/Nvidia system with 1,463,616 cores and a maximum speed of 238.7PFLOPS.
#5—Summit
An IBM system at Oak Ridge National Laboratory in Tennessee, Summit scored 148.6PFLOPS on the HPL benchmark. It has 4,356 nodes, each with two Power9 CPUs with 22 cores and six Nvidia Tesla V100 GPUs, each with 80 streaming multiprocessors (SM). The nodes are linked by a Mellanox dual-rail EDR InfiniBand network. It has 2,414,592 cores.
#6—Sierra
Similar in architecture to Summit, Sierra reached 94.64 PFLOPS. It has 4,320 nodes with two Power9 CPUs and four Nvidia Tesla V100 GPUs and a total of 1,572,480 cores. It is housed at the Lawrence Livermore National Laboratory, in California.
#7—Sunway TaihuLight
Sunway TaihuLight is a machine developed by National Research Center of Parallel Computer Engineering & Technology (NRCPC) in China and is installed in the city of Wuxi. It reached 93.01PFLOPS on the HPL benchmark. It has 10,649,600 cores.
#8—Perlmutter
The Perlmutter system is based on the HPE Cray Shasta platform and is a heterogeneous system with both AMD EPYC-based nodes and 1536 Nvidia A100-accelerated nodes. It has 761,856 cores. It achieved 70.87 PFLOPS. That’s an improvement of about 6PFLOP/s over last year’s score, but still not enough to catch Sunway TaihuLight.
#9—Selene
Selene is an Nvidia DGX A100 SuperPOD based on an AMD EPYC processor with Nvidia A100s for acceleration and a Mellanox HDR InfiniBand as a network. It has 555,520 cores. It achieved 63.46 PFLOPS and is installed in-house at Nvidia facilities in the US.
#10—Tianhe-2A (Milky Way-2A)
Powered by Intel Xeon CPUs and NUDT’s Matrix-2000 DSP accelerators, Tianhe-2A has 4,981,760 cores in the system to achieve 61.44 PFLOPS. It was developed by China’s National University of Defense Technology (NUDT) and is deployed at the National Supercomputer Center in Guangzhou, China.