Japan’s Fugaku supercomputer has retained its position as the number one computer system, followed by IBM’s Summit and Sierra in the U.S.
In the 56th edition of the TOP500, Japan, USA, Germany, Italy and France have dominated the top 100 of the world’s most powerful supercomputers. Japan’s Fugaku supercomputer retained its position as the number one computer system, followed by IBM’s Summit and Sierra in the U.S.
The Top500 project tracks the most powerful supercomputers in the world and is published biannually. The non-profit organization released its November 2020 results last week. This year, while two new systems made it to the top 10, the full list recorded the smallest number of new entries since the project’s inception in 1993.
As a result, Top500 representatives suggested that the current list reflects a flattening performance growth curve.
What the results show
The world’s supercomputers are now not only faster, but also smarter, and support a greater variety of workloads.
Nearly 70% (140) of the systems on the TOP500 list, including eight of the top 10, are powered by NVIDIA technology and are “increasingly using AI to help researchers make discoveries faster.”
NVIDIA’s end-to-end HGX AI supercomputing platform helps accelerate scientific computing, data analytics, and AI workloads.
“The shift to incorporating AI into HPC (High-Performance Computing), and a platform that extends beyond traditional supercomputing centers, represents a significant change in a field that, since Seymour Cray’s CDC 6600 was launched in 1964, has focused on harnessing ever larger, more powerful machines for compute-intensive simulation and modeling,” NVIDIA said in a blog post.
In October this year, Italian inter-university consortium CINECA had announced its plans to build the world’s fastest AI supercomputer. The new “Leonardo” system is expected to deliver 10 exaflops of FP16 performance in AI workloads. It will feature around 14,000 Nvidia A100 GPUs and around 3,500 Intel Sapphire Rapids CPUs.
In the case of a supercomputer, its performance capabilities are indicated using a standard rate called FLOPS or floating-point operations per second. This rate indicates the number of floating-point arithmetic calculations a system can perform per second. While a 1 petaFLOPS (PFLOPS) computer system can perform one quadrillion (10^15) floating-point operations per second, 1 exaFLOPS (EFLOPS) computer system can perform one quintillion (10^18) floating-point operations per second.
Meanwhile, Intel has remained the preferred processor provider, with 90% of the systems equipped with its Xeon or Xeon Phi chips.
With 212 systems on the list, China has exerted unequivocal domination over the list, followed by the U.S. with 113 systems, and Japan with 34. However, when it comes to aggregate performance, the U.S. continues to lead at 668.7 petaflops, compared to China’s 564.0 petaflops. Meanwhile, Japan has an aggregate performance of 593.7 petaflops.
Supercomputers, which have the unfortunate side effect of consuming enormous amounts of energy, are now also ranked on the Green500 by energy efficiency. On the Green500 list, the new NVIDIA DGX SuperPOD in the U.S. was found to be the most energy-efficient system. It was followed by MN-3 in Japan, and Atos-built JUWELS Booster Module in Germany.
Top 10 supercomputers
Notable developments in the top 10 include the addition of two new systems, and a new performance benchmark set by the Fugaku supercomputer.
Due to the addition of new hardware to beef up the system, Fugaku grew its High Performance LINPACK (HPL) benchmark to 442 petaflops, compared to the 416 petaflops the system achieved when it debuted in June 2020 – a 6.4% speed improvement.
The HPL benchmark, which indicates the performance of a dedicated system when solving a dense system of equations, is used to determine the Top500.
1. Fugaku, Japan
Built by Fujitsu, Fugaku is installed at the RIKEN Center for Computational Science (R-CCS) in Kobe, Japan. With its additional hardware, the system achieved a new world record of 442 petaflops result on HPL, making it three times ahead of the number two system in the list.
RIKEN’s director, Satoshi Matsuoka, stated that the improvement came as they were “finally being able to use the entire machine rather than just a good chunk of it.”
Since the June competition, his team has been able to fine-tune the code for maximum performance. “I don’t think we can improve much anymore,” Matsuoka said.
2. Summit, U.S.
Based at the Oak Ridge National Laboratory (ORNL) in Tennessee, Summit was built by IBM and is the fastest system in the US.. Launched in 2018, it has a performance of 148.8 petaflops and has 4,356 nodes, each one housing two 22-core Power9 CPUs and six NVIDIA Tesla V100 GPUs.
Recently, two teams working on Summit won the prestigious Gordon Bell Prize for outstanding achievement in high-performance computing, commonly referred to as the ‘Nobel Prize of supercomputing.’
3. Sierra, U.S.
A system at the Lawrence Livermore National Laboratory (LLNL) in California, Sierra has an HPL mark of 94.6 petaflops. With each of its 4,320 nodes equipped with two Power9 CPUs and four NVIDIA Tesla V100 GPUs, it has an architecture similar to that of Summit.
Sierra also made it to the 15th position on the Green500 List of the world’s most energy-efficient supercomputers.
4. Sunway TaihuLight, China
Installed at China’s National Supercomputing Center in Wuxi, Sunway TaihuLight previously held the Number 1 spot for two years (2016-2017). However, its rank has since fallen. While it was in third position last year, it has now slipped to fourth.
Built by China’s National Research Center of Parallel Computer Engineering & Technology (NRCPC), it achieved 93 petaflops on its HPL benchmark. It is powered exclusively by Sunway SW26010 processors.
5. Selene, U.S.
Installed in-house at NVIDIA Corp, Selene jumped to fifth position from seventh position in the June rankings. After a recent upgrade, Selene achieved 63.4 petaflops on HPL, nearly doubling its previous score of 27.6 petaflops.
NVIDIA unveiled Selene, its AI supercomputer, in June this year, after constructing and running it in less than a month. Its key uses include system development and testing, in-house AI workloads, and chip design work.
6. Tianhe-2A, China
With 61.4 petaflops, Tianhe-2A (aka MilkyWay-2A) climbed one spot to sixth position. Developed by China’s National University of Defense Technology (NUDT), it is installed at the National Supercomputer Center in Guangzhou.
Tianhe-2A is powered by Intel Xeon CPUs and NUDT’s Matrix-2000 DSP accelerators. It will be used for simulation, analysis, and government security applications. It held the Number 1 spot from June 2013 to November 2015.
7. JUWELS Booster Module, Germany
Atos-built JUWELS Booster Module is the newest entrant on the list. The BullSequana machine was recently installed at the Forschungszentrum Jülich (FZJ) in Germany and is the most powerful system in Europe, coming in at 44.1 HPL petaflops.
Based on a modular system architecture, JUWELS is powered by AMD processors and NVIDIA GPUs, similar to the Selene system.
8. HPC5, Italy
A Dell PowerEdge system, HPC5 was installed by the Italian company Eni S.p.A and is located inside Eni’s Green Data Center in Italy. One of the most powerful and sustainable computing systems in the world, HPC5 is used to research new sources of energy.
With a performance of 35.5 petaflops, it is the most powerful system in the list used for commercial purposes at a customer site. It uses Intel Xeon Gold CPUs and NVIDIA Tesla V100 GPUs.
9. Frontera, US
Installed at the Advanced Computing Center at the University of Texas at Austin in September year, Frontera is a Dell C6420 system equipped by Intel. Using 448,448 of its Intel Platinum Xeon cores, the system achieves 23.5 petaflops.
Frontera aids research across all areas of science including those related to quantum mechanics, drug design, eradicating emerging viruses, and the physics of black holes.
10. Dammam-7, Saudi Arabia
Dammam-7 is the second new addition at the top of the list. Installed at Saudi Aramco in Saudi Arabia, the HPE Cray CS-Storm systems uses Intel Gold Xeon CPUs and NVIDIA Tesla V100 GPUs. It achieves 22.4 petaflops and is the second commercial supercomputer in the top 10.