TOP500 list update: for the first time, a supercomputer based on ARM processors became the leader

The 55th edition of the ranking of the world’s highest-performing supercomputers has been published.
Read about the new leaders of the list and the capabilities of extra-class supercomputers under the cut.


The previous leader of the list – the Summit supercomputer (OLCF-4) of the Oak Ridge National Laboratory – became the second, giving the honorable first place to the new Japanese top-system Fugaku, which showed a High Performance Linpack (HPL) result of 415.5 petaflops. This indicator exceeds the capabilities of Summit by 2.8 times. Fugaku is powered by Fujitsu’s 48-core A64FX SoC processor, making Japanese development the first-ever TOP500 system equipped with ARM processors. With single or lower accuracy, which is often used for machine learning and artificial intelligence tasks, Fugaku peak performance is more than 1000 petaflops (1 exaflops). The new system is installed at the RIKEN Computing Science Center (R-CCS) in Kobe, Japan.

The aforementioned Summit, a supercomputer created by IBM, shows a performance of 148.8 petaflops in the HPL test. The system has 4356 nodes, each of which is equipped with two 22-core Power9 processors and six NVIDIA Tesla V100 graphics accelerators. The nodes are connected by an InfiniBand EDR network. Summit remains the fastest supercomputer in the United States.

In third place was also an American – supercomputer Sierra Livermore National Laboratory. Lawrence (LLNL), California, showing 94.6 petaflops. Its architecture is very similar to Summit: it is equipped with two Power9 processors and four NVIDIA Tesla V100 graphics accelerators in each of the 4320 nodes. Sierra uses the same InfiniBand Mellanox EDR as Sunway TaihuLight, a supercomputer developed by the China National Parallel Computing and Technology Research Center (NRCPC). He, by the way, dropped to fourth place on the list. The system is entirely based on the 260-core Sunway SW26010 processors. Its HPL mark of 93 petaflops has remained unchanged since it was installed at the National Supercomputer Center in Wuxi, China, in June 2016.

In fifth place is also the Chinese development – Tianhe-2A (Milky Way-2A), implemented by China National University of Defense Technology (NUDT). Its HPL 61.4 petaflops performance is the result of a hybrid architecture using Intel Xeon processors and specially designed Matrix-2000 coprocessors. It is deployed at the National Supercomputing Center in Guangzhou, China.

The newcomer to the list, HPC5, finished in sixth place with HPL performance of 35.5 petaflops. HPC5 is a PowerEdge system created by Dell and implemented by the Italian energy company Eni S.p.A, making it the fastest supercomputer in Europe.

Another new system, Selene, is in seventh place with an HPL of 27.58 petaflops. Selene is installed on NVIDIA in the USA.

Frontera, the Dell C6420 system installed in the Texas Computing Center (TACC) in the United States, is eighth in the list. Its 23.5 HPL petaflops is achieved using 448.448 Intel Xeon cores.

The second Italian supercomputer in the top 10 is Marconi-100, it is installed in the CINECA research center. Marconi-100 runs on IBM Power9 processors and NVIDIA V100 graphics accelerators, its performance is 21.6 petaflops, it took the ninth place in the list.
The top 10 with a 19.6 petaflops index is completed by the Cray XC50 system installed at the Swiss National Supercomputer Center (CSCS) in Lugano. It is equipped with Intel Xeon processors and NVIDIA P100 GPUs.

The Russian development – the Christofari supercomputer based on Xeon Platinum, Nvidia DGX-2 and Tesla V100 – is gaining 6.67 petaflops in the HPL test, taking only 36th place so far.

Green500 Results

The most energy-efficient system on the Green500 list is the MN-3, based on the new server from Preferred Networks. The supercomputer reached a record high of 21.1 gigaflops / watt with a performance of 1.62 petaflops. The system has excellent energy efficiency thanks to the MN-Core chip, an accelerator optimized for matrix arithmetic. It takes 395 place in the TOP500 list.

In second place is the new NVIDIA Selene supercomputer, the DGX A100 SuperPOD, powered by the new A100 graphics accelerators. In third place is the NA-1 system, the PEZY Computing / Exascaler system installed at NA Simulation in Japan. The supercomputer reached 18.4 gigaflops / watt and is at position 470 in the TOP500.

Main trends

  • The list’s aggregate performance is now 2.23 exaflops, compared to just 1.65 exaflops just six months ago. The lion’s share of such rapid growth is the merit of the new Fugaku supercomputer, which took 1st place in the list.
  • The total number of new systems in the list is only 51, which is an unrecord from the very beginning of the creation of the TOP500 list (since 1993).
  • China continues to dominate the TOP500 in terms of the number of systems (226), the United States in terms of the number of supercomputers in the list is third (114), Japan is third (30) .;
  • A total of 144 systems from the list use accelerators or coprocessors. As before, most systems use NVIDIA graphics accelerators.
  • X86 continues to be the dominant processor architecture, present in 481 of the 500 systems. Intel is used on 469 of them, AMD is installed in 11, Hygon – in the remaining.
  • Chinese manufacturers dominate the list: Lenovo (180), Sugon (68) and Inspur (64) account for 312 out of 500 systems.

Similar Posts

Leave a Reply

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