Published 53 edition of the list of the most high-performance supercomputers

Submitted by 53th edition rating 500 most high-performance computers in the world. In the new issue, the top ten remained unchanged, with the exception of promotion to fifth place in the ranking of the new cluster Border, produced by Dell for the Texas Computer Center. The cluster runs CentOS Linux 7 and includes more than 448 thousand cores based on a Xeon Platinum 8280 28C 2.7GHz. The total size of RAM is 1.5 PB, and performance reaches 23 petaflops, which is 6 times less than the leader in the rating.

Leading cluster in the ranking Summit deployed by IBM at the Oak Ridge National Laboratory (USA). The cluster runs Red Hat Enterprise Linux, includes 2.4 million processor cores (using 22-core IBM Power9 22C 3.07GHz CPUs and NVIDIA Tesla V100 accelerators), which provide 148 petaflops of performance.

The American cluster takes second place Sierra, installed at the Livermore National Laboratory by IBM on the basis of a platform similar to Summit and demonstrating performance at 94 petaflops (about 1.5 million cores). In third place is the Chinese cluster Sunway TaihuLight, operating at the National Supercomputing Center of China, including more than 10 million computing cores and showing performance of 93 petaflops. Despite similar performance indicators, the Sierra cluster consumes half as much energy as the Sunway TaihuLight. In fourth place is the Chinese Tianhe-2A cluster, which includes almost 5 million cores and demonstrates a performance of 61 petaflops.

The most interesting trends:

  • The most powerful domestic cluster, Lomonosov 2, moved from 72nd to 93rd place in the ranking over the year. Cluster in Roshydromet dropped from 172 to 365 place. The Lomonosov and Tornado clusters, which ranked 227th and 458th a year ago, were pushed out of the list. The number of domestic clusters in the ranking over the year decreased from 4 to 2 (in 2017 there were 5 domestic systems, and in 2012 - 12);
  • Distribution by number of supercomputers in different countries:
    • China: 219 (206 - a year ago);
    • USA: 116 (124);
    • Japan: 29 (36);
    • France: 19 (18);
    • UK: 18 (22);
    • Germany: 14 (21);
    • Ireland: 13 (7);
    • Netherlands: 13 (9);
    • Canada 8(6);
    • South Korea: 5 (7);
    • Italy: 5 (5);
    • Australia: 5 (5);
    • Singapore 5;
    • Switzerland 4;
    • Saudi Arabia, Brazil, India, South Africa: 3;
    • Russia, Finland, Sweden, Spain, Taiwan: 2;
  • In the ranking of operating systems used in supercomputers, only Linux has remained for two years;
  • Distribution by Linux distributions (one year ago in brackets):
    • 48.8% (50.8%) do not detail the distribution,
    • 27.8% (23.2%) use CentOS,
    • 7.6% (9.8%) - Cray Linux,
    • 3% (3.6%) - SUSE,
    • 4.8% (5%) - RHEL,
    • 1.6% (1.4%) - Ubuntu;
    • 0.4% (0.4%) - Scientific Linux
  • The minimum performance threshold for entering the Top500 has increased over the year from 715.6 to 1022 teraflops, i.e. now there are no clusters left in the ranking with a performance of less than a petaflop (a year ago, only 272 clusters showed a performance of more than a petaflop, two years ago - 138, three years ago - 94). For Top100, the entry threshold increased from 1703 to 2395 teraflops;
  • The total performance of all systems in the rating increased over the year from 1.22 to 1.559 exaflops (four years ago it was 361 petaflops). The system that closes the current ranking was in 404th place in the last issue, and 249th in the year before;
  • The general distribution of the number of supercomputers in different parts of the world is as follows:
    267 supercomputer is located in Asia (261 years ago),
    127 in America (131) and 98 in Europe (101), 5 in Oceania and 3 in Africa;

  • As a processor base, Intel CPUs are in the lead - 95.6% (a year ago it was 95%), in second place is IBM Power - 2.6% (from 3%), in third place is SPARC64 - 0.8% (1.2%), in fourth place is AMD - 0.4% (0.4%);
  • 33.2% (a year ago 13.8%) of all used processors have 20 cores, 16.8% (21.8%) - 16 cores, 11.2% (8.6%) - 18 cores, 11.2% (21%) - 12 cores, 7% (8.2% ) - 14 cores;
  • 133 out of 500 systems (a year ago - 110) additionally use accelerators or coprocessors, while 125 systems use NVIDIA chips (a year ago there were 96), 5 - Intel Xeon Phi (there were 7), 1 - PEZY (4), 1 uses hybrid solutions (there were 2), 1 uses Matrix-2000 (1). AMD GPUs are pushed out of the list;
  • Among cluster manufacturers, Lenovo took first place with 34.6% (a year ago 23.4%), Inspur took second place with 14.2% (13.6%), Sugon took third place with 12.6% (11%), and moved from second to fourth place. Hewlett-Packard - 8% (15.8%), fifth place is occupied by Cray 7.8% (10.6%), followed by Bull 4.2% (4.2%), Dell EMC 3% (2.6%), Fujitsu 2.6% (2.6%) , IBM 2.4% (3.6%), Penguin Computing - 1.8%, Huawei 1.4% (2.8%). Interestingly, five years ago the distribution among manufacturers was as follows: Hewlett-Packard 36%, IBM 35%, Cray 10.2% and SGI 3.8% (3.4%).

At the same time, a new release of the alternative rating of cluster systems is available graph 500, focused on evaluating the performance of supercomputer platforms associated with the simulation of physical processes and tasks for processing large amounts of data, typical for such systems. Rating Green500 separately more not issued and merged with the Top500 as energy efficiency is now reflected in the main Top500 rating (based on the ratio of LINPACK FLOPS to power consumption in watts).

Source: opennet.ru

Add a comment