An buga bugu 58 na kimar mafi girman manyan kwamfutoci

An buga bugu na 58 na kima na kwamfutoci 500 mafi inganci a duniya. A cikin sabon sakin, manyan goma ba su canza ba, amma 4 sabbin gungu na Rasha sun haɗa a cikin matsayi.

Rukunin Rasha na Chervonenkis, Galushkin, da Lyapunov, waɗanda Yandex ya ƙirƙira don magance matsalolin koyon injina da kuma samar da aikin petaflops 21.5, 16, da 12.8, bi da bi, sun ɗauki matsayi na 19, 36, da 40 a cikin jerin. Rukunin suna da ƙarfi ta hanyar Ubuntu 16.04 kuma suna da na'urori masu sarrafawa na AMD EPYC 7xxx da GPU na NVIDIA A100: ƙungiyar Chervonenkis tana da nodes 199 (cores 193 AMD EPYC 7702 64C 2GH da GPU na NVIDIA A100 80G 1592), Galushkin - nodes 136 (cores 134 AMD EPYC 7702 64C 2GH da GPU na NVIDIA A100 80G 1088), Lyapunov - nodes 137 (cores 130 AMD EPYC 7662 64C 2GHz da GPU na NVIDIA A100 40G 1096).

A matsayi na 43, sabon rukunin Sberbank, Christofari Neo, yana gudanar da NVIDIA DGX OS 5 (bugu). Ubuntu) kuma yana ba da petaflops 11.9 na aiki. Ƙungiyar tana da sama da ƙwayoyin kwamfuta 98 bisa ga AMD EPYC 7742 64C 2.25GHz CPU kuma tana zuwa tare da GPU na NVIDIA A100 80GB. Ƙungiyar Christofari da Sberbank ta tura a baya ta tashi daga matsayi na 61 zuwa na 72 a cikin jerin a cikin watanni shida kacal.

Har ila yau, wasu gungu na gida biyu sun kasance a cikin matsayi: Lomonosov 2 - ya koma daga 199 zuwa 241 (a cikin 2015, Lomonosov 2 cluster ya dauki matsayi 31, kuma Lomonosov wanda ya gabace shi a 2011 - 13 wuri) da MTS GROM - daga 240 zuwa 294 wuri . Don haka, adadin gungu na cikin gida a cikin martaba ya karu daga 3 zuwa 7 sama da watanni shida (don kwatanta, a cikin 2020 akwai tsarin gida 2 a cikin martaba, a cikin 2017 - 5, kuma a cikin 2012 - 12).

Dangane da babban matsayi, ƙungiyar Fugaku ta Japan, wacce aka gina ta amfani da na'urori masu sarrafawa na ARM, ta ci gaba da kasancewa a matsayi na farko. Ƙungiyar Fugaku tana cikin Cibiyar Binciken Jiki da Sinadarai ta RIKEN kuma tana ba da petaflops 442 na aiki. Ƙungiyar ta ƙunshi nodes 158976 bisa ga Fujitsu A64FX SoC, wacce aka sanye ta da CPU na Armv8.2-A SVE mai core 48 (512-bit SIMD) tare da mitar agogo na 2.2 GHz. Gabaɗaya, ƙungiyar tana da sama da cores na processor miliyan 7.6 (sau uku fiye da shugaban da ya gabata), 5 PB na RAM, da 150 PB na ajiya da aka raba bisa ga tsarin fayil ɗin Lustre. Ana amfani da Red Hat Enterprise a matsayin tsarin aiki. LinuxJimillar tsawon kebul na gani da ake amfani da shi wajen haɗa maɓallan ya kai kimanin kilomita 850.

A matsayi na biyu kuma, ƙungiyar Summit ce, wadda IBM ta tura a Oak Ridge National Laboratory (Amurka). Red Hat Enterprise ce ke kula da wannan ƙungiya. Linux, ya haɗa da cores masu sarrafawa miliyan 2.4 (ana amfani da CPUs na IBM Power9 22C 3.07GHz guda 22 da kuma masu saurin NVIDIA Tesla V100), waɗanda ke ba da aikin petaflops 148, wanda ya ninka aikin shugaban kimantawa sau uku.

Na uku ya zo ne ga rukunin American Sierra, wanda IBM ta sanya a Lawrence Livermore National Laboratory, bisa wani dandamali mai kama da Summit kuma yana nuna aiki a petaflops 94 (kimanin cores miliyan 1.5). Ana amfani da Red Hat Enterprise a matsayin tsarin aiki. Linux.

A matsayi na huɗu akwai ƙungiyar Sunway TaihuLight ta ƙasar Sin, wacce ke aiki a Cibiyar Supercomputer ta ƙasa ta ƙasar Sin, wadda ta ƙunshi sama da cores miliyan 10 kuma tana samar da aikin petaflops 93. Duk da irin wannan alkaluman aiki, ƙungiyar Sierra tana cinye rabin makamashin da Sunway TaihuLight ke amfani da shi. Tana amfani da tsarin aikinta. Linux-RaiseOS rarrabawa.

A matsayi na biyar akwai gungu na Perlmutter, wanda HPE ya samar kuma yana Cibiyar Nazarin Makamashi ta Ƙasa a Amurka. Tarin ya haɗa da nau'i-nau'i dubu 761 dangane da AMD EPYC 7763 64C 2.45GHz CPU kuma yana ba da aikin 71 petaflops. Tsarin aiki shine Cray OS.

Abubuwan da suka fi ban sha'awa:

• Rarraba ta adadin manyan kwamfutoci a ƙasashe daban-daban:
  • China: 173 (188 - watanni shida da suka wuce). A cikin duka, gungu na kasar Sin suna samar da kashi 17.5% na duk abin da ake samarwa (watanni shida da suka gabata - 19.4%);
  • Amurka: 149 (122). An kiyasta jimlar yawan aiki a 32.5% (watanni shida da suka gabata - 30.7%);
  • Japan: 32 (34);
  • Jamus: 26 (23);
  • Faransa: 19 (16);
  • Netherlands: 11 (16);
  • Birtaniya: 11 (11);
  • Kanada 11 (11);
  • Rasha 7 (3);
  • Koriya ta Kudu 7 (5)
  • Italiya: 6 (6);
  • Saudi Arabia 6 (6);
  • Brazil 5 (6);
  • Sweden 4 (3);
  • Poland 4 (4);
  • Australia, India, Switzerland, Finland: 3.
• A cikin jerin tsarin aiki da ake amfani da shi a cikin manyan kwamfutoci, tsawon shekaru huɗu da rabi yanzu, kawai Linux;
• Rarrabawa ta hanyar rarrabawa Linux (a cikin maƙallan - shekaru biyu da suka wuce):
  • 51.6% (49.6%) ba su dalla-dalla rarrabawar,
  • 18% (26.4%) amfani CentOS,
  • 7.6% (4.8%) - RHEL,
  • 7% (6.8%) — Cray Linux,
  • 5.4% (2%) — Ubuntu;
  • 4% (3%) - SUSE,
  • 0.2% (0.4%) — Kimiyya Linux
• Matsakaicin iyakar aiki don shigar da Top500 a cikin watanni 6 ya karu daga 1511 zuwa 1649 teraflops (shekaru uku da suka gabata, gungu 272 kawai sun nuna aikin fiye da petaflop, shekaru hudu da suka gabata - 138, shekaru biyar da suka gabata - 94). Don Top100, ƙofar shiga ya karu daga 4124 zuwa 4788 teraflops;
• Jimlar ayyukan duk tsarin a cikin ƙimar ya karu daga 2.8 zuwa 3 exaflops a cikin shekara (shekaru biyu da suka wuce 1.650 exaflops ne, kuma shekaru biyar da suka wuce - 566 petaflops). Tsarin da ya rufe matsayi na yanzu ya kasance a matsayi na 433 a cikin fitowar karshe, kuma a matsayi na 401 a shekarar da ta gabata;
• Babban rabon manyan kwamfutoci a sassa daban-daban na duniya shine kamar haka: 226 supercomputers suna cikin Asiya (245 - watanni shida da suka gabata), 160 a Arewacin Amurka (133) da 105 a Turai (113), 5 a Kudancin Amurka. Amurka (6), 3 a Oceania (2) da 1 a Afirka (1);
• A matsayin tushen processor, Intel CPUs suna kan gaba - 81.6% (shekaru biyu da suka gabata shine 94%), AMD yana matsayi na biyu tare da 14.6% (0.6% !!), IBM Power yana matsayi na uku - 1.4% ( ya kasance 2.8%. Akwai ci gaba mai ƙarfi na gungu dangane da na'urori na AMD; misali, duk sabbin tsarin da aka haɗa a cikin Top15 sanye take da AMD CPUs.
• 26.6% (shekaru biyu da suka wuce 35.6%) na duk na'urorin da aka yi amfani da su suna da nau'in nau'i na 20, 17.6% - 24 cores, 11.2% - 64 cores, 8.6% (13.8%) - 16 cores, 8.2% (11%) - 18 cores, 5.8 % (11.2%) - kashi 12.
• 149 daga cikin 500 tsarin (shekaru biyu da suka wuce - 144) kuma suna amfani da accelerators ko coprocessors, tare da tsarin 143 ta amfani da kwakwalwan kwamfuta na NVIDIA, 2 - Intel Xeon Phi (daga 5), ​​1 - PEZY (1), da 1 AMD Vega GPU;
• Daga cikin masana'antun gungu, Lenovo ya ɗauki matsayi na farko - 36.8% (shekaru biyu da suka wuce 34.8%), Inspur ya ɗauki matsayi na biyu - 11.6% (13.2%), Kamfanin Hewlett-Packard ya ɗauki matsayi na uku - 9% (7%), Sugon 7.8 ya biyo baya. % (14.2%), Atos - 7.2% (4.6%), Cray 6.4% (7%), Dell EMC 3.2% (2.2%), Fujitsu 3% (2.6%), NVIDIA 2.4 (1.2%), NEC 2% , Huawei 1.4% (2%), IBM 1.4% (2.6%), Penguin Computing - 1.4% (2.2%). Shekaru bakwai da suka gabata, rarraba tsakanin masana'antun ya kasance kamar haka: Hewlett-Packard 36%, IBM 35%, Cray 10.2% da SGI 3.8%;
• Ana amfani da Ethernet don haɗa nodes a cikin 49.4% (shekaru biyu da suka wuce 52%) na gungu, ana amfani da InfiniBand a cikin 33.6% (28%) na gungu, Omnipath - 8.4% (10%). Duban aikin gabaɗaya, tsarin tushen InfiniBand yana da kashi 43.3% na ayyukan Top500 gabaɗaya, yayin da Ethernet ke lissafin kashi 21.3%.

Nan gaba kadan, ana sa ran buga wani sabon bugu na madadin kima na tsarin tari mai lamba 500, wanda aka mayar da hankali kan tantance ayyukan dandali na manyan kwamfutoci da ke da alaƙa da simintin hanyoyin jiki da ayyuka don sarrafa bayanai masu yawa na irin waɗannan tsarin. Green500, HPCG (High-Performance Conjugate Gradient) da HPL-AI ana haɗe su tare da Top500 kuma suna nunawa a cikin babban ƙimar Top500.

source: budenet.ru