An riga an buga bugu na 60 na manyan 500

Bayan watanni 6 na buga lambar da ta gabata da kuma bin kalandar ɗaba'ar, da sabon bugu na 60 na martabar kwamfutoci 500 mafi girma a duniya.

A cikin sabon bugu, akwai canji ɗaya kawai a cikin manyan goma: kungiyar Leonardo, dake cikin cibiyar bincike ta Italiya CINECA, a matsayi na hudu. Tarin ya ƙunshi kusan nau'ikan sarrafawa miliyan 1,5 (Xeon Platinum 8358 32C 2,6 GHz CPU) kuma yana ba da aikin petaflops 255,75.

Manyan wurare goma sun haɗa da:

1. Frontier, wanda yake a Ma'aikatar Makamashi ta Oak Ridge National Laboratory. Tarin yana da kusan nau'ikan sarrafawa miliyan 9 (64GHz AMD EPYC 2C CPU, AMD Instinct MI250X accelerator) kuma yana ba da exaflops na 1.102, wanda ya kusan sau uku fiye da na biyu. wuri gungu.
2. Fugaku, wanda ke zaune a Cibiyar Nazarin Jiki da Kimiyya ta RIKEN (Japan). An gina gungu tare da masu sarrafa ARM (158976 nodes dangane da Fujitsu A64FX SoC, sanye take da 8.2-core 48GHz Armv2,2-A SVE CPU) yana ba da 442 petaflops na aiki.
3. An shirya LUMI a Cibiyar Supercomputing ta Turai (EuroHPC) a Finland kuma tana ba da 151 petaflops na aiki. Tarin ya dogara ne akan dandamalin HPE Cray EX235a iri ɗaya a matsayin jagora a cikin martaba, amma ya haɗa da na'urori masu sarrafawa miliyan 1,1 (AMD EPYC 64C 2GHz, AMD Instinct MI250X accelerator, Slingshot-11 network).
4. Leonardo ya karbi bakuncin a EuroHPC daban-daban a CINECA, Italiya. Yana da tsarin Atos BullSequana XH2000 tare da Xeon Platinum 8358 32C 2.6GHz a matsayin manyan masu sarrafawa, NVIDIA A100 SXM4 40 GB azaman masu haɓakawa da Quad-rail NVIDIA HDR100 Infiniband azaman haɗin haɗin gwiwa. Ya sami aikin Linpack na 174,7 Pflop/s.
5. Taron, wanda IBM ya gina kuma yana zaune a Oak Ridge National Laboratory (ORNL) a cikin Tennessee, Amurka, yanzu an sanya shi #5 tare da yin aiki na 148,8 Pflop/s akan ma'aunin HPL, wanda Ana amfani dashi don sanya jerin TOP500.
6. Saliyo, wanda aka shirya a Laboratory National Lawrence Livermore, CA, Amurka, gine-ginensa yayi kama da tsarin taron #5. An gina shi tare da nodes 4320 tare da CPUs POWER9 guda biyu da NVIDIA Tesla V100 GPUs guda hudu. Saliyo ya samu 94,6 Pflop/s.
7. Sunway TaihuLight, wani tsarin da cibiyar binciken fasahar kere-kere ta kasar Sin ta National Parallel Computer Engineering and Technology Research Center (NRCPC) ta kirkira, kuma aka sanya shi a cibiyar sarrafa kwamfuta ta kasa da ke Wuxi, na lardin Jiangsu na kasar Sin, ya samu matsayi na 7 tare da 93 Pflop/s.
8. Perlmutter a #8 ya dogara ne akan dandamalin HPE Cray “Shasta” da tsarin iri-iri tare da nodes na tushen AMD EPYC da 1536 NVIDIA A100 masu haɓaka nodes. Perlmutter ya buga 64,6 Pflop/s
9. Selene yanzu a No. 9 shine NVIDIA DGX A100 SuperPOD wanda aka shigar a cikin gida a NVIDIA a Amurka. Tsarin ya dogara ne akan na'ura mai sarrafa AMD EPYC tare da NVIDIA A100 don haɓakawa da Mellanox HDR InfiniBand azaman hanyar sadarwa kuma ya sami 63,4 Pflop/s.
10. Tianhe-2A (Milky Way-2A), tsarin da Jami'ar Tsaro ta kasar Sin (NUDT) ta kirkira kuma aka aiwatar a Cibiyar Supercomputer ta kasa da ke Guangzhou, kasar Sin, yanzu an jera shi a matsayin tsarin na 10 mai lamba 61,4, XNUMX Pflop/s. .

Amma ga supercomputers na gida, gungu Chervonenkis, Galushkin da Lyapunov wanda Yandex ya kirkira sun fadi daga wurare 22, 40 da 43 zuwa wurare 25, 44 da 47. An tsara waɗannan gungu don magance matsalolin koyon inji da ba da aikin 21,5, 16, da 12,8 petaflops, bi da bi.

Ƙungiyar Christofari Neo wanda Sberbank ya tura ya faɗi daga matsayi na 46 zuwa 50. Christofari Neo yana gudanar da NVIDIA DGX OS 5 (Ubuntu Edition) kuma yana ba da 11,9 petaflops na aiki. Tarin yana da fiye da 98k cores dangane da AMD EPYC 7742 64C 2.25GHz CPU kuma ya zo tare da NVIDIA A100 80GB GPU. Rukunin na biyu na Sberbank (Christofari) ya tashi daga matsayi na 80 zuwa 87 a cikin matsayi a cikin watanni shida.

Ga wani bangare na mafi ban sha'awa halaye ta Linux rabawa (a cikin baka - watanni 6 da suka gabata):
47.8% (47.8%) ba su dalla-dalla rarrabawa.
17,2% (18,2%) amfani da CentOS
9,6% (8,8%) - RHEL
9% (8%) - CrayLinux
5,4% (5,2%) - Ubuntu
3,8% (3,8%) - SUSE
0,8% (0,8%) - Alma Linux
0,8% (0,8%) - RockyLinux
0,2% (0,2%) - Linux na Kimiyya.

Matsakaicin iyakar aiki don shigar da Top500 na tsawon watanni 6 ya kasance 1,73 petaflops (watanni shida da suka gabata, 1,65 petaflops). Shekaru hudu da suka wuce gungu 272 ne kawai suka nuna kwazo a kan petaflops, shekaru biyar da suka gabata 138, shekaru shida da suka gabata 94). Ga Manyan 100, ƙofar shiga ta ƙaru daga 5,39 zuwa 9,22 petaflops.

Jimlar ayyukan duk tsarin a cikin martaba ya karu daga 4,4 zuwa 4,8 exaflops a cikin watanni 6 (shekaru uku da suka wuce 1.650 exaflops ne kuma shekaru biyar da suka gabata ya kasance 749 petaflops). Tsarin da ya rufe kima na yanzu an sanya shi 458 a cikin fitowar ta ƙarshe.

Finalmente idan kuna sha'awar ƙarin sani game da shi, zaka iya duba bayanan A cikin mahaɗin mai zuwa.