Ua paʻi ʻia ka paʻi 60th o ka pae o ka 500 o nā kamepiula hana kiʻekiʻe loa ma ka honua. I ka paʻi hou, hoʻokahi wale nō hoʻololi i ka ʻumi kiʻekiʻe - ʻo ka 4th wahi i lawe ʻia e ka hui ʻo Leonardo, aia ma ka hale noiʻi ʻepekema ʻItalia CINECA. Loaʻa i ka pūʻulu kahi kokoke i 1.5 miliona mau cores processor (CPU Xeon Platinum 8358 32C 2.6GHz) a hāʻawi i ka hana o 255.75 petaflops me ka hoʻohana mana o 5610 kilowatts.
ʻO nā ʻekolu kiʻekiʻe, e like me 6 mau mahina i hala aku nei, aia nā pūʻulu aʻe:
- Frontier - Noho ʻia ma ke Keʻena ʻOihana o ʻAmelika Hui Pū ʻIa o Oak Ridge National Laboratory. ʻO ka puʻupuʻu he kokoke i 9 miliona mau cores processor (AMD EPYC 64C 2GHz CPU, AMD Instinct MI250X accelerator) a hāʻawi i ka hana o 1.102 exaflops, kahi kokoke i ʻekolu mau manawa ma mua o ka puʻupuʻu lua (ʻoiai ʻo 30% ka haʻahaʻa o ka mana o Frontier).
- Fugaku - noho ʻia ma ka RIKEN Institute of Physical and Chemical Research (Iapana). Kūkulu ʻia ka pūʻulu me ka hoʻohana ʻana i nā kaʻina hana ARM (158976 nodes e pili ana i Fujitsu A64FX SoC, i lako me kahi 48-core Armv8.2-A SVE 2.2GHz CPU). Hāʻawi ʻo Fugaku i 442 petaflops o ka hana.
- Hoʻokipa ʻia ʻo LUMI ma ka European Supercomputing Center (EuroHPC) ma Finland a hāʻawi iā 151 petaflops o ka hana. Kūkulu ʻia ka pūʻulu ma luna o ka papa HPE Cray EX235a e like me ke alakaʻi o ka helu, akā e komo pū me 1.1 miliona mau cores processor (AMD EPYC 64C 2GHz, AMD Instinct MI250X accelerator, Slingshot-11 network).
No nā supercomputers home, ua hāʻule nā hui Chervonenkis, Galushkin a me Lyapunov i hana ʻia e Yandex mai 22, 40 a me 43 mau wahi i 25, 44 a me 47 mau wahi. Hoʻolālā ʻia kēia mau pūʻulu e hoʻoponopono i nā pilikia aʻo mīkini a hāʻawi i ka hana o 21.5, 16 a me 12.8 petaflops, kēlā me kēia. Holo nā pūʻulu i ka Ubuntu 16.04 a hoʻolako ʻia me nā kaʻina hana AMD EPYC 7xxx a me NVIDIA A100 GPUs: ʻo ka hui Chervonenkis he 199 nodes (193 tausani AMD EPYC 7702 64C 2GH cores a me 1592 NVIDIA A100 80G GPUs), Galushkin - EP136 tausani nodes AMD. cores 134 7702C 64GH a me 2 GPU NVIDIA A1088 100G), Lyapunov - 80 nodes (137 tausani cores AMD EPYC 130 7662C 64GHz a me 2 GPU NVIDIA A1096 100G).
ʻO ka pūʻulu Christofari Neo i hoʻoili ʻia e Sberbank i hāʻule mai ka 46th a i ka 50th wahi. Holo ʻo Christofari Neo i ka NVIDIA DGX OS 5 (Ubuntu edition) a hōʻike i ka hana o 11.9 petaflops. ʻOi aku ka nui o ka pūʻulu ma mua o 98 tausani mau cores computing e pili ana i ka AMD EPYC 7742 64C 2.25GHz CPU a hele mai me kahi NVIDIA A100 80GB GPU. Ua neʻe ka lua o ka pūʻulu o Sberbank (Christofari) mai 80th a i 87th wahi ma ka pae ma mua o ʻeono mahina.
ʻElua mau puʻupuʻu home ʻē aʻe i noho i ka pae: Lomonosov 2 - ua hoʻololi ʻia mai 262 a 290 wahi (i ka makahiki 2015, ua noho ka hui Lomonosov 2 i 31 wahi, a ʻo kona mua Lomonosov ma 2011 - 13 wahi) a me MTS GROM - ua hoʻololi ʻia mai 318 a 352 wahi . No laila, ʻaʻole i loli ka helu o nā puʻupuʻu home i ka pae a, e like me ʻeono mahina i hala aku nei, he 7 ʻōnaehana (no ka hoʻohālikelike ʻana, i ka makahiki 2020 aia nā ʻōnaehana home 2 i ka pae, ma 2017 - 5, a ma 2012 - 12).
ʻO nā hiʻohiʻona hoihoi loa:
- Ka mahele ʻana ma ka helu o nā supercomputers ma nā ʻāina like ʻole:
- Kina: 162 (173 eono mahina aku nei). ʻO ka huina, hoʻopuka nā hui Kina i 10% o nā huahana āpau (ʻeono mahina i hala - 12%);
- US: 127(127). Kuhi ʻia ka nui o ka hana ma 43.6% o ka hana helu holoʻokoʻa (ʻeono mahina i hala - 47.3%);
- Kelemānia: 34 (31). Ka huina hoonui - 4.5%;
- Iapana: 31 (34). Ka huina hoonui - 12.8%;
- Palani: 24 (22). Ka huina hoonui - 3.6%;
- UK: 15(12);
- Kanaka 10 (14);
- Netherlands: 8 (6);
- Kōlea Hema 8 (6)
- Palakila 8 (6);
- Rūsia 7 (7);
- Italia: 7 (6);
- Saudi Arabia 6 (6);
- Suedena 6 (5);
- Australia 5 (5);
- Irelani 5;
- Polani 5 (5);
- Kuikilana 4 (4);
- Finland: 3 (4).
- Singapore: 3;
- Inia: 3;
- Polani: 3;
- Norewai: 3.
- I ka helu ʻana o nā ʻōnaehana hana i hoʻohana ʻia i nā supercomputers, ʻo Linux wale nō i koe no ʻeono makahiki;
- Hāʻawi ʻia e nā māhele Linux (ma nā brackets - 6 mau mahina aku nei):
- 47.8% (47.8%) ʻaʻole kikoʻī i ka puʻunaue;
- 17.2% (18.2%) hoʻohana i ka CentOS;
- 9.6% (8.8%) - RHEL;
- 9% (8%) - Cray Linux;
- 5.4% (5.2%) - Ubuntu;
- 3.8% (3.8%) - SUSE;
- 0.8% (0.8%) - Alma Linux;
- 0.8% (0.8%) - Rocky Linux;
- 0.2% (0.2%) - Linux ʻepekema.
- ʻO ka paepae hana haʻahaʻa loa no ke komo ʻana i ka Top500 no 6 mau mahina he 1.73 petaflops (ʻeono mahina i hala - 1.65 petaflops). ʻEhā mau makahiki i hala aku nei, he 272 mau puʻupuʻu wale nō i hōʻike i ka hana ma mua o ka petaflop, ʻelima mau makahiki i hala - 138, ʻeono makahiki i hala - 94). No Top100, ua hoʻonui ka paepae komo mai 5.39 a i 9.22 petaflops;
- ʻO ka nui o ka hana o nā ʻōnaehana āpau i ka helu ma luna o 6 mau mahina i hoʻonui ʻia mai 4.4 a 4.8 exaflops (ʻekolu mau makahiki i hala aku nei he 1.650 exaflops, a ʻelima mau makahiki i hala - 749 petaflops). ʻO ka ʻōnaehana e pani ana i ka pae o kēia manawa aia ma kahi 458th ma ka helu hope loa;
- ʻO ka māhele nui o ka helu o nā supercomputers ma nā wahi like ʻole o ka honua penei: 218 supercomputers aia ma Asia (229 ʻeono mahina i hala), 137 ma ʻAmelika ʻĀkau (141) a me 131 ma ʻEulopa (118), 8 ma ʻAmelika Hema (6), 5 ma Oceania (5) a me 1 ma Africa (1);
- Ma keʻano he papa hana, aia nā Intel CPU i ke alakaʻi - 75.6% (ʻeono mahina i hala aku nei he 77.4%), aia ka AMD ma kahi lua me 20.2% (18.8%), a aia ʻo IBM Power ma ke kolu o kahi - 1.4% (ʻo 1.4). %).
- ʻO 22.2% (ʻeono mahina i hala aku nei 20%) o nā kaʻina hana i hoʻohana ʻia he 24 cores, 15.8% (15%) - 64 cores, 14.2% (19.2%) - 20 cores, 8.4% (8.8%) - 16 cores, 7.6% ( 8.2% ) - 18 cores, 6% - 28 cores, 5% (5.4%) - 12 cores.
- 177 mai loko o 500 ʻōnaehana (ʻeono mahina i hala aku nei - 167) hoʻohana hou i nā accelerators a i ʻole coprocessors, aʻo 161 mau ʻōnaehana hoʻohana i nā pahu NVIDIA, 9 - AMD, 2 - Intel Xeon Phi (mai 5), 1 - PEZY (1), 1 - MN- Core, 1 - Matrix-2000;
- Ma waena o nā mea hana cluster, ua lawe ʻo Lenovo i kahi mua me 32% (ʻeono mahina i hala aku nei he 32%), ua lawe ʻo Hewlett-Packard Enterprise i kahi lua me 20.2% (19.2%), ua lawe ʻo Inspur i ke kolu o kahi me 10% (10%), ukali ʻia e Atos - 8.6% (8.4%), Sugon 6.8% (7.2%), Dell EMC 3.6% (3.4%), NVIDIA 2.8% (2.8%), NEC 2.4% (2%), Fujitsu 2% (2.6%), MEGWARE 1.2 %, Penguin Computing - 1.2% (1.2%), IBM 1.2% (1.2%), Huawei 0.4% (1.4%).
- No ka hoʻohui ʻana i nā nodes ma 46.6% (ʻeono mahina i hala aku nei 45.4%) o nā pūʻulu, hoʻohana ʻia ʻo Ethernet, hoʻohana ʻia ʻo InfiniBand i 38.8% (39.2%) o nā pūʻulu, Omnipath - 7.2% (7.8%). Inā mākou e noʻonoʻo i ka nui o ka hana, a laila uhi nā ʻōnaehana InfiniBand-based 33.6% (32.4%) o nā hana Top500 āpau, a me Ethernet - 46.2% (45.1%).
I ka wā e hiki mai ana, manaʻo ʻia e paʻi ʻia kahi hoʻokuʻu hou o ka Graph 500 o nā ʻōnaehana cluster, e nānā ana i ka loiloi ʻana i ka hana o nā kahua supercomputer e pili ana i ka hoʻohālikelike ʻana i nā kaʻina hana kino a me nā hana no ka hoʻoili ʻana i ka nui o nā ʻikepili i loaʻa i loko o ia mau ʻōnaehana. Hoʻohui ʻia ka Green500, HPCG (High-Performance Conjugate Gradient) a me HPL-AI me ka Top500 a hōʻike ʻia i ka pae nui Top500.
Source: opennet.ru