{"id":59877,"date":"2020-11-16T22:18:19","date_gmt":"2020-11-16T21:18:19","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=59877"},"modified":"2020-11-16T22:18:19","modified_gmt":"2020-11-16T21:18:19","slug":"top500-expands-exaflops-capacity-amidst-low-turnover","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/59877-top500-expands-exaflops-capacity-amidst-low-turnover\/","title":{"rendered":"<span class=\"caps\">TOP500<\/span> Expands Exaflops Capacity Amidst Low Turnover"},"content":{"rendered":"<p class=\"cms-plugin cms-plugin-5967\"><span lang=\"EN-US\"><span class=\"caps\">FRANKFURT<\/span>, Ger\u00adma\u00adny; <span class=\"caps\">BERKELEY<\/span>, Calif.; and <span class=\"caps\">KNOXVILLE<\/span>, Tenn.\u2014<\/span>The 56th edi\u00adti\u00adon of the <span class=\"caps\">TOP500<\/span> saw the Japa\u00adne\u00adse Fug\u00ada\u00adku super\u00adcom\u00adpu\u00adter soli\u00addi\u00adfy its num\u00adber one sta\u00adtus in a list that reflects a flat\u00adtening per\u00adfor\u00admance growth cur\u00adve.&nbsp; Alt\u00adhough two new sys\u00adtems mana\u00adged to make it into the top 10, the full list recor\u00added the smal\u00adlest num\u00adber of new ent\u00adries sin\u00adce the pro\u00adject began in&nbsp;1993.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The ent\u00adry level to the list moved up to 1.32 peta\u00adflops on the High Per\u00adfor\u00admance Lin\u00adpack (<span class=\"caps\">HPL<\/span>) bench\u00admark, a small increase from 1.23 peta\u00adflops recor\u00added in the June 2020 ran\u00adkings. In a simi\u00adlar vein, the aggre\u00adga\u00adte per\u00adfor\u00admance of all 500 sys\u00adtems grew from 2.22 exa\u00adflops in June to just 2.43 exa\u00adflops on the latest list. Like\u00adwi\u00adse, avera\u00adge con\u00adcur\u00adren\u00adcy per sys\u00adtem bare\u00adly increased at all, gro\u00adwing from 145,363 cores six months ago to 145,465 cores in the cur\u00adrent&nbsp;list.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The\u00adre were, howe\u00adver, a few nota\u00adble deve\u00adlo\u00adp\u00adments in the top 10, inclu\u00adding two new sys\u00adtems, as well as a new high\u00adwa\u00adter mark set by the top-ran\u00adked&nbsp;<a name=\"_Hlk56172079\"><\/a>Fug\u00ada\u00adku super\u00adcom\u00adpu\u00adter. Thanks to addi\u00adtio\u00adnal hard\u00adware, Fug\u00ada\u00adku grew its <span class=\"caps\">HPL<\/span> per\u00adfor\u00admance to 442 peta\u00adflops, a mode\u00adst increase from the 416 peta\u00adflops the sys\u00adtem achie\u00adved when it debut\u00aded in June 2020. More signi\u00adfi\u00adcant\u00adly, Fug\u00ada\u00adku increased its per\u00adfor\u00admance on the new mixed pre\u00adcis\u00adi\u00adon <span class=\"caps\">HPC-AI<\/span> bench\u00admark to 2.0 exa\u00adflops, best\u00ading its 1.4 exa\u00adflops mark recor\u00added six months ago. The\u00adse repres\u00adents the first bench\u00admark mea\u00adsu\u00adre\u00adments abo\u00adve one exa\u00adflop for any pre\u00adcis\u00adi\u00adon on any type of hardware.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">Here is a brief rundown of cur\u00adrent top 10 systems:<\/p>\n<ul class=\"cms-plugin cms-plugin-5967\">\n<li>Fug\u00ada\u00adku remains at the top spot, gro\u00adwing its Arm <span class=\"caps\">A64FX<\/span> capa\u00adci\u00adty from 7,299,072 cores to 7,630,848 cores. The addi\u00adtio\u00adnal hard\u00adware enab\u00adled its new world record 442 peta\u00adflops result on <span class=\"caps\">HPL<\/span>. This puts it three times ahead of the num\u00adber two sys\u00adtem in the list. Fug\u00ada\u00adku was con\u00ads\u00adtruc\u00adted by Fuji\u00adtsu and is instal\u00adled at the <span class=\"caps\">RIKEN<\/span> Cen\u00adter for Com\u00adpu\u00adta\u00adtio\u00adnal Sci\u00adence (R\u2011<span class=\"caps\">CCS<\/span>) in Kobe,&nbsp;Japan.<\/li>\n<li>Sum\u00admit, an IBM-built sys\u00adtem at the Oak Ridge Natio\u00adnal Labo\u00adra\u00adto\u00adry (<span class=\"caps\">ORNL<\/span>) in Ten\u00adnes\u00adsee, remains the fas\u00adtest sys\u00adtem in the <span class=\"caps\">US<\/span> with a per\u00adfor\u00admance of 148.8 peta\u00adflops. Sum\u00admit has 4,356 nodes, each one housing two 22-core Power9 CPUs and six <span class=\"caps\">NVIDIA<\/span> Tes\u00adla <span class=\"caps\">V100<\/span>&nbsp;GPUs.<\/li>\n<li>Sier\u00adra, a sys\u00adtem at the Law\u00adrence Liver\u00admo\u00adre Natio\u00adnal Labo\u00adra\u00adto\u00adry in Cali\u00adfor\u00adnia, is ran\u00adked third with an <span class=\"caps\">HPL<\/span> mark of 94.6 peta\u00adflops. Its archi\u00adtec\u00adtu\u00adre is very simi\u00adlar to that of Sum\u00admit, with each of its 4,320 nodes equip\u00adped with two Power9 CPUs and four <span class=\"caps\">NVIDIA<\/span> Tes\u00adla <span class=\"caps\">V100<\/span>&nbsp;GPUs.<\/li>\n<li>Sun\u00adway Tai\u00adhu\u00adLight, a sys\u00adtem deve\u00adlo\u00adped by China\u2019s Natio\u00adnal Rese\u00adarch Cen\u00adter of Par\u00adal\u00adlel Com\u00adpu\u00adter Engi\u00adnee\u00adring <span class=\"amp\">&amp;<\/span> Tech\u00adno\u00adlo\u00adgy (<span class=\"caps\">NRCPC<\/span>) and instal\u00adled at the Natio\u00adnal Super\u00adcom\u00adpu\u00adting Cen\u00adter in Wuxi, is lis\u00adted at num\u00adber four. It is powered exclu\u00adsi\u00adve\u00adly by Sun\u00adway <span class=\"caps\">SW26010<\/span> pro\u00adces\u00adsors and achie\u00adves 93 peta\u00adflops on&nbsp;<span class=\"caps\">HPL<\/span>.<\/li>\n<li>At num\u00adber five is Sele\u00adne, an <span class=\"caps\">NVIDIA<\/span> <span class=\"caps\">DGX<\/span> <span class=\"caps\">A100<\/span> Super\u00adPOD instal\u00adled in-house at <span class=\"caps\">NVIDIA<\/span> Corp. It was lis\u00adted as num\u00adber seven in June but has dou\u00adbled in size, allo\u00adwing it to move up the list by two posi\u00adti\u00adons. The sys\u00adtem is based on <span class=\"caps\">AMD<\/span> <span class=\"caps\">EPYC<\/span> pro\u00adces\u00adsors with <span class=\"caps\">NVIDIA<\/span>\u2019s new <span class=\"caps\">A100<\/span> GPUs for acce\u00adle\u00adra\u00adti\u00adon. Sele\u00adne achie\u00adved 63.4 peta\u00adflops on <span class=\"caps\">HPL<\/span> as a result of the upgrade.<\/li>\n<li>Tian\u00adhe-2A (Mil\u00adky Way-2A), a sys\u00adtem deve\u00adlo\u00adped by China\u2019s Natio\u00adnal Uni\u00adver\u00adsi\u00adty of Defen\u00adse Tech\u00adno\u00adlo\u00adgy (<span class=\"caps\">NUDT<\/span>) and deploy\u00aded at the Natio\u00adnal Super\u00adcom\u00adpu\u00adter Cen\u00adter in Guang\u00adzho, is ran\u00adked 6th. It is powered by Intel Xeon CPUs and <span class=\"caps\">NUDT<\/span>\u2019s Matrix-2000 <span class=\"caps\">DSP<\/span> acce\u00adle\u00adra\u00adtors and achie\u00adves 61.4 peta\u00adflops on&nbsp;<span class=\"caps\">HPL<\/span>.<\/li>\n<li>A new super\u00adcom\u00adpu\u00adter, known as the <span class=\"caps\">JUWELS<\/span> Boos\u00adter Modu\u00adle, debuts at num\u00adber seven on the list. The Atos-built Bull\u00adSe\u00adqua\u00adna machi\u00adne was recent\u00adly instal\u00adled at the For\u00adschungs\u00adzen\u00adtrum J\u00fclich (<span class=\"caps\">FZJ<\/span>) in Ger\u00adma\u00adny. It is part of a modu\u00adlar sys\u00adtem archi\u00adtec\u00adtu\u00adre and a second Xeon based <span class=\"caps\">JUWELS<\/span> Modu\u00adle is lis\u00adted sepa\u00adra\u00adte\u00adly on the <span class=\"caps\">TOP500<\/span> at posi\u00adti\u00adon 44. The\u00adse modu\u00adles are inte\u00adgra\u00adted by using the Par\u00adTec Modu\u00adlo Clus\u00adter Soft\u00adware Suite. The Boos\u00adter Modu\u00adle uses <span class=\"caps\">AMD<\/span> <span class=\"caps\">EPYC<\/span> pro\u00adces\u00adsors with <span class=\"caps\">NVIDIA<\/span> <span class=\"caps\">A100<\/span> GPUs for acce\u00adle\u00adra\u00adti\u00adon simi\u00adlar to the num\u00adber five Sele\u00adne sys\u00adtem. Run\u00adning by its\u00adelf the <span class=\"caps\">JUWELS<\/span> Boos\u00adter Modu\u00adle was able to achie\u00adve 44.1 <span class=\"caps\">HPL<\/span> peta\u00adflops, which makes it the most powerful sys\u00adtem in Europe<\/li>\n<li><span class=\"caps\">HPC5<\/span>, a Dell PowerEdge sys\u00adtem instal\u00adled by the Ita\u00adli\u00adan com\u00adpa\u00adny Eni S.p.A., is ran\u00adked 8th. It achie\u00adves a per\u00adfor\u00admance of 35.5 peta\u00adflops using Intel Xeon Gold CPUs and <span class=\"caps\">NVIDIA<\/span> Tes\u00adla <span class=\"caps\">V100<\/span> GPUs. It is the most powerful sys\u00adtem in the list used for com\u00admer\u00adcial pur\u00adpo\u00adses at a cus\u00adto\u00admer&nbsp;site.<\/li>\n<li>Fron\u00adte\u00adra, a Dell <span class=\"caps\">C6420<\/span> sys\u00adtem that was instal\u00adled at the Texas Advan\u00adced Com\u00adpu\u00adting Cen\u00adter of the Uni\u00adver\u00adsi\u00adty of Texas last year is now lis\u00adted at num\u00adber nine. It achie\u00adves 23.5 peta\u00adflops using 448,448 of its Intel Pla\u00adti\u00adnum Xeon&nbsp;cores.<\/li>\n<li>The second new sys\u00adtem at the top of the list is Dammam\u20117, which is ran\u00adked 10th. It is instal\u00adled at Sau\u00addi Aram\u00adco in Sau\u00addi Ara\u00adbia and is the second com\u00admer\u00adcial super\u00adcom\u00adpu\u00adter in the cur\u00adrent top 10. The <span class=\"caps\">HPE<\/span> Cray CS-Storm sys\u00adtems uses Intel Gold Xeon CPUs and <span class=\"caps\">NVIDIA<\/span> Tes\u00adla <span class=\"caps\">V100<\/span> GPUs. It rea\u00adched 22.4 peta\u00adflops on the <span class=\"caps\">HPL<\/span> benchmark.<\/li>\n<\/ul>\n<p class=\"cms-plugin cms-plugin-5967\"><strong>Other <span class=\"caps\">TOP500<\/span> highlights<\/strong><\/p>\n<p class=\"cms-plugin cms-plugin-5967\">A total of 149 sys\u00adtems on the list are using acce\u00adle\u00adra\u00adtor\/\u00adco-pro\u00adces\u00adsor tech\u00adno\u00adlo\u00adgy, up from 146 six months ago. 140 of the\u00adse use <span class=\"caps\">NVIDIA<\/span> chips.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">Intel con\u00adti\u00adnues to domi\u00adna\u00adte in <span class=\"caps\">TOP500<\/span> pro\u00adces\u00adsor share with over 90 per\u00adcent of sys\u00adtems equip\u00adped with Xeon or Xeon Phi chips. Despi\u00adte the recent rise of alter\u00adna\u00adti\u00adve pro\u00adces\u00adsor archi\u00adtec\u00adtures in high per\u00adfor\u00admance com\u00adpu\u00adting, <span class=\"caps\">AMD<\/span> pro\u00adces\u00adsors (inclu\u00adding the Hygon chip) repre\u00adsent only 21 sys\u00adtems on the cur\u00adrent list, along with ten Power-based sys\u00adtems and just five Arm-based sys\u00adtems. Howe\u00adver, the num\u00adber of sys\u00adtems with AMD-based pro\u00adces\u00adsors dou\u00adbled from what it was six months ago.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The break\u00addown in sys\u00adtem inter\u00adcon\u00adnects is lar\u00adge\u00adly unch\u00adan\u00adged from recent lists, with Ether\u00adnet used in about half the sys\u00adtems (254), Infi\u00adni\u00adBand in about a third of sys\u00adtems (182), Omni\u00adPath in about one-tenth of sys\u00adtems (47), and Myri\u00adnet in one sys\u00adtem; the rema\u00adin\u00adder use cus\u00adtom inter\u00adcon\u00adnects (38) and pro\u00adprie\u00adta\u00adry net\u00adworks (6). Infi\u00adni\u00adBand-con\u00adnec\u00adted sys\u00adtems con\u00adti\u00adnue to domi\u00adna\u00adte in aggre\u00adga\u00adte capa\u00adci\u00adty with more than an exa\u00adflop of per\u00adfor\u00admance. Sin\u00adce Fug\u00ada\u00adku uses the pro\u00adprie\u00adta\u00adry Tofu D inter\u00adcon\u00adnect, the aggre\u00adga\u00adte per\u00adfor\u00admance in the six pro\u00adprie\u00adta\u00adry net\u00adworks sys\u00adtems (472.9 peta\u00adflops) is near\u00adly equal to that of the 254 Ether\u00adnet-based sys\u00adtems (477.7 petaflops)<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">Chi\u00adna con\u00adti\u00adnues to lead in sys\u00adtem share with 212 machi\u00adnes on the list, han\u00addi\u00adly bea\u00adting out the <span class=\"caps\">US<\/span> at with 113 sys\u00adtems and Japan with 34. Howe\u00adver, despi\u00adte the smal\u00adler num\u00adber of sys\u00adtems, the <span class=\"caps\">US<\/span> con\u00adti\u00adnues to lead the list in aggre\u00adga\u00adte per\u00adfor\u00admance with 668.7 peta\u00adflops to China\u2019s 564.0 peta\u00adflops. Thanks main\u00adly to the num\u00adber one Fug\u00ada\u00adku sys\u00adtem, Japan\u2019s aggre\u00adga\u00adte per\u00adfor\u00admance of 593.7 peta\u00adflops edges out that of&nbsp;China.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\"><strong>Green500 results<\/strong><\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The most ener\u00adgy-effi\u00adci\u00adent sys\u00adtem on the Green500 is the new <span class=\"caps\">NVIDIA<\/span> <span class=\"caps\">DGX<\/span> Super\u00adPOD in the <span class=\"caps\">US<\/span>. It achie\u00adved 26.2 gigaflops\/watt power-effi\u00adci\u00aden\u00adcy during its 2.4 <span class=\"caps\">HPL<\/span> per\u00adfor\u00admance run and is lis\u00adted at posi\u00adti\u00adon 172 in the <span class=\"caps\">TOP500<\/span>.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">Next on the list is the pre\u00advious Green500 champ, <span class=\"caps\">MN<\/span>\u20113. Alt\u00adhough it impro\u00adved its score from 21.1 to 26.0 gigaflops\/watt, it slips into the num\u00adber two posi\u00adti\u00adon. The sys\u00adtem uses the MN-Core chip, an acce\u00adle\u00adra\u00adtor opti\u00admi\u00adzed for matrix arith\u00adme\u00adtic. It is ran\u00adked num\u00adber 332 in the <span class=\"caps\">TOP500<\/span>.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">In the num\u00adber three Green500 is the Atos-built <span class=\"caps\">JUWELS<\/span> Boos\u00adter Modu\u00adle instal\u00adled at For\u00adschungs\u00adzen\u00adtrum J\u00fclich (<span class=\"caps\">FZJ<\/span>) in Ger\u00adma\u00adny. It achie\u00adves 25.0 gigaflops\/watt and is ran\u00adked seventh in the <span class=\"caps\">TOP500<\/span>.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">In fourth posi\u00adti\u00adon is Spartan\u20112, ano\u00adther Atos-built machi\u00adne. It achie\u00adves 24.3 gigaflops\/watt on <span class=\"caps\">HPL<\/span> and is ran\u00adked at posi\u00adti\u00adon 148 on the <span class=\"caps\">TOP500<\/span> list.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The fifth-ran\u00adked sys\u00adtem on the Green500 is Sele\u00adne, with an effi\u00adci\u00aden\u00adcy of 24.0 gigaflops\/watt. It also occu\u00adp\u00adies the num\u00adber five spot on the <span class=\"caps\">TOP500<\/span>.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">With the excep\u00adti\u00adon of the <span class=\"caps\">MN<\/span>\u20113 sys\u00adtem, the remai\u00adning top five Green500 sys\u00adtems are using the new <span class=\"caps\">NVIDIA<\/span> <span class=\"caps\">A100<\/span> <span class=\"caps\">GPU<\/span> as an acce\u00adle\u00adra\u00adtor. All four of the\u00adse sys\u00adtems use <span class=\"caps\">AMD<\/span> <span class=\"caps\">EPYC<\/span> as their main&nbsp;<span class=\"caps\">CPU<\/span>.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">Of the top 40 sys\u00adtems on the Green500, 37 levera\u00adge acce\u00adle\u00adra\u00adtors, 2 use <span class=\"caps\">A64FX<\/span> vec\u00adtor-pro\u00adces\u00adsors, and one (Tai\u00adhu\u00adLight) a Sun\u00adway many-core processor.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">Extra\u00adpo\u00adla\u00adting the power effi\u00adci\u00aden\u00adcy value of 26.2 gigaflops\/watt of the <span class=\"caps\">NVIDIA<\/span> <span class=\"caps\">DGX<\/span> Super\u00adPOD out line\u00adar\u00adly to an exa\u00adflop would result in a power con\u00adsump\u00adti\u00adon of 38 <span class=\"caps\">MW<\/span> (igno\u00adring addi\u00adtio\u00adnal hard\u00adware nee\u00added for scaling).<\/p>\n<p class=\"cms-plugin cms-plugin-5967\"><strong><span class=\"caps\">HPCG<\/span> Results<\/strong><\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The <span class=\"caps\">TOP500<\/span> list has incor\u00adpo\u00adra\u00adted the High-Per\u00adfor\u00admance Con\u00adju\u00adga\u00adte Gra\u00addi\u00adent (<span class=\"caps\">HPCG<\/span>) Bench\u00admark results, which pro\u00advi\u00addes an alter\u00adna\u00adti\u00adve metric for asses\u00adsing super\u00adcom\u00adpu\u00adter per\u00adfor\u00admance and is meant to com\u00adple\u00adment the <span class=\"caps\">HPL<\/span> measurement.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The list-lea\u00adding Fug\u00ada\u00adku expan\u00added its <span class=\"caps\">HPCG<\/span> result with a record 16.0 HPCG-peta\u00adflops. The two <span class=\"caps\">US<\/span> Depart\u00adment of Ener\u00adgy sys\u00adtems, Sum\u00admit at <span class=\"caps\">ORNL<\/span> and Sier\u00adra at <span class=\"caps\">LLNL<\/span>, are second and third, respec\u00adtively, on the <span class=\"caps\">HPCG<\/span> bench\u00admark. Sum\u00admit achie\u00adved 2.93 HPCG-peta\u00adflops and Sier\u00adra 1.80 HPCG-peta\u00adflops. The only other sys\u00adtems to break the peta\u00adflops bar\u00adri\u00ader on <span class=\"caps\">HPCG<\/span> are the upgraded Sele\u00adne sys\u00adtem at 1.62 peta\u00adflops and the new <span class=\"caps\">JUWELS<\/span> Boos\u00adter Modu\u00adle at 1.28 petaflops.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\"><strong><span class=\"caps\">HPL-AI<\/span> Results<\/strong><\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The <span class=\"caps\">HPL-AI<\/span> bench\u00admark seeks to high\u00adlight the con\u00adver\u00adgence of <span class=\"caps\">HPC<\/span> and arti\u00adfi\u00adci\u00adal intel\u00adli\u00adgence (<span class=\"caps\">AI<\/span>) workloads based on machi\u00adne lear\u00adning and deep lear\u00adning by sol\u00adving a sys\u00adtem of line\u00adar equa\u00adtions using novel, mixed-pre\u00adcis\u00adi\u00adon algo\u00adrith\u00adms that exploit modern hardware.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The top-ran\u00adked sys\u00adtem for this bench\u00admark is <span class=\"caps\">RIKEN<\/span>\u2019s Fug\u00ada\u00adku sys\u00adtem, which achie\u00adved 2.0 exa\u00adflops of mixed pre\u00adcis\u00adi\u00adon com\u00adpu\u00adta\u00adti\u00adon. At num\u00adber two is <span class=\"caps\">ORNL<\/span>\u2019s Sum\u00admit super\u00adcom\u00adpu\u00adter, which achie\u00adved 0.55 exa\u00adflops, fol\u00adlo\u00adwed by <span class=\"caps\">NVIDIA<\/span>\u2019s Sele\u00adne which tur\u00adned in an <span class=\"caps\">HPL-AI<\/span> result of 0.25 exaflops.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\"><strong>About the <span class=\"caps\">TOP500<\/span> List<\/strong><\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The first ver\u00adsi\u00adon of what beca\u00adme today\u2019s <span class=\"caps\">TOP500<\/span> list star\u00adted as an exer\u00adcise for a small con\u00adfe\u00adrence in Ger\u00adma\u00adny in June 1993. Out of curio\u00adsi\u00adty, the aut\u00adhors deci\u00added to revi\u00adsit the list in Novem\u00adber 1993 to see how things had chan\u00adged. About that time, they rea\u00adli\u00adzed they might be onto some\u00adthing and deci\u00added to con\u00adti\u00adnue com\u00adpi\u00adling the list, which is now a much-anti\u00adci\u00adpa\u00adted, much-wat\u00adched and much-deba\u00adted twice-year\u00adly&nbsp;event.<\/p>\n<p class=\"cms-plugin cms-plugin-5967\">The <span class=\"caps\">TOP500<\/span> list is com\u00adpi\u00adled by Erich Stroh\u00admai\u00ader and Horst Simon of Law\u00adrence Ber\u00adke\u00adley Natio\u00adnal Labo\u00adra\u00adto\u00adry; Jack Don\u00adgar\u00adra of the Uni\u00adver\u00adsi\u00adty of Ten\u00adnes\u00adsee, Knox\u00adville; and Mar\u00adtin Meu\u00ader of <span class=\"caps\">ISC<\/span> Group, Germany.<\/p>\n","protected":false},"excerpt":{"rendered":"<p><span class=\"caps\">FRANKFURT<\/span>, Ger\u00adma\u00adny; <span class=\"caps\">BERKELEY<\/span>, Calif.; and <span class=\"caps\">KNOXVILLE<\/span>, Tenn.\u2014The 56th edi\u00adti\u00adon of the <span class=\"caps\">TOP500<\/span> saw the Japa\u00adne\u00adse Fug\u00ada\u00adku super\u00adcom\u00adpu\u00adter soli\u00addi\u00adfy its num\u00adber one sta\u00adtus in a list that reflects a flat\u00adtening per\u00adfor\u00admance growth cur\u00adve.&nbsp; Alt\u00adhough two new sys\u00adtems mana\u00adged to make it into the top 10, the full list recor\u00added the smal\u00adlest num\u00adber of new ent\u00adries sin\u00adce the pro\u00adject began in&nbsp;1993.<\/p>\n<p>The ent\u00adry level to the list moved up to 1.32 peta\u00adflops on the High Per\u00adfor\u00admance Lin\u00adpack (<span class=\"caps\">HPL<\/span>) bench\u00admark, a small increase from 1.23 peta\u00adflops recor\u00added in the June 2020 ran\u00adkings. In a simi\u00adlar vein, the aggre\u00adga\u00adte per\u00adfor\u00admance of all 500 sys\u00adtems grew from 2.22 exa\u00adflops in June to just 2.43 exa\u00adflops on the latest list. Like\u00adwi\u00adse, avera\u00adge con\u00adcur\u00adren\u00adcy per sys\u00adtem bare\u00adly increased at all, gro\u00adwing from 145,363 cores six months ago to 145,465 cores in the cur\u00adrent list. (\u2026) <a class=\"moretag\" href=\"https:\/\/www.planet3dnow.de\/cms\/59877-top500-expands-exaflops-capacity-amidst-low-turnover\/\">Wei\u00adter\u00adle\u00adsen&nbsp;\u00bb<\/a><\/p>\n","protected":false},"author":593,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wp_typography_post_enhancements_disabled":false,"ngg_post_thumbnail":0,"footnotes":""},"categories":[22],"tags":[1598,759],"class_list":["post-59877","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","tag-supercomputer","tag-top500","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/59877","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/users\/593"}],"replies":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/comments?post=59877"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/59877\/revisions"}],"predecessor-version":[{"id":59878,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/59877\/revisions\/59878"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=59877"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=59877"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=59877"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}