{"id":56743,"date":"2020-07-09T08:30:21","date_gmt":"2020-07-09T06:30:21","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=56743"},"modified":"2020-07-09T08:30:21","modified_gmt":"2020-07-09T06:30:21","slug":"nvidia-ampere-gpus-come-to-google-cloud-at-speed-of-light","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/56743-nvidia-ampere-gpus-come-to-google-cloud-at-speed-of-light\/","title":{"rendered":"NVIDIA<\/span> Ampere GPUs Come to Google Cloud at Speed of Light"},"content":{"rendered":"

Just weeks after launch, NVIDIA<\/span> A100<\/span> GPUs land on Goog\u00adle Com\u00adpu\u00adte Engine.<\/p>\n

<\/a><\/p>\n

The NVIDIA<\/span> A100<\/span> Ten\u00adsor Core GPU<\/span><\/a> has lan\u00added on Goog\u00adle Cloud.<\/p>\n

Available in alpha on Goog\u00adle Com\u00adpu\u00adte Engi\u00adne just over a month after its intro\u00adduc\u00adtion, A100<\/span> has come to the cloud fas\u00adter than any NVIDIA<\/span> GPU<\/span> in history.<\/p>\n

Today\u2019s intro\u00adduc\u00adtion of the Acce\u00adle\u00adra\u00adtor-Opti\u00admi\u00adzed VM<\/span> (A2<\/span>) ins\u00adtance fami\u00adly fea\u00adturing A100<\/span> makes Goog\u00adle the first cloud ser\u00advice pro\u00advi\u00adder to offer the new NVIDIA<\/span> GPU<\/span>.<\/p>\n

A100<\/span>, which is built on the new\u00adly intro\u00addu\u00adced NVIDIA<\/span> Ampere archi\u00adtec\u00adtu\u00adre<\/a>, deli\u00advers NVIDIA<\/span>\u2019s grea\u00adtest gene\u00adra\u00adtio\u00adnal leap ever. It boosts trai\u00adning and infe\u00adrence com\u00adpu\u00adting per\u00adfor\u00admance by 20x over its pre\u00adde\u00adces\u00adsors, pro\u00advi\u00adding tre\u00admen\u00addous spee\u00addups for workloads to power the AI<\/span> revolution.<\/p>\n

\u201c<\/span>Goog\u00adle Cloud cus\u00adto\u00admers often look to us to pro\u00advi\u00adde the latest hard\u00adware and soft\u00adware ser\u00advices to help them dri\u00adve inno\u00adva\u00adti\u00adon on AI<\/span> and sci\u00aden\u00adti\u00adfic com\u00adpu\u00adting workloads, \u201d said Manish Saina\u00adni, direc\u00adtor of Pro\u00adduct Manage\u00adment at Goog\u00adle Cloud. \u201cWith our new A2<\/span> VM<\/span> fami\u00adly, we are proud to be the first major cloud pro\u00advi\u00adder to mar\u00adket NVIDIA<\/span> A100<\/span> GPUs, just as we were with NVIDIA<\/span> T4<\/span> GPUs. We are exci\u00adted to see what our cus\u00adto\u00admers will do with the\u00adse new capabilities.\u201d<\/p>\n

In cloud data cen\u00adters, A100<\/span> can power a broad ran\u00adge of com\u00adpu\u00adte-inten\u00adsi\u00adve appli\u00adca\u00adti\u00adons, inclu\u00adding AI<\/span> trai\u00adning and infe\u00adrence, data ana\u00adly\u00adtics, sci\u00aden\u00adti\u00adfic com\u00adpu\u00adting, geno\u00admics, edge video ana\u00adly\u00adtics, 5G<\/span> ser\u00advices, and more.<\/p>\n

Fast-gro\u00adwing, cri\u00adti\u00adcal indus\u00adtries will be able to acce\u00adle\u00adra\u00adte their dis\u00adco\u00adveries with the breakth\u00adrough per\u00adfor\u00admance of A100<\/span> on Goog\u00adle Com\u00adpu\u00adte Engi\u00adne. From sca\u00adling up AI<\/span> trai\u00adning and sci\u00aden\u00adti\u00adfic com\u00adpu\u00adting, to sca\u00adling out infe\u00adrence appli\u00adca\u00adti\u00adons, to enab\u00adling real-time con\u00adver\u00adsa\u00adtio\u00adnal AI<\/span><\/a>, A100<\/span> acce\u00adle\u00adra\u00adtes com\u00adplex and unpre\u00addic\u00adta\u00adble workloads of all sizes run\u00adning in the cloud. <\/b><\/p>\n

NVIDIA<\/span> CUDA<\/span> 11<\/a>, coming to gene\u00adral avai\u00adla\u00adbi\u00adli\u00adty soon, makes acces\u00adsi\u00adble to deve\u00adlo\u00adpers the new capa\u00adbi\u00adli\u00adties of NVIDIA<\/span> A100<\/span> GPUs, inclu\u00adding Ten\u00adsor Cores, mixed-pre\u00adcis\u00adi\u00adon modes, mul\u00adti-ins\u00adtance GPU<\/span>, advan\u00adced memo\u00adry manage\u00adment and stan\u00addard C++\/Fortran par\u00adal\u00adlel lan\u00adguage constructs.<\/p>\n

Breakthrough A100<\/span> Performance in the Cloud for Every Size Workload<\/b><\/h2>\n

The new A2<\/span> VM<\/span> ins\u00adtances can deli\u00adver dif\u00adfe\u00adrent levels of per\u00adfor\u00admance to effi\u00adci\u00adent\u00adly acce\u00adle\u00adra\u00adte workloads across CUDA-enab\u00adled machi\u00adne lear\u00adning trai\u00adning and infe\u00adrence, data ana\u00adly\u00adtics, as well as high per\u00adfor\u00admance computing.<\/p>\n

For lar\u00adge, deman\u00adding workloads, Goog\u00adle Com\u00adpu\u00adte Engi\u00adne offers cus\u00adto\u00admers the a2-megag\u00adpu-16g ins\u00adtance, which comes with 16 A100<\/span> GPUs, offe\u00adring a total of 640GB<\/span> of GPU<\/span> memo\u00adry and 1.3TB<\/span> of sys\u00adtem memo\u00adry \u2014 all con\u00adnec\u00adted through NVS\u00adwitch<\/a> with up to 9.6TB<\/span>\/s of aggre\u00adga\u00adte bandwidth.<\/p>\n

For tho\u00adse with smal\u00adler workloads, Goog\u00adle Com\u00adpu\u00adte Engi\u00adne is also offe\u00adring A2<\/span> VMs in smal\u00adler con\u00adfi\u00adgu\u00adra\u00adti\u00adons to match spe\u00adci\u00adfic appli\u00adca\u00adti\u00adons\u2019 needs.<\/p>\n

Goog\u00adle Cloud announ\u00adced that addi\u00adtio\u00adnal NVIDIA<\/span> A100<\/span> sup\u00adport is coming soon to Goog\u00adle Kuber\u00adnetes Engi\u00adne, Cloud AI<\/span> Plat\u00adform and other Goog\u00adle Cloud ser\u00advices. For more infor\u00adma\u00adti\u00adon, inclu\u00adding tech\u00adni\u00adcal details on the new A2<\/span> VM<\/span> fami\u00adly and how to sign up for access, visit the Goog\u00adle Cloud blog<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"

Just weeks after launch, NVIDIA<\/span> A100<\/span> GPUs land on Goog\u00adle Com\u00adpu\u00adte Engi\u00adne. The NVIDIA<\/span> A100<\/span> Ten\u00adsor Core GPU<\/span> has lan\u00added on Goog\u00adle Cloud. Available in alpha on Goog\u00adle Com\u00adpu\u00adte Engi\u00adne just over a month after its intro\u00adduc\u00adtion, A100<\/span> has come to the cloud fas\u00adter than any NVIDIA<\/span> GPU<\/span> in histo\u00adry. Today\u2019s intro\u00adduc\u00adtion of the Acce\u00ad\u00adle\u00adra\u00ad\u00adtor-Opti\u00ad\u00admi\u00ad\u00adzed VM<\/span> (A2<\/span>) ins\u00adtance (\u2026) Wei\u00adter\u00adle\u00adsen \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":[2332,2021,992],"class_list":["post-56743","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","tag-a100","tag-google-cloud","tag-nvidia","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/56743","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=56743"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/56743\/revisions"}],"predecessor-version":[{"id":56745,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/56743\/revisions\/56745"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=56743"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=56743"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=56743"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}