{"id":52934,"date":"2019-11-19T23:38:41","date_gmt":"2019-11-19T22:38:41","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=52934"},"modified":"2019-11-19T23:38:41","modified_gmt":"2019-11-19T22:38:41","slug":"penguin-computing-upgrades-corona-with-latest-amd-radeon-instinct-gpu-technology-for-enhanced-ml-and-ai-capabilities","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/52934-penguin-computing-upgrades-corona-with-latest-amd-radeon-instinct-gpu-technology-for-enhanced-ml-and-ai-capabilities\/","title":{"rendered":"Penguin Computing Upgrades Corona with Latest <span class=\"caps\">AMD<\/span> Radeon Instinct <span class=\"caps\">GPU<\/span> Technology for Enhanced <span class=\"caps\">ML<\/span> and <span class=\"caps\">AI<\/span> Capabilities"},"content":{"rendered":"<div class=\"wpb_row vc_row vc_row-fluid mk-fullwidth-false attched-false js-master-row mk-grid mk-in-viewport\" data-mk-stretch-content=\"true\">\n<div class=\"vc_col-sm-12 wpb_column column_container _ height-full\">\n<div class=\" vc_custom_1574114544920\">\n<div id=\"text-block-3\" class=\"mk-text-block \">\n<p><strong>Fre\u00admont, <span class=\"caps\">CA<\/span>.,<\/strong>&nbsp;\u2013&nbsp;<strong>Novem\u00adber 18, 2019<\/strong>&nbsp;\u2013&nbsp;<a href=\"https:\/\/www.penguincomputing.com\/\" target=\"_blank\" rel=\"noopener noreferrer\">Pen\u00adgu\u00adin Com\u00adpu\u00adting<\/a>, a lea\u00adder in high-per\u00adfor\u00admance com\u00adpu\u00adting (<span class=\"caps\">HPC<\/span>), arti\u00adfi\u00adci\u00adal intel\u00adli\u00adgence (<span class=\"caps\">AI<\/span>), and enter\u00adpri\u00adse data cen\u00adter solu\u00adti\u00adons and ser\u00advices, today announ\u00adced that&nbsp;<em>Coro\u00adna<\/em>, an <span class=\"caps\">HPC<\/span> clus\u00adter first deli\u00adver\u00aded to Law\u00adrence Liver\u00admo\u00adre Natio\u00adnal Lab (<span class=\"caps\">LLNL<\/span>) in late 2018, has been upgraded with the newest <span class=\"caps\">AMD<\/span> Rade\u00adon Instinct<sup><span class=\"caps\">TM<\/span><\/sup>&nbsp;<span class=\"caps\">MI60<\/span> acce\u00adle\u00adra\u00adtors, based on Vega which, per <span class=\"caps\">AMD<\/span>, is the world\u2019s 1<sup>st<\/sup>&nbsp;7nm <span class=\"caps\">GPU<\/span> archi\u00adtec\u00adtu\u00adre that brings PCIe\u00ae 4.0 sup\u00adport. This upgrade is the latest exam\u00adp\u00adle of Pen\u00adgu\u00adin Com\u00adpu\u00adting and <span class=\"caps\">LLNL<\/span>\u2019s ongo\u00ading col\u00adla\u00adbo\u00adra\u00adti\u00adon aimed at pro\u00advi\u00adding addi\u00adtio\u00adnal capa\u00adbi\u00adli\u00adties to the <span class=\"caps\">LLNL<\/span> user community.<\/p>\n<p>As pre\u00advious\u00adly released, the clus\u00adter con\u00adsists of 170 two-socket nodes with 24-core <span class=\"caps\">AMD<\/span> <span class=\"caps\">EPYC<\/span><sup><span class=\"caps\">TM<\/span><\/sup>&nbsp;7401 pro\u00adces\u00adsors and a PCIe 1.6 Tera\u00adbyte (<span class=\"caps\">TB<\/span>) non\u00advo\u00adla\u00adti\u00adle (solid-sta\u00adte) memo\u00adry device. Each Coro\u00adna com\u00adpu\u00adte node is GPU-rea\u00addy with half of tho\u00adse nodes today uti\u00adli\u00adzing four <span class=\"caps\">AMD<\/span> Rade\u00adon Instinct <span class=\"caps\">MI25<\/span> acce\u00adle\u00adra\u00adtors per node, deli\u00adve\u00adring 4.2 peta\u00adFLOPS of <span class=\"caps\">FP32<\/span> peak per\u00adfor\u00admance. With the <span class=\"caps\">MI60<\/span> upgrade, the clus\u00adter increa\u00adses its poten\u00adti\u00adal <span class=\"caps\">PFLOPS<\/span> peak per\u00adfor\u00admance to 9.45 peta\u00adFLOPS of <span class=\"caps\">FP32<\/span> peak per\u00adfor\u00admance. This brings signi\u00adfi\u00adcant\u00adly grea\u00adter per\u00adfor\u00admance and <span class=\"caps\">AI<\/span> capa\u00adbi\u00adli\u00adties to the rese\u00adarch communities.<\/p>\n<p><span class=\"dquo\">\u201c<\/span>The Pen\u00adgu\u00adin Com\u00adpu\u00adting <span class=\"caps\">DOE<\/span> team con\u00adti\u00adnues our col\u00adla\u00adbo\u00adra\u00adti\u00adve ven\u00adture with our ven\u00addor part\u00adners <span class=\"caps\">AMD<\/span> and Mel\u00adlan\u00adox to ensu\u00adre the Liver\u00admo\u00adre Coro\u00adna <span class=\"caps\">GPU<\/span> enhance\u00adments expand the capa\u00adbi\u00adli\u00adties to con\u00adti\u00adnue their mis\u00adsi\u00adon out\u00adreach within various machi\u00adne lear\u00adning com\u00admu\u00adni\u00adties,\u201d said Ken Guden\u00adrath, Direc\u00adtor of Fede\u00adral Sys\u00adtems at Pen\u00adgu\u00adin Computing.<\/p>\n<p>Coro\u00adna is being made available to indus\u00adtry through <span class=\"caps\">LLNL<\/span>\u2019s High Per\u00adfor\u00admance Com\u00adpu\u00adting Inno\u00adva\u00adti\u00adon Cen\u00adter (<span class=\"caps\">HPCIC<\/span>). Fun\u00added through the Com\u00admo\u00addi\u00adty Tech\u00adno\u00adlo\u00adgy Sys\u00adtems (<span class=\"caps\">CTS<\/span>\u20111) con\u00adtract with the Natio\u00adnal Nuclear Secu\u00adri\u00adty Admi\u00adnis\u00adtra\u00adti\u00adon (<span class=\"caps\">NNSA<\/span>), the upgrade will help <span class=\"caps\">LLNL<\/span> rese\u00adar\u00adchers and their indus\u00adtry part\u00adners impro\u00adve capa\u00adbi\u00adli\u00adties in sca\u00adlable deep lear\u00adning, big data ana\u00adly\u00adtics and data sci\u00adence, while enhan\u00adcing <span class=\"caps\">NNSA<\/span>\u2019s abili\u00adty to assess future archi\u00adtec\u00adtures and meet the needs of the <span class=\"caps\">NNSA<\/span>\u2019s Advan\u00adced Simu\u00adla\u00adti\u00adon <span class=\"amp\">&amp;<\/span> Com\u00adpu\u00adting pro\u00adgram. It will also pro\u00advi\u00adde a hig\u00adher level of per\u00adfor\u00admance for rese\u00adar\u00adching cogni\u00adti\u00adve com\u00adpu\u00adting and deve\u00adlo\u00adping pre\u00addic\u00adti\u00adve simu\u00adla\u00adti\u00adons for appli\u00adca\u00adti\u00adons such as iner\u00adti\u00adal con\u00adfi\u00adne\u00adment fusi\u00adon and mole\u00adcu\u00adlar dyna\u00admics simu\u00adla\u00adti\u00adons for pre\u00adcis\u00adi\u00adon medicine.<\/p>\n<p><span class=\"dquo\">\u201c<\/span>This upgrade signi\u00adfi\u00adcant\u00adly increa\u00adses the capa\u00adbi\u00adli\u00adty available on Coro\u00adna,\u201d said Bro\u00adnis R. de Supin\u00adski, Chief Tech\u00adni\u00adcal Offi\u00adcer for Liver\u00admo\u00adre Com\u00adpu\u00adting. \u201cThe new Vega GPUs offer sub\u00adstan\u00adti\u00adal dou\u00adble-pre\u00adcis\u00adi\u00adon per\u00adfor\u00admance, in addi\u00adti\u00adon to much more sin\u00adgle-pre\u00adcis\u00adi\u00adon per\u00adfor\u00admance. <span class=\"caps\">LLNL<\/span> sci\u00aden\u00adtists will use the com\u00adbi\u00adna\u00adti\u00adon to under\u00adstand the poten\u00adti\u00adal of mixed-pre\u00adcis\u00adi\u00adon algo\u00adrith\u00adms for a varie\u00adty of domains.\u201d<\/p>\n<p><span class=\"caps\">AMD<\/span>\u2019s Rade\u00adon Instinct <span class=\"caps\">MI60<\/span> acce\u00adle\u00adra\u00adtors bring many new fea\u00adtures that impro\u00adve per\u00adfor\u00admance, inclu\u00adding the Vega 7nm <span class=\"caps\">GPU<\/span> archi\u00adtec\u00adtu\u00adre and the <span class=\"caps\">AMD<\/span> Infi\u00adni\u00adty Fabric<sup><span class=\"caps\">TM<\/span><\/sup>&nbsp;Link tech\u00adno\u00adlo\u00adgy, a peer-to-peer <span class=\"caps\">GPU<\/span> com\u00admu\u00adni\u00adca\u00adti\u00adons tech\u00adno\u00adlo\u00adgy that deli\u00advers up to 184 <span class=\"caps\">GB<\/span>\/s trans\u00adfer speeds bet\u00adween GPUs \u2013 which is 5.<span class=\"caps\">75X<\/span> fas\u00adter than PCIe Gen 3, and full-chip Error-cor\u00adrec\u00adting code (<span class=\"caps\">ECC<\/span>)11 and Relia\u00adbi\u00adli\u00adty, Acces\u00adsi\u00adbi\u00adli\u00adty and Ser\u00advicea\u00adbi\u00adli\u00adty (<span class=\"caps\">RAS<\/span>) 12 tech\u00adno\u00adlo\u00adgies. The new acce\u00adle\u00adra\u00adtors also uti\u00adli\u00adze the latest ROCm open source soft\u00adware stack, which is now inte\u00adgra\u00adted into lea\u00adding frame\u00adworks like Ten\u00adsor\u00adFlow and PyTorch and maps workloads to the hete\u00adro\u00adge\u00adneous com\u00adpu\u00adte resour\u00adces of the under\u00adly\u00ading hardware.<\/p>\n<p><span class=\"dquo\">\u201c<\/span><span class=\"caps\">AMD<\/span> is plea\u00adsed to con\u00adti\u00adnue col\u00adla\u00adbo\u00adra\u00adti\u00adon with <span class=\"caps\">LLNL<\/span> and the <span class=\"caps\">NNSA<\/span> in advan\u00adcing open acce\u00adle\u00adra\u00adtor solu\u00adti\u00adons. Access to sys\u00adtems like Coro\u00adna enable next gene\u00adra\u00adti\u00adon sci\u00aden\u00adti\u00adfic dis\u00adco\u00advery as we move to the exas\u00adca\u00adle era,\u201d said Ogi Brkic, Cor\u00adpo\u00adra\u00adte Vice Pre\u00adsi\u00addent and Gene\u00adral Mana\u00adger of the Data Cen\u00adter <span class=\"caps\">GPU<\/span> Busi\u00adness Unit at&nbsp;<span class=\"caps\">AMD<\/span>.<\/p>\n<p>For more infor\u00adma\u00adti\u00adon about <span class=\"caps\">AMD<\/span> Rade\u00adon Instinct, plea\u00adse visit&nbsp;<a href=\"https:\/\/www.amd.com\/en\/graphics\/servers-radeon-instinct-mi\" target=\"_blank\" rel=\"noopener noreferrer\">www.amd.com\/en\/graphics\/servers-radeon-instinct-mi<\/a>.<\/p>\n<div class=\"clearboth\">&nbsp;<\/div>\n<\/div>\n<\/div>\n<div id=\"padding-4\" class=\"mk-padding-divider clearfix\">&nbsp;<\/div>\n<\/div>\n<\/div>\n<div class=\"wpb_row vc_row vc_row-fluid mk-fullwidth-true attched-false vc_custom_1522173355982 js-master-row mk-in-viewport\" data-mk-full-width=\"true\" data-mk-full-width-init=\"true\">\n<div class=\"mk-grid\">\n<div class=\"vc_col-sm-12 wpb_column column_container _ height-full\">\n<div id=\"padding-6\" class=\"mk-padding-divider clearfix\">&nbsp;<\/div>\n<div class=\"wpb_row vc_inner vc_row vc_row-fluid attched-false \">\n<div class=\"wpb_column vc_column_container vc_col-sm-12\">\n<div class=\"vc_column-inner\">\n<div class=\"wpb_wrapper\">\n<div class=\" vc_custom_1560809608644\">\n<div id=\"text-block-7\" class=\"mk-text-block \">\n<h4>About Penguin Computing<\/h4>\n<p>For 20 years, the Pen\u00adgu\u00adin Com\u00adpu\u00adting team of arti\u00adfi\u00adci\u00adal intel\u00adli\u00adgence (<span class=\"caps\">AI<\/span>), engi\u00adnee\u00adring, and com\u00adpu\u00adter sci\u00adence experts has reim\u00adagi\u00adned how start\u00adups, For\u00adtu\u00adne 500, govern\u00adment, and aca\u00adde\u00admic orga\u00adniza\u00adti\u00adons sol\u00adve com\u00adplex tech\u00adno\u00adlo\u00adgy chal\u00adlenges and achie\u00adve their orga\u00adniza\u00adtio\u00adnal goals. Pen\u00adgu\u00adin Com\u00adpu\u00adting is focu\u00adsed on open plat\u00adforms, inclu\u00adding Open Com\u00adpu\u00adte Pro\u00adject (<span class=\"caps\">OCP<\/span>) sys\u00adtems. We spe\u00adcia\u00adli\u00adze in inno\u00adva\u00adti\u00adve on-pre\u00admi\u00adse high-per\u00adfor\u00admance com\u00adpu\u00adting (<span class=\"caps\">HPC<\/span>), bare metal <span class=\"caps\">HPC<\/span> in the cloud, <span class=\"caps\">AI<\/span>, and sto\u00adrage tech\u00adno\u00adlo\u00adgies cou\u00adpled with lea\u00adding-edge design, imple\u00admen\u00adta\u00adti\u00adon, hos\u00adting, and mana\u00adged ser\u00advices inclu\u00adding sys-admin and sto\u00adrage-as-a-ser\u00advice, and high\u00adly rated cus\u00adto\u00admer sup\u00adport. More infor\u00adma\u00adti\u00adon at&nbsp;<a href=\"https:\/\/www.penguincomputing.com\/\" target=\"_blank\" rel=\"noopener noreferrer\">www.penguincomputing.com<\/a>.<\/p>\n<div class=\"clearboth\">&nbsp;<\/div>\n<\/div>\n<\/div>\n<div id=\"padding-8\" class=\"mk-padding-divider clearfix\">&nbsp;<\/div>\n<div class=\"vc_separator wpb_content_element vc_separator_align_center vc_sep_width_100 vc_sep_pos_align_center vc_separator_no_text vc_sep_color_grey\">&nbsp;<\/div>\n<div id=\"padding-9\" class=\"mk-padding-divider clearfix\">&nbsp;<\/div>\n<div class=\" vc_custom_1574114466896\">\n<div id=\"text-block-10\" class=\"mk-text-block \">\n<p class=\"pr-co-info-attrib\">Pen\u00adgu\u00adin Com\u00adpu\u00adting, Rel\u00adion and Frost\u00adByte are trade\u00admarks or regis\u00adtered trade\u00admarks of Pen\u00adgu\u00adin Com\u00adpu\u00adting, Inc. Pen\u00adgu\u00adin Com\u00adpu\u00adting is a sub\u00adsi\u00addia\u00adry of&nbsp;<a href=\"http:\/\/ir.smartm.com\/\" target=\"_blank\" rel=\"noopener noreferrer\"><span class=\"caps\">SMART<\/span> Glo\u00adbal Hol\u00addings Inc.<\/a>, (<span class=\"caps\">NASDAQ<\/span>: <span class=\"caps\">SGH<\/span>).<\/p>\n<p class=\"pr-co-info-attrib\">PCIe is a regis\u00adtered trade\u00admark of <span class=\"caps\">PCI-SIG<\/span>.<\/p>\n<p class=\"pr-co-info-attrib\"><span class=\"caps\">AMD<\/span>, the <span class=\"caps\">AMD<\/span> logo, <span class=\"caps\">EPYC<\/span>, Infi\u00adni\u00adty Fabric, Rade\u00adon Instinct, and com\u00adbi\u00adna\u00adti\u00adons the\u00adreof are trade\u00admarks or regis\u00adtered trade\u00admarks of Advan\u00adced Micro Devices, Inc.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>Fre\u00admont, <span class=\"caps\">CA<\/span>.,&nbsp;\u2013&nbsp;Novem\u00adber 18, 2019&nbsp;\u2013&nbsp;Pen\u00adgu\u00adin Com\u00adpu\u00adting, a lea\u00adder in high-per\u00ad\u00adfor\u00ad\u00admance com\u00adpu\u00adting (<span class=\"caps\">HPC<\/span>), arti\u00adfi\u00adci\u00adal intel\u00adli\u00adgence (<span class=\"caps\">AI<\/span>), and enter\u00adpri\u00adse data cen\u00adter solu\u00adti\u00adons and ser\u00advices, today announ\u00adced that&nbsp;Coro\u00adna, an <span class=\"caps\">HPC<\/span> clus\u00adter first deli\u00adver\u00aded to Law\u00adrence Liver\u00admo\u00adre Natio\u00adnal Lab (<span class=\"caps\">LLNL<\/span>) in late 2018, has been upgraded with the newest <span class=\"caps\">AMD<\/span> Rade\u00adon InstinctTM&nbsp;<span class=\"caps\">MI60<\/span> acce\u00adle\u00adra\u00adtors, based on Vega which, per <span class=\"caps\">AMD<\/span>, is (\u2026) <a class=\"moretag\" href=\"https:\/\/www.planet3dnow.de\/cms\/52934-penguin-computing-upgrades-corona-with-latest-amd-radeon-instinct-gpu-technology-for-enhanced-ml-and-ai-capabilities\/\">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":[966,2127,1579,1598],"class_list":["post-52934","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","tag-amd","tag-penguin-computing","tag-radeon-instinct","tag-supercomputer","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/52934","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=52934"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/52934\/revisions"}],"predecessor-version":[{"id":52935,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/52934\/revisions\/52935"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=52934"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=52934"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=52934"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}