{"id":13251,"date":"2014-11-14T14:54:22","date_gmt":"2014-11-14T13:54:22","guid":{"rendered":"http:\/\/www.planet3dnow.de\/cms\/?p=13251"},"modified":"2014-11-21T11:16:26","modified_gmt":"2014-11-21T10:16:26","slug":"amd-firepro-gpus-enable-gsi-research-facilitys-heavy-ion-studies","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/13251-amd-firepro-gpus-enable-gsi-research-facilitys-heavy-ion-studies\/","title":{"rendered":"AMD<\/span> FirePro GPUs Enable GSI<\/span> Research Facility\u2019s Heavy Ion Studies"},"content":{"rendered":"

AMD<\/span> Fire\u00adPro S9150<\/span> GPU<\/span> powers more than 3 peta\u00adFLOPS super\u00adcom\u00adpu\u00adter clus\u00adter for Open\u00adCL\u2122 appli\u00adca\u00adti\u00adons at GSI<\/span><\/p>\n

SUNNYVALE<\/span>, Calif. 11\/13\/2014<\/em><\/p>\n

AMD<\/span><\/a> (NYSE<\/span>: AMD<\/span>) lea\u00adder\u00adship in hete\u00adro\u00adge\u00adneous high per\u00adfor\u00admance com\u00adpu\u00adting (HPC<\/span>) is deli\u00adve\u00adring more than 3 peta\u00adFLOPS of AMD<\/span> Fire\u00adPro\u2122 S9150<\/span><\/a> ser\u00adver GPU<\/span> per\u00adfor\u00admance for the GSI<\/span> Helm\u00adholtz\u00adzen\u00adtrum f\u00fcr Schwer\u00adio\u00adnen\u00adfor\u00adschung GmbH in sup\u00adport of phy\u00adsics rese\u00adarch rela\u00adted to hea\u00advy ion stu\u00addies. Each of the 160 ASUS<\/span> ESC4000<\/span> G2S<\/span> nodes1<\/sup> of the new L\u2011CSC<\/span> clus\u00adter include four AMD<\/span> Fire\u00adPro S9150<\/span> ser\u00adver GPUs, for a total GPU<\/span> peak of 3.25 peta\u00adFLOPS sin\u00adgle pre\u00adcis\u00adi\u00adon and 1.62 peta\u00adFLOPS dou\u00adble pre\u00adcis\u00adi\u00adon per\u00adfor\u00admance. The clus\u00adter is enab\u00adling Lat\u00adti\u00adce Quan\u00adtum Chro\u00admo Dyna\u00admics (Lat\u00adti\u00adce QCD<\/span>) com\u00adpu\u00adta\u00adtio\u00adnal rese\u00adarch using one of the fas\u00adtest Open\u00adCL\u2122 imple\u00admen\u00adta\u00adti\u00adons for rese\u00adarch appli\u00adca\u00adti\u00adons in the world. Lat\u00adti\u00adce QCD<\/span> cal\u00adcu\u00adla\u00adti\u00adons are appli\u00adca\u00adble to a ran\u00adge of stu\u00addies in high ener\u00adgy and hea\u00advy ion physics.<\/p>\n

\u201c<\/span>AMD<\/span> is proud to col\u00adla\u00adbo\u00adra\u00adte with ASUS<\/span>, the Frank\u00adfurt Insti\u00adtu\u00adte for Advan\u00adced Stu\u00addies, (FIAS<\/span>) and GSI<\/span> to sup\u00adport such important phy\u00adsics and com\u00adpu\u00adter sci\u00adence rese\u00adarch,\u201d said David Cum\u00admings, seni\u00ador direc\u00adtor and gene\u00adral mana\u00adger, pro\u00adfes\u00adsio\u00adnal gra\u00adphics, AMD<\/span>. \u201cThis instal\u00adla\u00adti\u00adon reaf\u00adfirms AMD<\/span>\u2019s lea\u00adding role in HPC<\/span> with the imple\u00admen\u00adta\u00adti\u00adon of the AMD<\/span> Fire\u00adPro S9150<\/span> ser\u00adver GPUs in this three peta\u00adFLOPS super\u00adcom\u00adpu\u00adter clus\u00adter. AMD<\/span> and ASUS<\/span> are enab\u00adling Open\u00adCL\u2122 appli\u00adca\u00adti\u00adons for cri\u00adti\u00adcal sci\u00adence rese\u00adarch usa\u00adge for this clus\u00adter. We\u2019re com\u00admit\u00adted to buil\u00adding our HPC<\/span> lea\u00adder\u00adship posi\u00adti\u00adon in the indus\u00adtry as a fore\u00admost pro\u00advi\u00adder of com\u00adpu\u00adting appli\u00adca\u00adti\u00adons, tools and technologies.\u201d<\/p>\n

\u201c<\/span>We had excel\u00adlent coope\u00adra\u00adti\u00adon with ASUS<\/span> and AMD<\/span> to make this pro\u00adject hap\u00adpen in such a short time\u00adframe,\u201d said Pro\u00adfes\u00adsor Doc\u00adtor Vol\u00adker Lin\u00addenstruth, pro\u00adfes\u00adsor at Goe\u00adthe Uni\u00adver\u00adsi\u00adty of Frank\u00adfurt and chair\u00adman of Frank\u00adfurt Insti\u00adtu\u00adte for Advan\u00adced Stu\u00addies. \u201cThe ASUS<\/span> ESC4000<\/span> G2S<\/span> ser\u00advers and the AMD<\/span> Fire\u00adPro S9150<\/span> GPUs are an extre\u00adme\u00adly powerful basis for the L\u2011CSC<\/span> clus\u00adter and they pro\u00advi\u00adde the com\u00adpu\u00adte capa\u00adbi\u00adli\u00adties we need for our rese\u00adarch. The lar\u00adge 16GB<\/span> of memo\u00adry of the AMD<\/span> Fire\u00adPro S9150<\/span> ser\u00adver GPUs allows us to run most LQCD<\/span> com\u00adpu\u00adta\u00adti\u00adons on one GPU<\/span> wit\u00adhout inter-GPU and inter-node com\u00admu\u00adni\u00adca\u00adti\u00adon resul\u00adting in very effi\u00adci\u00adent LQCD<\/span> appli\u00adca\u00adti\u00adon ope\u00adra\u00adti\u00adon. We have cho\u00adsen a mul\u00adti-GPU approach with four GPU<\/span> boards per ser\u00adver to enable an extre\u00adme\u00adly cost-effi\u00adci\u00adent and ener\u00adgy-effi\u00adci\u00adent configuration.\u201d<\/p>\n

\u201c<\/span>ASUS<\/span> is pas\u00adsio\u00adna\u00adte about the HPC<\/span> field, in the same way that AMD<\/span> is dedi\u00adca\u00adted to deli\u00adve\u00adring extre\u00adme-per\u00adfor\u00admance GPU<\/span> solu\u00adti\u00adons. Tog\u00ade\u00adther, ASUS<\/span> and AMD<\/span> have rea\u00adched a new level with the GSI<\/span> pro\u00adject, reve\u00adal\u00ading the true power of ASUS<\/span> ESC4000<\/span> G2S<\/span> GPU<\/span> ser\u00advers and AMD<\/span> Fire\u00adPro S9150<\/span> GPUs,\u201d said Tom Lin, ASUS<\/span> Gene\u00adral Mana\u00adger. \u201cWe\u2019re proud to offer our cus\u00adto\u00admers solu\u00adti\u00adons that deli\u00adver unpar\u00adal\u00adle\u00adled per\u00adfor\u00admance and utmost ener\u00adgy effi\u00adci\u00aden\u00adcy. The\u00adre is no doubt that our joint GSI<\/span> pro\u00adject deli\u00advers top-tier per\u00adfor\u00admance for cri\u00adti\u00adcal HPC<\/span> applications.\u201d<\/p>\n

GSI<\/span> (Gesell\u00adschaft f\u00fcr Schwer\u00adio\u00adnen\u00adfor\u00adschung), loca\u00adted in Darm\u00adstadt, Ger\u00adma\u00adny, is a rese\u00adarch faci\u00adli\u00adty for hea\u00advy ion rese\u00adarch inclu\u00adding can\u00adcer the\u00adra\u00adpy. The com\u00adpu\u00adte clus\u00adter was desi\u00adgned and built in con\u00adjunc\u00adtion with FIAS<\/span>, a rese\u00adarch insti\u00adtu\u00adte for phy\u00adsics, neu\u00adro sci\u00adence, life sci\u00adence, and com\u00adpu\u00adter sci\u00adence. Within the next few years, GSI<\/span> will build FAIR<\/span> (Faci\u00adli\u00adty for Anti-Pro\u00adton and Hea\u00advy-Ion Rese\u00adarch), a new par\u00adtic\u00adle acce\u00adle\u00adra\u00adtor and detec\u00adtor faci\u00adli\u00adty. Rese\u00adar\u00adchers from around the world use this faci\u00adli\u00adty for expe\u00adri\u00adments that help them make fasci\u00adna\u00adting rese\u00adarch discoveries.<\/p>\n

High ener\u00adgy phy\u00adsics expe\u00adri\u00adments, such as at GSI<\/span>, need huge com\u00adpu\u00adte farms for simu\u00adla\u00adti\u00adons. The L\u2011CSC<\/span> clus\u00adter is aimed at QCD<\/span> com\u00adpu\u00adta\u00adti\u00adons, which requi\u00adre extre\u00adme\u00adly high memo\u00adry band\u00adwidth. The imple\u00admen\u00adta\u00adti\u00adon of the Lat\u00adti\u00adce QCD<\/span> algo\u00adrithm was deve\u00adlo\u00adped at FIAS<\/span> using an Open\u00adCL\u2122 based imple\u00admen\u00adta\u00adti\u00adon for por\u00adta\u00adbi\u00adli\u00adty and flexibility.<\/p>\n

AMD<\/span> is exhi\u00adbi\u00adting with AMD<\/span> Fire\u00adPro S9150<\/span> ser\u00adver GPUs and AMD<\/span> ser\u00adver tech\u00adno\u00adlo\u00adgies at SC<\/span> \u201914 in New Orleans Ernest Memo\u00adri\u00adal Con\u00adven\u00adti\u00adon Cen\u00adter booth #839 from Novem\u00adber 17 to 20.<\/p>\n

Supporting Resources<\/h3>\n