{"id":18020,"date":"2015-07-08T14:27:03","date_gmt":"2015-07-08T12:27:03","guid":{"rendered":"http:\/\/www.planet3dnow.de\/cms\/?p=18020"},"modified":"2015-07-08T14:27:03","modified_gmt":"2015-07-08T12:27:03","slug":"amd-delivers-worlds-first-server-gpu-with-industry-leading-32gb-memory-for-high-performance-compute","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/18020-amd-delivers-worlds-first-server-gpu-with-industry-leading-32gb-memory-for-high-performance-compute\/","title":{"rendered":"AMD<\/span> Delivers World\u2019s First Server GPU<\/span> with Industry-Leading 32GB<\/span> Memory for High Performance Compute"},"content":{"rendered":"

Mit der neu\u00aden AMD<\/span> Fire\u00adPro\u2122 S9170<\/span> Gra\u00adfik\u00adkar\u00adte pr\u00e4\u00adsen\u00adtiert AMD<\/span> die welt\u00adweit ers\u00adte und schnells\u00adte Ein\u00adzel-Gra\u00adfik\u00adkar\u00adte f\u00fcr Work\u00adsta\u00adtions mit 32 GB<\/span> Spei\u00adcher und Open\u00adCL\u2122 2.0 Unter\u00adst\u00fct\u00adzung. Die AMD<\/span> Fire\u00adPro\u2122 S9170<\/span> Gra\u00adfik\u00adkar\u00adte basiert auf der Archi\u00adtek\u00adtur der zwei\u00adten Gene\u00adra\u00adti\u00adon der AMD<\/span> Gra\u00adphics Core Next (GCN<\/span>) Gra\u00adfik\u00adpro\u00adzes\u00adso\u00adren und ist in der Lage in der Spit\u00adze bis zu 5,24 TFLOPS<\/span> Rechen\u00adleis\u00adtung (sin\u00adgle-pre\u00adcis\u00adi\u00adon) zu erbrin\u00adgen. Bei dua\u00adler Genau\u00adig\u00adkeit (dou\u00adble-pre\u00adcis\u00adi\u00adon) erreicht die Gra\u00adfik\u00adkar\u00adte Spit\u00adzen\u00adwer\u00adte von bis zu 2,62 TFLOPS<\/span>.<\/span><\/p>\n

AMD<\/span> Deli\u00advers World\u2019s First Ser\u00adver GPU<\/span> with Indus\u00adtry-Lea\u00adding 32GB<\/span> Memo\u00adry for High Per\u00adfor\u00admance Compute<\/strong><\/p>\n

\u2014 AMD<\/span> Fire\u00adPro\u2122 S9170<\/span> Ser\u00adver GPU<\/span> Offers Unmat\u00adched Onboard Memo\u00adry to Sup\u00adport Lar\u00adge Data\u00adset Computations \u2014<\/em><\/p>\n

Sun\u00adny\u00adva\u00adle, CA<\/span>. \u2014 July 8, 2015 \u2014<\/strong> AMD<\/span><\/a> (NASDAQ<\/span>: AMD<\/span>) today announ\u00adced the new AMD<\/span> Fire\u00adPro\u2122 S9170<\/span> ser\u00adver GPU<\/span>, the world\u2019s first and fas\u00adtest 32GB<\/span> sin\u00adgle-GPU ser\u00adver card for DGEMM<\/span> hea\u00advy dou\u00adble-pre\u00adcis\u00adi\u00adon workloads1<\/sup>, with sup\u00adport for Open\u00adCL\u2122 2.0. Based on the second-gene\u00adra\u00adti\u00adon AMD<\/span> Gra\u00adphics Core Next (GCN<\/span>) GPU<\/span> archi\u00adtec\u00adtu\u00adre, this new addi\u00adti\u00adon to the AMD<\/span> Fire\u00adPro\u2122 ser\u00adver GPU<\/span> fami\u00adly is capa\u00adble of deli\u00adve\u00adring up to 5.24 TFLOPS<\/span> of peak sin\u00adgle pre\u00adcis\u00adi\u00adon com\u00adpu\u00adte per\u00adfor\u00admance while enab\u00adling full through\u00adput dou\u00adble pre\u00adcis\u00adi\u00adon per\u00adfor\u00admance, pro\u00advi\u00adding up to 2.62 TFLOPS<\/span> of peak dou\u00adble pre\u00adcis\u00adi\u00adon performance.<\/p>\n

Desi\u00adgned with com\u00adpu\u00adte-inten\u00adsi\u00adve work\u00adflows in mind, the AMD<\/span> Fire\u00adPro S9170<\/span> ser\u00adver GPU<\/span> is ide\u00adal for data cen\u00adter mana\u00adgers who over\u00adsee clus\u00adters within aca\u00adde\u00admic or govern\u00adment bodies, oil and gas indus\u00adtries, or deep neu\u00adral net\u00adwork com\u00adpu\u00adte clus\u00adter development.<\/p>\n

\u201c<\/span>AMD<\/span> is reco\u00adgni\u00adzed as an HPC<\/span> indus\u00adtry inno\u00adva\u00adtor as the gra\u00adphics pro\u00advi\u00adder with the top spot on the Novem\u00adber 2014 Green500 List.<\/a> Today the best GPU<\/span> for com\u00adpu\u00adte just got bet\u00adter with the intro\u00adduc\u00adtion of the AMD<\/span> Fire\u00adPro S9170<\/span> ser\u00adver GPU<\/span> to com\u00adple\u00adment AMD<\/span>\u2019s impres\u00adsi\u00adve array of ser\u00adver gra\u00adphics offe\u00adrings for high per\u00adfor\u00admance com\u00adpu\u00adte envi\u00adron\u00adments,\u201d said Sean Bur\u00adke, cor\u00adpo\u00adra\u00adte vice pre\u00adsi\u00addent and gene\u00adral mana\u00adger, AMD<\/span> Pro\u00adfes\u00adsio\u00adnal Gra\u00adphics group. \u201cThe AMD<\/span> Fire\u00adPro S9170<\/span> ser\u00adver GPU<\/span> can acce\u00adle\u00adra\u00adte com\u00adplex workloads in sci\u00aden\u00adti\u00adfic com\u00adpu\u00adting, data ana\u00adly\u00adtics, or seis\u00admic pro\u00adces\u00adsing, wiel\u00adding an indus\u00adtry-lea\u00adding 32GB<\/span> of memo\u00adry. We desi\u00adgned the new offe\u00adring for super\u00adcom\u00adpu\u00adters to achie\u00adve mas\u00adsi\u00adve com\u00adpu\u00adte per\u00adfor\u00admance while maxi\u00admi\u00adzing available power budgets.\u201d<\/p>\n

\u201c<\/span>The\u00adre are some HPC<\/span> workloads which requi\u00adre as much data as pos\u00adsi\u00adble to stay resi\u00addent on the device, and so the 32GB<\/span> of memo\u00adry pro\u00advi\u00added by AMD<\/span> Fire\u00adPro S9170<\/span>, the lar\u00adgest available on a sin\u00adgle GPU<\/span>, will enable the acce\u00adle\u00adra\u00adti\u00adon of sci\u00aden\u00adti\u00adfic cal\u00adcu\u00adla\u00adti\u00adons that were pre\u00advious\u00adly impos\u00adsi\u00adble,\u201d said Simon McIn\u00adtosh-Smith, head of the Microelec\u00adtro\u00adnics Rese\u00adarch Group at the Uni\u00adver\u00adsi\u00adty of Bris\u00adtol. \u201cFor exam\u00adp\u00adle, our new Open\u00adCL ver\u00adsi\u00adon of the SNAP<\/span> trans\u00adport code from Los Ala\u00admos Natio\u00adnal Labo\u00adra\u00adto\u00adry needs to keep as much data resi\u00addent on the device as pos\u00adsi\u00adble, and so the 32GB<\/span> of memo\u00adry will let us run pro\u00adblems of a much more inte\u00adres\u00adt\u00ading size fas\u00adter than ever befo\u00adre. The lar\u00adge memo\u00adry, com\u00adbi\u00adned with the 320GB<\/span>\/s memo\u00adry band\u00adwidth and dou\u00adble pre\u00adcis\u00adi\u00adon floa\u00adting point per\u00adfor\u00admance, will make the AMD<\/span> Fire\u00adPro S9170<\/span> ser\u00adver GPU<\/span> a \u2018kil\u00adler\u2019 solu\u00adti\u00adon device for many HPC<\/span> applications.\u201d<\/p>\n

The AMD<\/span> Fire\u00adPro S9170<\/span> offers unpar\u00adal\u00adle\u00adled GPU<\/span> com\u00adpu\u00adte per\u00adfor\u00admance and power effi\u00adci\u00aden\u00adcy with the fol\u00adlo\u00adwing benefits:<\/p>\n