AMD Awarded $32 Million for 쁄xtreme Scale High-Performance Computing Research Focused on HSA, APUs and Memory

AMD to Expand Rese짯arch on Exas짯ca짯le Com짯pu짯ters for the Depart짯ment of Energy

SUNNYVALE, CA (Mar짯ket짯wired) 11/14/14 AMD (NYSE: AMD) today announ짯ced that for the third straight year it was award짯ed rese짯arch grants for deve짯lo짯p짯ment of cri짯ti짯cal tech짯no짯lo짯gies nee짯ded for extre짯me-sca짯le com짯pu짯ting in con짯junc짯tion with pro짯jects asso짯cia짯ted with the U.S. Depart짯ment of Ener짯gy (DOE) Extre짯me-Sca짯le Com짯pu짯ting Rese짯arch and Deve짯lo짯p짯ment Pro짯gram, known as 쏤ast짯For짯ward 2. The two DOE awards, tota짯ling more than $32 mil짯li짯on, will fund rese짯arch focu짯sed on exas짯ca짯le appli짯ca짯ti짯ons for AMD Acce짯le짯ra짯ted Pro짯ces짯sing Units (APUs) based on the open-stan짯dard Hete짯ro짯ge짯neous Sys짯tem Archi짯tec짯tu짯re (HSA), as well as future memo짯ry sys짯tems to power a gene짯ra짯ti짯on of exas짯ca짯le super짯com짯pu짯ters capa짯ble of deli짯ve짯ring 3060 times more per짯for짯mance than today셲 fas짯test supercomputers.

Fast짯For짯ward 2 is a joint짯ly-fun짯ded col짯la짯bo짯ra짯ti짯on bet짯ween DOE Office of Sci짯ence and the Natio짯nal Nuclear Secu짯ri짯ty Admi짯nis짯tra짯ti짯on (NNSA) focu짯sed on initia짯ting part짯ner짯ships with mul짯ti짯ple com짯pa짯nies to acce짯le짯ra짯te the rese짯arch and deve짯lo짯p짯ment of cri짯ti짯cal tech짯no짯lo짯gies nee짯ded to enable exas짯ca짯le com짯pu짯ting. Spe짯ci짯fi짯cal짯ly, the Fast짯For짯ward 2 rese짯arch grants award짯ed to AMD are for:

  • Node Archi짯tec짯tu짯re: AMD will con짯duct rese짯arch for an inte짯gra짯ted exas짯ca짯le node archi짯tec짯tu짯re based on its HSA-enab짯led APUs. Par짯ti짯cu짯lar are짯as of rese짯arch include power effi짯ci짯en짯cy, relia짯bi짯li짯ty, pro짯gramma짯bi짯li짯ty, com짯po짯nent and net짯work inter짯face inte짯gra짯ti짯on, APU micro짯ar짯chi짯tec짯tu짯re, advan짯ced memo짯ry archi짯tec짯tures and effi짯ci짯ent data movement.
  • Memo짯ry Tech짯no짯lo짯gy: AMD will col짯la짯bo짯ra짯te with the DOE and others to help defi짯ne a new stan짯dard for memo짯ry inter짯faces that meets the needs of future-gene짯ra짯ti짯on memo짯ry devices, inclu짯ding non-vola짯ti짯le memo짯ry and pro짯ces짯sing-in-memo짯ry (PIM) architectures.

We쇑춓e hono짯red to be sel짯ec짯ted once again by DOE to aid rese짯arch efforts for exas짯ca짯le-com짯pu짯ting, said Alan Lee, AMD셲 cor짯po짯ra짯te vice pre짯si짯dent for Rese짯arch and Advan짯ced Deve짯lo짯p짯ment. 쏰ur cus짯to짯mers have long loo짯ked to AMD to deve짯lop inno짯va짯ti짯ve com짯pu짯te, gra짯phics, and memo짯ry tech짯no짯lo짯gies and pro짯ducts that sol짯ve real-world pro짯blems. This rese짯arch is focu짯sed on ener짯gy-effi짯ci짯ent node archi짯tec짯tures and memo짯ry sys짯tems to impro짯ve the capa짯bi짯li짯ty of the world셲 fas짯test supercomputers.

The DOE stra짯te짯gic plan seeks to address the nation셲 most pres짯sing sci짯en짯ti짯fic chal짯lenges by advan짯cing simu짯la짯ti짯on-based sci짯en짯ti짯fic dis짯co짯very made pos짯si짯ble by the world셲 hig짯hest per짯forming exas짯ca짯le super짯com짯pu짯ters. The tech짯no짯lo짯gies deve짯lo짯ped as part of this initia짯ti짯ve could posi짯tively impact low-power embedded, cloud/datacenter and mid-ran짯ge high-per짯for짯mance com짯pu짯ting appli짯ca짯ti짯ons for diver짯se fields ran짯ging from medi짯cal sci짯ence to astro짯phy짯sics and cli짯ma짯te mode짯ling. Fast짯For짯ward 2 invest짯ments will result in an increase in DOE셲 abili짯ty to levera짯ge com짯mer짯cial deve짯lo짯p짯ments for future systems.

AMD Opte짯ron꽓 com짯pu짯te pro짯ces짯sors and AMD Fire짯Pro꽓 gra짯phics pro짯ces짯sors are used today in many of the world셲 lea짯ding super짯com짯pu짯ters.

About AMD
AMD (NYSE: AMD) designs and inte짯gra짯tes tech짯no짯lo짯gy that powers mil짯li짯ons of intel짯li짯gent devices, inclu짯ding per짯so짯nal com짯pu짯ters, tablets, game con짯so짯les and cloud ser짯vers that defi짯ne the new era of sur짯round com짯pu짯ting. AMD solu짯ti짯ons enable peo짯p짯le ever짯y짯whe짯re to rea짯li짯ze the full poten짯ti짯al of their favo짯ri짯te devices and appli짯ca짯ti짯ons to push the boun짯da짯ries of what is pos짯si짯ble. For more infor짯ma짯ti짯on, visit www.amd.com.

AMD, the AMD Arrow logo, AMD Opte짯ron, AMD Fire짯Pro and com짯bi짯na짯ti짯ons the짯reof, are trade짯marks of Advan짯ced Micro Devices, Inc. Other names are for infor짯ma짯tio짯nal pur짯po짯ses only and may be trade짯marks of their respec짯ti짯ve owners.