U.S. Department of Energy and Cray to Deliver Record-Setting Frontier Supercomputer at ORN

Exascale system expected to be world’s most powerful computer for science and innovation

OAK RIDGE, Tenn., May 7, 2019—The U.S. Depart­ment of Ener­gy today announ­ced a con­tract with Cray Inc. to build the Fron­tier super­com­pu­ter at Oak Ridge Natio­nal Labo­ra­to­ry, which is anti­ci­pa­ted to debut in 2021 as the world’s most power­ful com­pu­ter with a per­for­mance of grea­ter than 1.5 exaflops.

Sche­du­led for deli­very in 2021, Fron­tier will acce­le­ra­te inno­va­ti­on in sci­ence and tech­no­lo­gy and main­tain U.S. lea­ders­hip in high-per­for­mance com­pu­ting and arti­fi­cial intel­li­gence. The total con­tract award is valued at more than $600 mil­li­on for the sys­tem and tech­no­lo­gy deve­lo­p­ment.  The sys­tem will be based on Cray’s new Shas­ta archi­tec­tu­re and Slings­hot inter­con­nect and will fea­ture high-per­for­mance AMD EPYC CPU and AMD Rade­on Instinct GPU technology.

By sol­ving cal­cu­la­ti­ons up to 50 times fas­ter than today’s top supercomputers—exceeding a quin­til­li­on, or 1018, cal­cu­la­ti­ons per second—Frontier will enab­le rese­ar­chers to deli­ver bre­akthroughs in sci­en­ti­fic dis­co­very, ener­gy assuran­ce, eco­no­mic com­pe­ti­ti­ve­ness, and natio­nal secu­ri­ty. As a second-genera­ti­on AI system—following the world-lea­ding Sum­mit sys­tem deploy­ed at ORNL in 2018—Frontier will pro­vi­de new capa­bi­li­ties for deep lear­ning, machi­ne lear­ning and data ana­ly­tics for app­li­ca­ti­ons ran­ging from manu­fac­tu­ring to human health.

Frontier’s record-brea­king per­for­mance will ensu­re our country’s abi­li­ty to lead the world in sci­ence that impro­ves the lives and eco­no­mic pro­spe­ri­ty of all Ame­ri­cans and the ent­i­re world,” said U.S. Secreta­ry of Ener­gy Rick Per­ry. “Fron­tier will acce­le­ra­te inno­va­ti­on in AI by giving Ame­ri­can rese­ar­chers world-class data and com­pu­ting resour­ces to ensu­re the next gre­at inven­ti­ons are made in the United States.”

Sin­ce 2005, Oak Ridge Natio­nal Labo­ra­to­ry has deploy­ed Jagu­ar, Titan, and Sum­mit, each the world’s fas­test com­pu­ter in its time. The com­bi­na­ti­on of tra­di­tio­nal pro­ces­sors with gra­phics pro­ces­sing units to acce­le­ra­te the per­for­mance of lea­ders­hip-class sci­en­ti­fic super­com­pu­ters is an approach pionee­red by ORNL and its part­ners and suc­cess­ful­ly demons­tra­ted through ORNL’s No.1 ran­ked Titan and Sum­mit supercomputers. 

ORNL’s visi­on is to sus­tain the nation’s pree­mi­nence in sci­ence and tech­no­lo­gy by deve­lo­ping and deploy­ing lea­ders­hip com­pu­ting for rese­arch and inno­va­ti­on at an unpre­ce­den­ted sca­le,” said ORNL Direc­tor Tho­mas Zacha­ria. “Fron­tier fol­lows the well-estab­lis­hed com­pu­ting path char­ted by ORNL and its part­ners that will pro­vi­de the rese­arch com­mu­ni­ty with an exas­ca­le sys­tem rea­dy for sci­ence on day one.”

Rese­ar­chers with DOE’s Exas­ca­le Com­pu­ting Pro­ject are deve­lo­ping exas­ca­le sci­en­ti­fic app­li­ca­ti­ons today on ORNL’s 200-peta­flop Sum­mit sys­tem and will seam­less­ly tran­si­ti­on their sci­en­ti­fic app­li­ca­ti­ons to Fron­tier in 2021. In addi­ti­on, the lab’s Cen­ter for Acce­le­ra­ted App­li­ca­ti­on Rea­di­ness is now accep­t­ing pro­po­sals from sci­en­tists to pre­pa­re their codes to run on Frontier.

Rese­ar­chers will harness Frontier’s power­ful archi­tec­tu­re to advan­ce sci­ence in such app­li­ca­ti­ons as sys­tems bio­lo­gy, mate­ri­als sci­ence, ener­gy pro­duc­tion, addi­ti­ve manu­fac­tu­ring and health data sci­ence. Visit the Fron­tier web­site to learn more about what rese­ar­chers plan to accom­plish in the­se and other sci­en­ti­fic fields.

Fron­tier will offer best-in-class tra­di­tio­nal sci­en­ti­fic mode­ling and simu­la­ti­on capa­bi­li­ties while also lea­ding the world in arti­fi­cial intel­li­gence and data ana­ly­tics. Clo­se­ly inte­gra­ting arti­fi­cial intel­li­gence with data ana­ly­tics and mode­ling and simu­la­ti­on will drasti­cal­ly redu­ce the time to dis­co­very by auto­ma­ti­cal­ly reco­gni­zing pat­terns in data and gui­ding simu­la­ti­ons bey­ond the limits of tra­di­tio­nal approaches.

We are hono­red to be part of this his­to­ric moment as we embark on sup­por­ting extre­me-sca­le sci­en­ti­fic endea­vors to deli­ver the next U.S. exas­ca­le super­com­pu­ter to the Depart­ment of Ener­gy and ORNL,” said Peter Unga­ro, pre­si­dent and CEO of Cray. “Fron­tier will incor­po­ra­te foun­da­tio­nal new tech­no­lo­gies from Cray and AMD that will enab­le the new exas­ca­le era—characterized by data-inten­si­ve workloads and the con­ver­gence of mode­ling, simu­la­ti­on, ana­ly­tics, and AI for sci­en­ti­fic dis­co­very, engi­nee­ring and digi­tal transformation.”

Fron­tier will incor­po­ra­te several novel tech­no­lo­gies co-desi­gned spe­ci­fi­cal­ly to deli­ver a balan­ced sci­en­ti­fic capa­bi­li­ty for the user com­mu­ni­ty. The sys­tem will be com­po­sed of more than 100 Cray Shas­ta cabi­nets with high den­si­ty com­pu­te bla­des powe­red by HPC and AI- opti­mi­zed AMD EPYC pro­ces­sors and Rade­on Instinct GPU acce­le­ra­tors pur­po­se-built for the needs of exas­ca­le com­pu­ting. The new acce­le­ra­tor-centric com­pu­te bla­des will sup­port a 4:1 GPU to CPU ratio with high speed AMD Infi­ni­ty Fab­ric links and cohe­rent memo­ry bet­ween them wit­hin the node. Each node will have one Cray Slings­hot inter­con­nect net­work port for every GPU with stream­li­ned com­mu­ni­ca­ti­on bet­ween the GPUs and net­work to enab­le opti­mal per­for­mance for high-per­for­mance com­pu­ting and AI workloads at exascale.

To make this per­for­mance seam­less to con­su­me by deve­lo­pers, Cray and AMD are co-designing and deve­lo­ping enhan­ced GPU pro­gramming tools opti­mi­zed for per­for­mance, pro­duc­ti­vi­ty and por­ta­bi­li­ty. This will inclu­de new capa­bi­li­ties in the Cray Pro­gramming Envi­ron­ment and AMD’s ROCm open com­pu­te plat­form that will be inte­gra­ted tog­e­ther into the Cray Shas­ta soft­ware stack for Frontier.

AMD is proud to be working with Cray, Oak Ridge Natio­nal Labo­ra­to­ry and the Depart­ment of Ener­gy to push the bounda­ries of high per­for­mance com­pu­ting with Fron­tier,” said Lisa Su, AMD pre­si­dent and CEO. “Today’s announ­ce­ment repres­ents the power of col­la­bo­ra­ti­on bet­ween pri­va­te indus­try and public rese­arch insti­tu­ti­ons to deli­ver ground­brea­king inno­va­tions that sci­en­tists can use to sol­ve some of the world’s big­gest problems.”

Fron­tier lever­a­ges a deca­de of exas­ca­le tech­no­lo­gy invest­ments by DOE. The con­tract award inclu­des tech­no­lo­gy deve­lo­p­ment fun­ding, a cen­ter of excel­lence, several ear­ly-deli­very sys­tems, the main Fron­tier sys­tem, and mul­ti-year sys­tems sup­port. The Fron­tier sys­tem is expec­ted to be deli­ve­r­ed in 2021, and accep­t­ance is anti­ci­pa­ted in 2022. 

Fron­tier will be part of the Oak Ridge Lea­ders­hip Com­pu­ting Faci­li­ty, a DOE Office of Sci­ence User Faci­li­ty. ORNL is mana­ged by UT–Battelle for DOE’s Office of Sci­ence, the sin­gle lar­gest sup­por­ter of basic rese­arch in the phy­si­cal sci­en­ces in the United Sta­tes. DOE’s Office of Sci­ence is working to address some of the most pres­sing chal­len­ges of our time. For more infor­ma­ti­on, plea­se visit https://science.energy.gov/.

Durch die weitere Nutzung der Seite stimmst du der Verwendung von Cookies zu. Weitere Informationen

Die Cookie-Einstellungen auf dieser Website sind auf "Cookies zulassen" eingestellt, um das beste Surferlebnis zu ermöglichen. Wenn du diese Website ohne Änderung der Cookie-Einstellungen verwendest oder auf "Akzeptieren" klickst, erklärst du dich damit einverstanden.