Optimized for AI Accelerator Applications, GLOBALFOUNDRIES 12LP+ FinFET Solution Ready for Production

Built on a pro­ven plat­form with a robust pro­duc­tion eco­sys­tem, 12LP+ offers AI appli­ca­ti­on desi­gners an effi­ci­ent deve­lo­p­ment expe­ri­ence and fast time-to-market

San­ta Cla­ra, Calif., June 30, 2020 – GLOBALFOUNDRIES® (GF®), the world’s lea­ding spe­cial­ty foundry, today announ­ced its most advan­ced Fin­FET solu­ti­on, 12LP+, has com­ple­ted tech­no­lo­gy qua­li­fi­ca­ti­on and is rea­dy for production.

GF’s dif­fe­ren­tia­ted 12LP+ solu­ti­on is opti­mi­zed for arti­fi­ci­al intel­li­gence (AI) trai­ning and infe­rence appli­ca­ti­ons. Built on a pro­ven plat­form with a robust pro­duc­tion eco­sys­tem, 12LP+ offers chip desi­gners an effi­ci­ent deve­lo­p­ment expe­ri­ence and a fast time-to-market.

Con­tri­bu­ting to its best-in-class com­bi­na­ti­on of per­for­mance, power and area, 12LP+ intro­du­ces new fea­tures inclu­ding an updated stan­dard cell libra­ry, an inter­po­ser for 2.5D pack­a­ging, and a low-power 0.5V Vmin SRAM bit­cell that sup­ports the low laten­cy and power-effi­ci­ent shut­t­ling of data bet­ween the AI pro­ces­sors and memo­ry. The result is a semi­con­duc­tor solu­ti­on engi­nee­red to meet the spe­ci­fic needs of the fast-gro­wing AI market.

Arti­fi­ci­al intel­li­gence is on a tra­jec­to­ry to beco­me the most dis­rup­ti­ve tech­no­lo­gy of our life­time,” said Amir Fain­tuch, seni­or vice pre­si­dent and gene­ral mana­ger of Com­pu­ting and Wired Infra­struc­tu­re at GF. “It is incre­asing­ly clear that the power effi­ci­en­cy of AI sys­tems – in par­ti­cu­lar how many ope­ra­ti­ons you can wrest from a watt of power – will be among the most cri­ti­cal fac­tors a com­pa­ny con­siders when deci­ding to invest in data cen­ters or edge AI appli­ca­ti­ons. Our new 12LP+ solu­ti­on tack­les this chall­enge head-on. It has been engi­nee­red and opti­mi­zed, obses­si­ve­ly so, with AI in mind.”

12LP+ builds upon GF’s estab­lished 14nm/12LP plat­form, of which GF has ship­ped more than one mil­li­on wafers. GF’s 12LP is being used by com­pa­nies inclu­ding Enfla­meTen­stor­rent, and others for AI acce­le­ra­tor appli­ca­ti­ons. By part­ne­ring clo­se­ly and lear­ning from AI cli­ents, GF deve­lo­ped 12LP+ to pro­vi­de grea­ter dif­fe­ren­tia­ti­on and increased value for desi­gners in the AI space while mini­mi­zing their deve­lo­p­ment and pro­duc­tion costs.

Dri­ving the enhan­ced per­for­mance of 12LP+ are fea­tures inclu­ding a 20-per­cent SoC-level logic per­for­mance boost over 12LP, and a 10-per­cent impro­ve­ment in logic area sca­ling. The­se advance­ments are achie­ved in 12LP+ through its next-gene­ra­ti­on stan­dard cell libra­ry with per­for­mance-dri­ven area opti­mi­zed com­pon­ents, sin­gle Fin cells, a new low-vol­ta­ge SRAM bit­cell, and impro­ved ana­log lay­out design rules.

12LP+, a spe­cia­li­zed appli­ca­ti­on solu­ti­on, is aug­men­ted by GF’s AI design refe­rence packa­ge, as well as GF’s co-deve­lo­p­ment, pack­a­ging, and post-fab turn­key ser­vices – which tog­e­ther enable a holi­stic expe­ri­ence for desig­ning low-power, cost-effec­ti­ve cir­cuits opti­mi­zed for AI appli­ca­ti­ons. Clo­se col­la­bo­ra­ti­on bet­ween GF and its eco­sys­tem part­ners results in cost-effec­ti­ve deve­lo­p­ment cos­ts and a quicker time to market.

In addi­ti­on to 12LP’s exis­ting IP port­fo­lio, GF will expand the IP vali­da­ti­ons for 12LP+ to include PCIe 3/4/5 and USB 2/3 to host pro­ces­sors, HBM2/2e, DDR/LPDDR4/4x and GDDR6 to exter­nal memo­ry, and chip-to-chip inter­con­nect for desi­gners and cli­ents pur­suing chip­let architectures.

GFs’ 12LP+ solu­ti­on has been qua­li­fied and is curr­ent­ly rea­dy for pro­duc­tion at GF’s Fab 8 in Mal­ta, New York. Seve­ral 12LP+ tape-outs are sche­du­led for the second half of 2020. GF recent­ly announ­ced it would bring its Fab 8 faci­li­ty into com­pli­ance with both the U.S. Inter­na­tio­nal Traf­fic in Arms Regu­la­ti­ons (ITAR) stan­dards and the Export Admi­nis­tra­ti­on Regu­la­ti­ons (EAR). The­se new con­trol assu­ran­ces, which will go into effect later this year, will make con­fi­den­tia­li­ty and inte­gri­ty pro­tec­tions available for defen­se-rela­ted appli­ca­ti­ons, devices or com­pon­ents manu­fac­tu­red at Fab 8.

Click here to learn more about GF’s 12LP 12nm Fin­FET technology.

About GF

GLOBALFOUNDRIES (GF) is the world’s lea­ding spe­cial­ty foundry. GF deli­vers dif­fe­ren­tia­ted fea­ture-rich solu­ti­ons that enable its cli­ents to deve­lop inno­va­ti­ve pro­ducts for high-growth mar­ket seg­ments. GF pro­vi­des a broad ran­ge of plat­forms and fea­tures with a uni­que mix of design, deve­lo­p­ment and fabri­ca­ti­on ser­vices. With an at-sca­le manu­fac­tu­ring foot­print span­ning the U.S., Euro­pe and Asia, GF has the fle­xi­bi­li­ty and agi­li­ty to meet the dyna­mic needs of cli­ents across the glo­be. GF is owned by Muba­d­a­la Invest­ment Com­pa­ny. For more infor­ma­ti­on, visit www.globalfoundries.com.