{"id":65396,"date":"2022-05-31T16:51:04","date_gmt":"2022-05-31T14:51:04","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=65396"},"modified":"2022-05-31T16:51:04","modified_gmt":"2022-05-31T14:51:04","slug":"ayar-labs-to-accelerate-development-and-application-of-optical-interconnects-in-artificial-intelligence-machine-learning-architectures-with-nvidia","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/65396-ayar-labs-to-accelerate-development-and-application-of-optical-interconnects-in-artificial-intelligence-machine-learning-architectures-with-nvidia\/","title":{"rendered":"Ayar Labs to Accelerate Development and Application of Optical Interconnects in Artificial Intelligence\/Machine Learning Architectures with NVIDIA<\/span>"},"content":{"rendered":"

SANTA<\/span> CLARA<\/span>, Calif. \u2013<\/strong>May 25, 2022 \u2013 Ayar Labs<\/a>, the lea\u00adder in chip-to-chip opti\u00adcal con\u00adnec\u00adti\u00advi\u00adty, is deve\u00adlo\u00adping with NVIDIA<\/span> ground\u00adbrea\u00adking arti\u00adfi\u00adci\u00adal intel\u00adli\u00adgence (AI<\/span>) infra\u00adstruc\u00adtu\u00adre based on opti\u00adcal I\/O tech\u00adno\u00adlo\u00adgy to meet future demands of AI<\/span> and high per\u00adfor\u00admance com\u00adpu\u00adting (HPC<\/span>) workloads. The col\u00adla\u00adbo\u00adra\u00adti\u00adon will focus on inte\u00adgra\u00adting Ayar Labs\u2019 tech\u00adno\u00adlo\u00adgy to deve\u00adlop sca\u00adle-out archi\u00adtec\u00adtures enab\u00adled by high-band\u00adwidth, low-laten\u00adcy and ultra-low-power opti\u00adcal-based inter\u00adcon\u00adnects for future NVIDIA<\/span> pro\u00adducts. Tog\u00ade\u00adther, the com\u00adpa\u00adnies plan to acce\u00adle\u00adra\u00adte the deve\u00adlo\u00adp\u00adment and adop\u00adti\u00adon of opti\u00adcal I\/O tech\u00adno\u00adlo\u00adgy to sup\u00adport the explo\u00adsi\u00adve growth of AI<\/span> and machi\u00adne lear\u00adning (ML<\/span>) appli\u00adca\u00adti\u00adons and data volumes.<\/p>\n

Mee\u00adting Future Per\u00adfor\u00admance and Power Requi\u00adre\u00adments with Opti\u00adcal I\/O<\/strong><\/p>\n

Opti\u00adcal I\/O uni\u00adque\u00adly chan\u00adges the per\u00adfor\u00admance and power tra\u00adjec\u00adto\u00adries of sys\u00adtem designs by enab\u00adling com\u00adpu\u00adte, memo\u00adry and net\u00adwor\u00adking ASICs to com\u00admu\u00adni\u00adca\u00adte with dra\u00adma\u00adti\u00adcal\u00adly increased band\u00adwidth, at lower laten\u00adcy, over lon\u00adger distances and at a frac\u00adtion of the power of exis\u00adting elec\u00adtri\u00adcal I\/O solu\u00adti\u00adons. The tech\u00adno\u00adlo\u00adgy is also foun\u00adda\u00adtio\u00adnal to enab\u00adling emer\u00adging hete\u00adro\u00adge\u00adneous com\u00adpu\u00adte sys\u00adtems, disaggregated\/pooled designs, and uni\u00adfied memo\u00adry archi\u00adtec\u00adtures that are cri\u00adti\u00adcal to acce\u00adle\u00adra\u00adting future data cen\u00adter innovation.<\/p>\n

\u201c<\/span>Today\u2019s sta\u00adte-of-the-art AI<\/span>\/ML<\/span> trai\u00adning archi\u00adtec\u00adtures are limi\u00adt\u00aded by cur\u00adrent cop\u00adper-based com\u00adpu\u00adte-to-com\u00adpu\u00adte inter\u00adcon\u00adnects to build sca\u00adle-out sys\u00adtems for tomorrow\u2019s requi\u00adre\u00adments,\u201d said Charles Wuisch\u00adpard, CEO<\/span> of Ayar Labs. \u201cOur work with NVIDIA<\/span> to deve\u00adlop next-gene\u00adra\u00adti\u00adon solu\u00adti\u00adons based on opti\u00adcal I\/O pro\u00advi\u00addes the foun\u00adda\u00adti\u00adon for the next leap in AI<\/span> capa\u00adbi\u00adli\u00adties to address the world\u2019s most sophisti\u00adca\u00adted problems.\u201d<\/p><\/blockquote>\n

Deli\u00adve\u00adring the \u2018Next Million\u2011X\u2019 Spee\u00addup for AI<\/span> with Opti\u00adcal Interconnect
\n<\/strong><\/p>\n

\u201c<\/span>Over the past deca\u00adde, NVI\u00adDIA-acce\u00adle\u00adra\u00adted com\u00adpu\u00adting has deli\u00adver\u00aded a million\u2011X spee\u00addup in AI<\/span>,\u201d said Rob Ober, Chief Plat\u00adform Archi\u00adtect for Data Cen\u00adter Pro\u00adducts at NVIDIA<\/span>. \u201cThe next million\u2011X will requi\u00adre new, advan\u00adced tech\u00adno\u00adlo\u00adgies like opti\u00adcal I\/O to sup\u00adport the band\u00adwidth, power and sca\u00adle requi\u00adre\u00adments of future AI<\/span> and ML<\/span> workloads and sys\u00adtem architectures.\u201d<\/p><\/blockquote>\n

As AI<\/span> model sizes con\u00adti\u00adnue to grow, by 2023 NVIDIA<\/span> belie\u00adves that models will have 100 tril\u00adli\u00adon or more con\u00adnec\u00adtions<\/a> \u2013 a 600X<\/span> increase from 2021 \u2013 excee\u00adding the tech\u00adni\u00adcal capa\u00adbi\u00adli\u00adties of exis\u00adting plat\u00adforms. Tra\u00addi\u00adtio\u00adnal elec\u00adtri\u00adcal-based inter\u00adcon\u00adnects will reach their band\u00adwidth limits, dri\u00adving lower appli\u00adca\u00adti\u00adon per\u00adfor\u00admance, hig\u00adher laten\u00adcy and increased power con\u00adsump\u00adti\u00adon. New inter\u00adcon\u00adnect solu\u00adti\u00adons and sys\u00adtem archi\u00adtec\u00adtures are nee\u00added to address the sca\u00adle, per\u00adfor\u00admance and power demands of the next gene\u00adra\u00adti\u00adon of AI<\/span>. Ayar Labs\u2019 col\u00adla\u00adbo\u00adra\u00adti\u00adon with NVIDIA<\/span> is focu\u00adsed on addres\u00adsing the\u00adse future chal\u00adlenges by deve\u00adlo\u00adping next-gene\u00adra\u00adti\u00adon archi\u00adtec\u00adtures with opti\u00adcal I\/O.<\/p>\n

To learn more about Ayar Labs\u2019 chip-to-chip opti\u00adcal tech\u00adno\u00adlo\u00adgy, plea\u00adse visit: https:\/\/ayarlabs.com\/<\/a><\/p>\n

About Ayar Labs<\/strong><\/p>\n

Ayar Labs is dis\u00adrupt\u00ading the tra\u00addi\u00adtio\u00adnal per\u00adfor\u00admance, cost, and effi\u00adci\u00aden\u00adcy cur\u00adves of the semi\u00adcon\u00adduc\u00adtor and com\u00adpu\u00adting indus\u00adtries by dri\u00adving a 1000x impro\u00adve\u00adment in inter\u00adcon\u00adnect band\u00adwidth den\u00adsi\u00adty at 10x lower power.  Ayar Labs\u2019 paten\u00adted approach uses indus\u00adtry stan\u00addard cost-effec\u00adti\u00adve sili\u00adcon pro\u00adces\u00adsing tech\u00adni\u00adques to deve\u00adlop high speed, high den\u00adsi\u00adty, low power opti\u00adcal-based inter\u00adcon\u00adnect \u201cchip\u00adlets\u201d and lasers to replace tra\u00addi\u00adtio\u00adnal elec\u00adtri\u00adcal based I\/O. The com\u00adpa\u00adny was foun\u00added in 2015 and is fun\u00added by a num\u00adber of dome\u00adstic and inter\u00adna\u00adtio\u00adnal ven\u00adture capi\u00adtal firms as well as stra\u00adte\u00adgic inves\u00adtors.  For more infor\u00adma\u00adti\u00adon, visit www.ayarlabs.com<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"

SANTA<\/span> CLARA<\/span>, Calif. \u2013<\/strong>May 25, 2022 \u2013 Ayar Labs<\/a>, the lea\u00adder in chip-to-chip opti\u00adcal con\u00adnec\u00adti\u00advi\u00adty, is deve\u00adlo\u00adping with NVIDIA<\/span> ground\u00adbrea\u00adking arti\u00adfi\u00adci\u00adal intel\u00adli\u00adgence (AI<\/span>) infra\u00adstruc\u00adtu\u00adre based on opti\u00adcal I\/O tech\u00adno\u00adlo\u00adgy to meet future demands of AI<\/span> and high per\u00adfor\u00admance com\u00adpu\u00adting (HPC<\/span>) workloads. The col\u00adla\u00adbo\u00adra\u00adti\u00adon will focus on inte\u00adgra\u00adting Ayar Labs\u2019 tech\u00adno\u00adlo\u00adgy to deve\u00adlop sca\u00adle-out archi\u00adtec\u00adtures enab\u00adled by high-band\u00adwidth, low-laten\u00adcy and ultra-low-power opti\u00adcal-based inter\u00adcon\u00adnects for future NVIDIA<\/span> pro\u00adducts. Tog\u00ade\u00adther, the com\u00adpa\u00adnies plan to acce\u00adle\u00adra\u00adte the deve\u00adlo\u00adp\u00adment and adop\u00adti\u00adon of opti\u00adcal I\/O tech\u00adno\u00adlo\u00adgy to sup\u00adport the explo\u00adsi\u00adve growth of AI<\/span> and machi\u00adne lear\u00adning (ML<\/span>) appli\u00adca\u00adti\u00adons and data volu\u00admes. (\u2026) Wei\u00adter\u00adle\u00adsen \u00bb<\/a><\/p>\n","protected":false},"author":593,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wp_typography_post_enhancements_disabled":false,"ngg_post_thumbnail":0,"footnotes":""},"categories":[22],"tags":[2735,992,2736],"class_list":["post-65396","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","tag-ayar-labs","tag-nvidia","tag-optical-i-o","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/65396","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/users\/593"}],"replies":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/comments?post=65396"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/65396\/revisions"}],"predecessor-version":[{"id":65397,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/65396\/revisions\/65397"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=65396"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=65396"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=65396"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}