{"id":67514,"date":"2023-01-10T19:57:17","date_gmt":"2023-01-10T18:57:17","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=67514"},"modified":"2023-01-10T19:57:17","modified_gmt":"2023-01-10T18:57:17","slug":"the-greenest-generation-nvidia-intel-and-partners-supercharge-ai-computing-efficiency","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/67514-the-greenest-generation-nvidia-intel-and-partners-supercharge-ai-computing-efficiency\/","title":{"rendered":"The Greenest Generation: NVIDIA<\/span>, Intel and Partners Supercharge AI<\/span> Computing Efficiency"},"content":{"rendered":"

Acce\u00adle\u00adra\u00adted NVIDIA<\/span> Hop\u00adper sys\u00adtems with 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors \u2014 inclu\u00adding NVIDIA<\/span> DGX<\/span> H100<\/span> and 60+ sys\u00adtems from NVIDIA<\/span> part\u00adners \u2014 pro\u00advi\u00adde 25x more effi\u00adci\u00aden\u00adcy than tra\u00addi\u00adtio\u00adnal data cen\u00adter ser\u00advers to save big on ener\u00adgy costs.<\/p>\n

AI<\/span> is at the heart of humanity\u2019s most trans\u00adfor\u00adma\u00adti\u00adve inno\u00adva\u00adtions \u2014 from deve\u00adlo\u00adping COVID<\/span> vac\u00adci\u00adnes at unpre\u00adce\u00adden\u00adted speeds and dia\u00adgno\u00adsing can\u00adcer to powe\u00adring auto\u00adno\u00admous vehic\u00adles and under\u00adstan\u00adding cli\u00adma\u00adte change.<\/p>\n

Vir\u00adtual\u00adly every indus\u00adtry will bene\u00adfit from adop\u00adting AI<\/span>, but the tech\u00adno\u00adlo\u00adgy has beco\u00adme more resour\u00adce inten\u00adsi\u00adve as neu\u00adral net\u00adworks have increased in com\u00adple\u00adxi\u00adty. To avo\u00adid pla\u00adcing unsus\u00adtainable demands on elec\u00adtri\u00adci\u00adty gene\u00adra\u00adti\u00adon to run this com\u00adpu\u00adting infra\u00adstruc\u00adtu\u00adre, the under\u00adly\u00ading tech\u00adno\u00adlo\u00adgy must be as effi\u00adci\u00adent as possible.<\/p>\n

Acce\u00adle\u00adra\u00adted com\u00adpu\u00adting powered by NVIDIA<\/span> GPUs and the NVIDIA<\/span> AI<\/span> plat\u00adform offer the effi\u00adci\u00aden\u00adcy that enables data cen\u00adters to sus\u00adtain\u00adab\u00adly dri\u00adve the next gene\u00adra\u00adti\u00adon of breakthroughs.<\/p>\n

And now, timed with the launch of 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors, NVIDIA<\/span> and its part\u00adners have kicked off a new gene\u00adra\u00adti\u00adon of acce\u00adle\u00adra\u00adted com\u00adpu\u00adting sys\u00adtems that are built for ener\u00adgy-effi\u00adci\u00adent AI<\/span>. When com\u00adbi\u00adned with NVIDIA<\/span> H100<\/span> Ten\u00adsor Core GPUs<\/a>, the\u00adse sys\u00adtems can deli\u00adver dra\u00adma\u00adti\u00adcal\u00adly hig\u00adher per\u00adfor\u00admance, grea\u00adter sca\u00adle and hig\u00adher effi\u00adci\u00aden\u00adcy than the pri\u00ador gene\u00adra\u00adti\u00adon, pro\u00advi\u00adding more com\u00adpu\u00adta\u00adti\u00adon and pro\u00adblem-sol\u00adving per watt.<\/p>\n

The new Intel CPUs will be used in NVIDIA<\/span> DGX<\/span> H100<\/span> sys\u00adtems<\/a>, as well as in more than 60 ser\u00advers fea\u00adturing H100<\/span> GPUs from NVIDIA<\/span> part\u00adners around the world.<\/p>\n

Supercharging Speed, Efficiency and Savings for Enterprise AI<\/span><\/b><\/h2>\n

The coming NVIDIA<\/span> and Intel-powered sys\u00adtems will help enter\u00adpri\u00adses run workloads an avera\u00adge of 25x more effi\u00adci\u00adent\u00adly than tra\u00addi\u00adtio\u00adnal CPU-only data cen\u00adter ser\u00advers. This incre\u00addi\u00adble per\u00adfor\u00admance per watt means less power is nee\u00added to get jobs done, which helps ensu\u00adre the power available to data cen\u00adters is used as effi\u00adci\u00adent\u00adly as pos\u00adsi\u00adble to super\u00adchar\u00adge the most important work.<\/p>\n

Com\u00adpared to pri\u00ador-gene\u00adra\u00adti\u00adon acce\u00adle\u00adra\u00adted sys\u00adtems, this new gene\u00adra\u00adti\u00adon of NVI\u00adDIA-acce\u00adle\u00adra\u00adted ser\u00advers speed trai\u00adning and infe\u00adrence to boost ener\u00adgy effi\u00adci\u00aden\u00adcy<\/a> by 3.5x \u2013 which trans\u00adla\u00adtes into real cost savings, with AI<\/span> data cen\u00adters deli\u00adve\u00adring over 3x lower total cost of ownership.<\/p>\n

New 4th Gen Intel Xeon CPUs Move More Data to Accelerate NVIDIA<\/span> AI<\/span><\/b><\/h2>\n

Among the fea\u00adtures of the new 4th Gen Intel Xeon CPU<\/span> is sup\u00adport for PCIe Gen 5, which can dou\u00adble the data trans\u00adfer rates from CPU<\/span> to NVIDIA<\/span> GPUs and net\u00adwor\u00adking. Increased PCIe lanes allow for a grea\u00adter den\u00adsi\u00adty of GPUs and high-speed net\u00adwor\u00adking within each server.<\/p>\n

Fas\u00adter memo\u00adry band\u00adwidth also impro\u00adves the per\u00adfor\u00admance of data-inten\u00adsi\u00adve workloads such as AI<\/span>, while net\u00adwor\u00adking speeds \u2014 up to 400 giga\u00adbits per second (Gbps) per con\u00adnec\u00adtion \u2014 sup\u00adport fas\u00adter data trans\u00adfers bet\u00adween ser\u00advers and storage.<\/p>\n

NVIDIA<\/span> DGX<\/span> H100<\/span> sys\u00adtems and ser\u00advers from NVIDIA<\/span> part\u00adners with H100<\/span> PCIe GPUs come with a licen\u00adse for NVIDIA<\/span> AI<\/span> Enter\u00adpri\u00adse<\/a>, an end-to-end, secu\u00adre, cloud-nati\u00adve suite of AI<\/span> deve\u00adlo\u00adp\u00adment and deploy\u00adment soft\u00adware, pro\u00advi\u00adding a com\u00adple\u00adte plat\u00adform for excel\u00adlence in effi\u00adci\u00adent enter\u00adpri\u00adse AI<\/span>.<\/p>\n

NVIDIA<\/span> DGX<\/span> H100<\/span> Systems Supercharge Efficiency for Supersize AI<\/span><\/b><\/h2>\n

Each DGX<\/span> H100<\/span> sys\u00adtem fea\u00adtures eight NVIDIA<\/span> H100<\/span> GPUs, 10 NVIDIA<\/span> ConnectX\u20117<\/a> net\u00adwork adap\u00adters and dual 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors to deli\u00adver the per\u00adfor\u00admance requi\u00adred to build lar\u00adge gene\u00adra\u00adti\u00adve AI<\/span> models, lar\u00adge lan\u00adguage models<\/a>, recom\u00admen\u00adder sys\u00adtems<\/a> and more.<\/p>\n

Com\u00adbi\u00adned with NVIDIA<\/span> net\u00adwor\u00adking, this archi\u00adtec\u00adtu\u00adre super\u00adchar\u00adges effi\u00adci\u00adent com\u00adpu\u00adting at sca\u00adle by deli\u00adve\u00adring up to 9x more per\u00adfor\u00admance than the pre\u00advious gene\u00adra\u00adti\u00adon and 20x to 40x more per\u00adfor\u00admance than unac\u00adce\u00adle\u00adra\u00adted X86<\/span> dual-socket ser\u00advers for AI<\/span> trai\u00adning and HPC<\/span> workloads. If a lan\u00adguage model pre\u00advious\u00adly requi\u00adred 40 days to train on a clus\u00adter of X86-only ser\u00advers, the NVIDIA<\/span> DGX<\/span> H100<\/span> using Intel Xeon CPUs and ConnectX\u20117 powered net\u00adwor\u00adking could com\u00adple\u00adte the same work in as litt\u00adle as 1\u20132 days.<\/p>\n

NVIDIA<\/span> DGX<\/span> H100<\/span> sys\u00adtems are the buil\u00adding blocks of an enter\u00adpri\u00adse-rea\u00addy, turn\u00adkey NVIDIA<\/span> DGX<\/span> Super\u00adPOD<\/a>, which deli\u00advers up to one exa\u00adflop<\/a> of AI<\/span> per\u00adfor\u00admance, pro\u00advi\u00adding a leap in effi\u00adci\u00aden\u00adcy for lar\u00adge-sca\u00adle enter\u00adpri\u00adse AI<\/span> deployment.<\/p>\n

NVIDIA<\/span> Partners Boost Data Center Efficiency <\/b><\/h2>\n

For AI<\/span> data cen\u00adter workloads, NVIDIA<\/span> H100<\/span> GPUs enable enter\u00adpri\u00adses to build and deploy appli\u00adca\u00adti\u00adons more efficiently.<\/p>\n

Brin\u00adging a new gene\u00adra\u00adti\u00adon of per\u00adfor\u00admance and ener\u00adgy effi\u00adci\u00aden\u00adcy to enter\u00adpri\u00adses world\u00adwi\u00adde, a broad port\u00adfo\u00adlio of sys\u00adtems with H100<\/span> GPUs and 4th Gen Intel Xeon Sca\u00adlable CPUs are coming soon from NVIDIA<\/span> part\u00adners, inclu\u00adding ASUS<\/span>, Atos, Cis\u00adco, Dell Tech\u00adno\u00adlo\u00adgies, Fuji\u00adtsu, GIGABYTE<\/span>, Hew\u00adlett Packard Enter\u00adpri\u00adse, Leno\u00advo, QCT<\/span> and Supermicro.<\/p>\n

As the bell\u00adwe\u00adther of the effi\u00adci\u00aden\u00adcy gains to come, the Fla\u00adti\u00adron Institute\u2019s Leno\u00advo Think\u00adSys\u00adtem with NVIDIA<\/span> H100<\/span> GPUs<\/a> tops the latest Green500 list \u2014 and NVIDIA<\/span> tech\u00adno\u00adlo\u00adgies power 23 of the top 30 sys\u00adtems on the list. The Fla\u00adti\u00adron sys\u00adtem uses pri\u00ador-gene\u00adra\u00adti\u00adon Intel CPUs, so even more effi\u00adci\u00aden\u00adcy is expec\u00adted from the sys\u00adtems now coming to market.<\/p>\n

Addi\u00adtio\u00adnal\u00adly, con\u00adnec\u00adting ser\u00advers with NVIDIA<\/span> ConnectX\u20117 net\u00adwor\u00adking and Intel 4th Gen Xeon Sca\u00adlable pro\u00adces\u00adsors will increase effi\u00adci\u00aden\u00adcy and redu\u00adce infra\u00adstruc\u00adtu\u00adre and power consumption.<\/p>\n

NVIDIA<\/span> ConnectX\u20117 adap\u00adters sup\u00adport PCIe Gen 5 and 400 Gbps per con\u00adnec\u00adtion using Ether\u00adnet or Infi\u00adni\u00adBand, doubling net\u00adwor\u00adking through\u00adput bet\u00adween ser\u00advers and to sto\u00adrage. The adap\u00adters sup\u00adport advan\u00adced net\u00adwor\u00adking, sto\u00adrage and secu\u00adri\u00adty off\u00adloads. ConnectX\u20117 redu\u00adces the num\u00adber of cables and switch ports nee\u00added, saving 17% or more on elec\u00adtri\u00adci\u00adty nee\u00added for the net\u00adwor\u00adking of lar\u00adge GPU-acce\u00adle\u00adra\u00adted HPC<\/span> and AI<\/span> clus\u00adters and con\u00adtri\u00adbu\u00adting to the bet\u00adter ener\u00adgy effi\u00adci\u00aden\u00adcy of the\u00adse new servers.<\/p>\n

NVIDIA<\/span> AI<\/span> Enterprise Software Delivers Full-Stack AI<\/span> Solution<\/b><\/h2>\n

Run\u00adning on NVIDIA<\/span> H100<\/span>, NVIDIA<\/span> AI<\/span> Enter\u00adpri\u00adse acce\u00adle\u00adra\u00adtes the data sci\u00adence pipe\u00adline and stream\u00adli\u00adnes the deve\u00adlo\u00adp\u00adment and deploy\u00adment of pre\u00addic\u00adti\u00adve AI<\/span> models to auto\u00adma\u00adte essen\u00adti\u00adal pro\u00adces\u00adses and gain rapid insights from data.<\/p>\n

With an exten\u00adsi\u00adve libra\u00adry of full-stack soft\u00adware, inclu\u00adding AI<\/span> work\u00adflows of refe\u00adrence appli\u00adca\u00adti\u00adons, frame\u00adworks, pre\u00adtrai\u00adned models and infra\u00adstruc\u00adtu\u00adre opti\u00admiza\u00adti\u00adon, the soft\u00adware pro\u00advi\u00addes an ide\u00adal foun\u00adda\u00adti\u00adon for sca\u00adling enter\u00adpri\u00adse AI<\/span> success.<\/p>\n

To try out NVIDIA<\/span> H100<\/span> run\u00adning AI<\/span> work\u00adflows and frame\u00adworks sup\u00adport\u00aded in NVIDIA<\/span> AI<\/span> Enter\u00adpri\u00adse, sign up for NVIDIA<\/span> Launch\u00adPad<\/a> free of charge.<\/p>\n

Watch NVIDIA<\/span> foun\u00adder and CEO<\/span> Jen\u00adsen Huang <\/a>speak at the 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsor launch event.<\/p>\n","protected":false},"excerpt":{"rendered":"

Acce\u00adle\u00adra\u00adted NVIDIA<\/span> Hop\u00adper sys\u00adtems with 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors \u2014 inclu\u00adding NVIDIA<\/span> DGX<\/span> H100<\/span> and 60+ sys\u00adtems from NVIDIA<\/span> part\u00adners \u2014 pro\u00advi\u00adde 25x more effi\u00adci\u00aden\u00adcy than tra\u00addi\u00adtio\u00adnal data cen\u00adter ser\u00advers to save big on ener\u00adgy costs.
\n (\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":[2164,2818,1004,992],"class_list":["post-67514","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","tag-ai","tag-dgx-h100","tag-intel","tag-nvidia","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/67514","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=67514"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/67514\/revisions"}],"predecessor-version":[{"id":67515,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/67514\/revisions\/67515"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=67514"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=67514"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=67514"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}

The\u00adse next-gene\u00adra\u00adti\u00adon sys\u00adtems also deli\u00adver a leap for\u00adward in ope\u00adra\u00adtio\u00adnal effi\u00adci\u00aden\u00adcy as they\u2019re opti\u00admi\u00adzed for the NVIDIA<\/span> AI<\/span> Enter\u00adpri\u00adse soft\u00adware suite<\/a>.<\/p>\n

As the fourth gene\u00adra\u00adti\u00adon of the world\u2019s pre\u00admier pur\u00adpo\u00adse-built AI<\/span> infra\u00adstruc\u00adtu\u00adre, NVIDIA<\/span> DGX<\/span> H100<\/span> sys\u00adtems pro\u00advi\u00adde a ful\u00adly opti\u00admi\u00adzed plat\u00adform powered by the ope\u00adra\u00adting sys\u00adtem of the acce\u00adle\u00adra\u00adted data cen\u00adter, NVIDIA<\/span> Base Com\u00admand<\/a> software.<\/p>\n