{"id":67516,"date":"2023-01-10T20:08:28","date_gmt":"2023-01-10T19:08:28","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=67516"},"modified":"2023-01-10T20:08:28","modified_gmt":"2023-01-10T19:08:28","slug":"intel-launches-4th-gen-xeon-scalable-processors-max-series-cpus","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/67516-intel-launches-4th-gen-xeon-scalable-processors-max-series-cpus\/","title":{"rendered":"Intel Launches 4th Gen Xeon Scalable Processors, Max Series CPUs"},"content":{"rendered":"<p>Intel high\u00adlights broad indus\u00adtry adop\u00adti\u00adon across all major CSPs, OEMs, ODMs and ISVs, and show\u00adca\u00adses increased per\u00adfor\u00admance in <span class=\"caps\">AI<\/span>, net\u00adwor\u00adking and high per\u00adfor\u00admance computing.<\/p>\n<p><strong><span class=\"caps\">NEWS<\/span> <span class=\"caps\">HIGHLIGHTS<\/span><\/strong><br>\n&nbsp;<\/p>\n<ul>\n<li>Expan\u00adsi\u00adve&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/4th-Gen-Xeon-Customer-Fact-Sheet.pdf\">cus\u00adto\u00admer and part\u00adner adop\u00adti\u00adon<\/a>&nbsp;from <span class=\"caps\">AWS<\/span>, Cis\u00adco, Clou\u00adde\u00adra, Core\u00adWea\u00adve, Dell Tech\u00adno\u00adlo\u00adgies, Drop\u00adbox, Erics\u00adson, Fuji\u00adtsu, Goog\u00adle Cloud, Hew\u00adlett Packard Enter\u00adpri\u00adse, <span class=\"caps\">IBM<\/span> Cloud, Inspur Infor\u00adma\u00adti\u00adon, <span class=\"caps\">IONOS<\/span>, Leno\u00advo, Los Ala\u00admos Natio\u00adnal Labo\u00adra\u00adto\u00adry, Micro\u00adsoft Azu\u00adre, <span class=\"caps\">NVIDIA<\/span>, Ora\u00adcle Cloud, OVH\u00adcloud, phoe\u00adnix\u00adNAP, Red\u00adHat, <span class=\"caps\">SAP<\/span>, Super\u00adMi\u00adcro, Tele\u00adfo\u00adni\u00adca and VMware, among others.<\/li>\n<li>With the most built-in acce\u00adle\u00adra\u00adtors of any <span class=\"caps\">CPU<\/span> in the world for key workloads such as <span class=\"caps\">AI<\/span>, ana\u00adly\u00adtics, net\u00adwor\u00adking, secu\u00adri\u00adty, sto\u00adrage and high per\u00adfor\u00admance com\u00adpu\u00adting (<span class=\"caps\">HPC<\/span>), 4th Gen Intel Xeon Sca\u00adlable and Intel Max Series fami\u00adlies deli\u00adver lea\u00adder\u00adship per\u00adfor\u00admance in a pur\u00adpo\u00adse-built workload-first approach.<\/li>\n<li>4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors are Intel\u2019s most sus\u00adtainable data cen\u00adter pro\u00adces\u00adsors, deli\u00adve\u00adring a ran\u00adge of fea\u00adtures for opti\u00admi\u00adzing power and per\u00adfor\u00admance, making opti\u00admal use of <span class=\"caps\">CPU<\/span> resour\u00adces to help achie\u00adve cus\u00adto\u00admers\u2019 sus\u00adtaina\u00adbi\u00adli\u00adty&nbsp;goals.<\/li>\n<li>When com\u00adpared with pri\u00ador gene\u00adra\u00adti\u00adons, 4th Gen Xeon cus\u00adto\u00admers can expect a 2.9x<sup>1<\/sup>&nbsp;avera\u00adge per\u00adfor\u00admance per watt effi\u00adci\u00aden\u00adcy impro\u00adve\u00adment for tar\u00adge\u00adted workloads when uti\u00adli\u00adzing built-in acce\u00adle\u00adra\u00adtors, up to 70-watt<sup>2<\/sup>&nbsp;power savings per <span class=\"caps\">CPU<\/span> in opti\u00admi\u00adzed power mode with mini\u00admal per\u00adfor\u00admance loss for sel\u00adect workloads and a 52% to 66% lower total cost of owner\u00adship (<span class=\"caps\">TCO<\/span>)<sup>3<\/sup>.<\/li>\n<\/ul>\n<p><span class=\"caps\">SANTA<\/span> <span class=\"caps\">CLARA<\/span>, Calif., Jan. 10, 2023 \u2013 Intel today mark\u00aded one of the most important pro\u00adduct laun\u00adches in com\u00adpa\u00adny histo\u00adry with the unvei\u00adling of 4th Gen Intel\u00ae Xeon\u00ae Sca\u00adlable pro\u00adces\u00adsors (code-named Sap\u00adphi\u00adre Rapids), the Intel\u00ae Xeon\u00ae <span class=\"caps\">CPU<\/span> Max Series (code-named Sap\u00adphi\u00adre Rapids <span class=\"caps\">HBM<\/span>) and the Intel\u00ae Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series (code-named Pon\u00adte Vec\u00adchio), deli\u00adve\u00adring for its cus\u00adto\u00admers a leap in data cen\u00adter per\u00adfor\u00admance, effi\u00adci\u00aden\u00adcy, secu\u00adri\u00adty and new capa\u00adbi\u00adli\u00adties for <span class=\"caps\">AI<\/span>, the cloud, the net\u00adwork and edge, and the world\u2019s most powerful supercomputers.<\/p>\n<p>Working along\u00adside its cus\u00adto\u00admers and part\u00adners with 4th Gen Xeon, Intel is deli\u00adve\u00adring dif\u00adfe\u00adren\u00adtia\u00adted solu\u00adti\u00adons and sys\u00adtems at sca\u00adle to tack\u00adle their big\u00adgest com\u00adpu\u00adting chal\u00adlenges. Intel\u2019s uni\u00adque approach to pro\u00advi\u00adding pur\u00adpo\u00adse-built, workload-first acce\u00adle\u00adra\u00adti\u00adon and high\u00adly opti\u00admi\u00adzed soft\u00adware tun\u00aded for spe\u00adci\u00adfic workloads enables the com\u00adpa\u00adny to deli\u00adver the right per\u00adfor\u00admance at the right power for opti\u00admal over\u00adall total cost of ownership.<\/p>\n<p><strong>Press Kit:<\/strong>&nbsp;<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/newsroom\/resources\/press-kit-4th-gen-intel-xeon-scalable-processors.html\">4th Gen Xeon Sca\u00adlable Processors<\/a><\/p>\n<p>Addi\u00adtio\u00adnal\u00adly, as Intel\u2019s&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/4th-Gen-Xeon-Sustainability-Fact-Sheet.pdf\">most sus\u00adtainable data cen\u00adter pro\u00adces\u00adsors<\/a>, 4th Gen Xeon pro\u00adces\u00adsors deli\u00adver cus\u00adto\u00admers a ran\u00adge of fea\u00adtures for mana\u00adging power and per\u00adfor\u00admance, making the opti\u00admal use of <span class=\"caps\">CPU<\/span> resour\u00adces to help achie\u00adve their sus\u00adtaina\u00adbi\u00adli\u00adty&nbsp;goals.<\/p>\n<p><span class=\"dquo\">\u201c<\/span>The launch of 4th Gen Xeon Sca\u00adlable pro\u00adces\u00adsors and the Max Series pro\u00adduct fami\u00adly is a pivo\u00adtal moment in fue\u00adling Intel\u2019s tur\u00adn\u00adaround, reig\u00adni\u00adting our path to lea\u00adder\u00adship in the data cen\u00adter and gro\u00adwing our foot\u00adprint in new are\u00adnas,\u201d said San\u00addra Rive\u00adra, Intel exe\u00adcu\u00adti\u00adve vice pre\u00adsi\u00addent and gene\u00adral mana\u00adger of the Data Cen\u00adter and <span class=\"caps\">AI<\/span> Group. \u201cIntel\u2019s 4th Gen Xeon and the Max Series pro\u00adduct fami\u00adly deli\u00adver what cus\u00adto\u00admers tru\u00adly want \u2013 lea\u00adder\u00adship per\u00adfor\u00admance and relia\u00adbi\u00adli\u00adty within a secu\u00adre envi\u00adron\u00adment for their real-world requi\u00adre\u00adments \u2013 dri\u00adving fas\u00adter time to value and powe\u00adring their pace of innovation.\u201d<\/p>\n<p>Unli\u00adke any other data cen\u00adter pro\u00adces\u00adsor on the mar\u00adket and alre\u00ada\u00addy in the hands of cus\u00adto\u00admers today, the 4th Gen Xeon fami\u00adly great\u00adly expands on Intel\u2019s pur\u00adpo\u00adse-built, workload-first stra\u00adtegy and approach.&nbsp;<\/p>\n<p><strong>Lea\u00adding Per\u00adfor\u00admance and Sus\u00adtaina\u00adbi\u00adli\u00adty Bene\u00adfits with the Most Built-In Acceleration<\/strong><\/p>\n<p>Today, the\u00adre are over 100 mil\u00adli\u00adon Xeons instal\u00adled in the mar\u00adket \u2013 from on-prem ser\u00advers run\u00adning <span class=\"caps\">IT<\/span> ser\u00advices, inclu\u00adding new as-a-ser\u00advice busi\u00adness models, to net\u00adwor\u00adking equip\u00adment mana\u00adging Inter\u00adnet traf\u00adfic, to wire\u00adless base sta\u00adti\u00adon com\u00adpu\u00adting at the edge, to cloud services.<\/p>\n<p>Buil\u00adding on deca\u00addes of data cen\u00adter, net\u00adwork and intel\u00adli\u00adgent edge inno\u00adva\u00adti\u00adon and lea\u00adder\u00adship, new 4th Gen Xeon pro\u00adces\u00adsors deli\u00adver lea\u00adding per\u00adfor\u00admance with the&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/4th-Gen-Xeon-Accelerator-Fact-Sheet.pdf\">most built-in acce\u00adle\u00adra\u00adtors of any <span class=\"caps\">CPU<\/span><\/a>&nbsp;in the world to tack\u00adle cus\u00adto\u00admers\u2019 most important com\u00adpu\u00adting chal\u00adlenges across <span class=\"caps\">AI<\/span>, ana\u00adly\u00adtics, net\u00adwor\u00adking, secu\u00adri\u00adty, sto\u00adrage and&nbsp;<span class=\"caps\">HPC<\/span>.<\/p>\n<p>&nbsp;<\/p>\n<p><a href=\"https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-67517\" src=\"https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080.jpg\" alt width=\"1920\" height=\"1080\" srcset=\"https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080.jpg 1920w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080-300x169.jpg 300w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080-1024x576.jpg 1024w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080-768x432.jpg 768w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080-1536x864.jpg 1536w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-4th-gen-xeon-data.jpg.rendition.intel_.web_.1920.1080-1568x882.jpg 1568w\" sizes=\"auto, (max-width: 1920px) 100vw, 1920px\"><\/a><\/p>\n<p>When com\u00adpa\u00adring with pri\u00ador gene\u00adra\u00adti\u00adons, 4th Gen Intel Xeon cus\u00adto\u00admers can expect a 2.9x<sup>1<\/sup>&nbsp;avera\u00adge per\u00adfor\u00admance per watt effi\u00adci\u00aden\u00adcy impro\u00adve\u00adment for tar\u00adge\u00adted workloads when uti\u00adli\u00adzing built-in acce\u00adle\u00adra\u00adtors, up to 70-watt<sup>2<\/sup>&nbsp;power savings per <span class=\"caps\">CPU<\/span> in opti\u00admi\u00adzed power mode with mini\u00admal per\u00adfor\u00admance loss, and a 52% to 66% lower <span class=\"caps\">TCO<\/span><sup>3<\/sup>.<\/p>\n<p><strong>Sus\u00adtaina\u00adbi\u00adli\u00adty<\/strong><\/p>\n<p>The expan\u00adsi\u00adve\u00adness of built-in acce\u00adle\u00adra\u00adtors included in 4th Gen Xeon means Intel deli\u00advers plat\u00adform-level power savings, les\u00adsening the need for addi\u00adtio\u00adnal dis\u00adcrete acce\u00adle\u00adra\u00adti\u00adon and hel\u00adping our cus\u00adto\u00admers achie\u00adve their sus\u00adtaina\u00adbi\u00adli\u00adty goals. Addi\u00adtio\u00adnal\u00adly, the new Opti\u00admi\u00adzed Power Mode can deli\u00adver up to 20% socket power savings with a less than 5% per\u00adfor\u00admance impact for sel\u00adec\u00adted workloads<sup>11<\/sup>. New inno\u00adva\u00adtions in air and liquid coo\u00adling redu\u00adce total data cen\u00adter ener\u00adgy con\u00adsump\u00adti\u00adon fur\u00adther; and for the manu\u00adfac\u00adtu\u00adring of 4th Gen Xeon, it\u2019s been built with&nbsp; 90% or more rene\u00adwa\u00adble elec\u00adtri\u00adci\u00adty at Intel sites with sta\u00adte-of-the-art water recla\u00adma\u00adti\u00adon facilities.<\/p>\n<p><strong>Arti\u00adfi\u00adci\u00adal Intelligence<\/strong><\/p>\n<p>In <span class=\"caps\">AI<\/span>, and com\u00adpared to pre\u00advious gene\u00adra\u00adti\u00adon, 4th Gen Xeon pro\u00adces\u00adsors achie\u00adve up to 10x<sup>5,6<\/sup>&nbsp;hig\u00adher PyTorch real-time infe\u00adrence and trai\u00adning per\u00adfor\u00admance with built-in Intel\u00ae Advan\u00adced Matrix Exten\u00adsi\u00adon (Intel\u00ae <span class=\"caps\">AMX<\/span>) acce\u00adle\u00adra\u00adtors. Intel\u2019s 4th Gen Xeon unlocks new levels of per\u00adfor\u00admance for infe\u00adrence and trai\u00adning across a wide breadth of <span class=\"caps\">AI<\/span> workloads. The Xeon <span class=\"caps\">CPU<\/span> Max Series expands on the\u00adse capa\u00adbi\u00adli\u00adties for natu\u00adral lan\u00adguage pro\u00adces\u00adsing, with cus\u00adto\u00admers see\u00ading up to a 20x<sup>12<\/sup>&nbsp;speed-up on lar\u00adge lan\u00adguage models. With the deli\u00advery of Intel\u2019s <span class=\"caps\">AI<\/span> soft\u00adware suite, deve\u00adlo\u00adpers can use their <span class=\"caps\">AI<\/span> tool of choice, while incre\u00adasing pro\u00adduc\u00adti\u00advi\u00adty and spee\u00adding time to <span class=\"caps\">AI<\/span> deve\u00adlo\u00adp\u00adment. The suite is por\u00adta\u00adble from the work\u00adsta\u00adtion, enab\u00adling it to sca\u00adle out in the cloud and all the way out to the edge. And it has been vali\u00adda\u00adted with over 400 machi\u00adne lear\u00adning and deep lear\u00adning <span class=\"caps\">AI<\/span> models across the most com\u00admon <span class=\"caps\">AI<\/span> uses cases in every busi\u00adness segment.&nbsp;<\/p>\n<p><strong>Net\u00adwor\u00adking<\/strong><\/p>\n<p>4th Gen Xeon offers a fami\u00adly of pro\u00adces\u00adsors spe\u00adci\u00adfi\u00adcal\u00adly&nbsp;<a href=\"https:\/\/community.intel.com\/t5\/Blogs\/Tech-Innovation\/Edge-5G\/Breakthrough-Network-and-Edge-Performance-for-a-New-Era-of\/post\/1443547\">opti\u00admi\u00adzed for high-per\u00adfor\u00admance, low-laten\u00adcy net\u00adwork and edge workloads<\/a>. The\u00adse pro\u00adces\u00adsors are a cri\u00adti\u00adcal part of the foun\u00adda\u00adti\u00adon dri\u00adving a more soft\u00adware-defi\u00adned future for indus\u00adtries ran\u00adging from tele\u00adcom\u00admu\u00adni\u00adca\u00adti\u00adons and retail to manu\u00adfac\u00adtu\u00adring and smart cities. For <span class=\"caps\">5G<\/span> core workloads, built-in acce\u00adle\u00adra\u00adtors help increase through\u00adput and decrease laten\u00adcy, while advan\u00adces in power manage\u00adment enhan\u00adce both the respon\u00adsi\u00adve\u00adness and the effi\u00adci\u00aden\u00adcy of the plat\u00adform. And, when com\u00adpared to pre\u00advious gene\u00adra\u00adti\u00adons, 4th Gen Xeon deli\u00advers up to twice the vir\u00adtua\u00adli\u00adzed radio access net\u00adwork (vRAN) capa\u00adci\u00adty wit\u00adhout incre\u00adasing power con\u00adsump\u00adti\u00adon. This enables com\u00admu\u00adni\u00adca\u00adti\u00adons ser\u00advice pro\u00advi\u00adders to dou\u00adble the per\u00adfor\u00admance-per-watt to meet their cri\u00adti\u00adcal per\u00adfor\u00admance, sca\u00adling and ener\u00adgy effi\u00adci\u00aden\u00adcy&nbsp;needs.<\/p>\n<p><strong><a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/products\/docs\/processors\/xeon\/xeon-max-series-product-brief.html\">High Per\u00adfor\u00admance Computing<\/a><\/strong><\/p>\n<p>4th Gen Xeon and the&nbsp;<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/newsroom\/news\/introducing-intel-max-series-product-family.html\">Intel Max Series pro\u00adduct fami\u00adly<\/a>&nbsp;bring a sca\u00adlable, balan\u00adced archi\u00adtec\u00adtu\u00adre that inte\u00adgra\u00adtes <span class=\"caps\">CPU<\/span> and <span class=\"caps\">GPU<\/span> with&nbsp;<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/newsroom\/news\/oneapi-2023-tools-maximize-value-intel-hardware.html\">oneAPI\u2019s open soft\u00adware eco\u00adsys\u00adtem<\/a>&nbsp;for deman\u00adding com\u00adpu\u00adting workloads in <span class=\"caps\">HPC<\/span> and <span class=\"caps\">AI<\/span>, sol\u00adving the world\u2019s most chal\u00adlen\u00adging problems.<\/p>\n<p>The Xeon <span class=\"caps\">CPU<\/span> Max Series is the first and only x86-based pro\u00adces\u00adsor with high band\u00adwidth memo\u00adry, acce\u00adle\u00adra\u00adting many <span class=\"caps\">HPC<\/span> workloads wit\u00adhout the need for code chan\u00adges. The Intel Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series is Intel\u2019s hig\u00adhest-den\u00adsi\u00adty pro\u00adces\u00adsor and will be available in seve\u00adral form fac\u00adtors that address dif\u00adfe\u00adrent cus\u00adto\u00admer&nbsp;needs.<\/p>\n<p>The Xeon <span class=\"caps\">CPU<\/span> Max Series offers 64 giga\u00adbytes of high band\u00adwidth memo\u00adry (HBM2e) on the packa\u00adge, signi\u00adfi\u00adcant\u00adly incre\u00adasing data through\u00adput for <span class=\"caps\">HPC<\/span> and <span class=\"caps\">AI<\/span> workloads. Com\u00adpared with top-end 3rd Gen Intel\u00ae Xeon\u00ae Sca\u00adlable pro\u00adces\u00adsors, the Xeon <span class=\"caps\">CPU<\/span> Max Series pro\u00advi\u00addes up to 3.7 times<sup>10<\/sup>&nbsp;more per\u00adfor\u00admance on a ran\u00adge of real-world appli\u00adca\u00adti\u00adons like ener\u00adgy and earth sys\u00adtems modeling.<\/p>\n<p>Fur\u00adther, the Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series packs over 100 bil\u00adli\u00adon tran\u00adsis\u00adtors into a 47-tile packa\u00adge, brin\u00adging new levels of through\u00adput to chal\u00adlen\u00adging workloads like phy\u00adsics, finan\u00adcial ser\u00advices and life sci\u00aden\u00adces. When pai\u00adred with the Xeon <span class=\"caps\">CPU<\/span> Max Series, the com\u00adbi\u00adned plat\u00adform achie\u00adves up to 12.8 times<sup>13<\/sup>&nbsp;grea\u00adter per\u00adfor\u00admance than the pri\u00ador gene\u00adra\u00adti\u00adon when run\u00adning the <span class=\"caps\">LAMMPS<\/span> mole\u00adcu\u00adlar dyna\u00admics simulator.<\/p>\n<p><strong>Most Fea\u00adture-Rich and Secu\u00adre Xeon Plat\u00adform&nbsp;Yet<\/strong><\/p>\n<p>Signi\u00adfy\u00ading the big\u00adgest plat\u00adform trans\u00adfor\u00adma\u00adti\u00adon Intel has deli\u00adver\u00aded, not only is&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/4th-Gen-Xeon-Scalable-Product-Brief.pdf\">4th Gen Xeon a mar\u00advel of acce\u00adle\u00adra\u00adti\u00adon, but it is also an achie\u00adve\u00adment in manu\u00adfac\u00adtu\u00adring<\/a>, com\u00adbi\u00adning up to four Intel 7\u2011built tiles on a sin\u00adgle packa\u00adge, con\u00adnec\u00adted using Intel <span class=\"caps\">EMIB<\/span> (embedded mul\u00adti-die inter\u00adcon\u00adnect bridge) pack\u00ada\u00adging tech\u00adno\u00adlo\u00adgy and deli\u00adve\u00adring new fea\u00adtures inclu\u00adding increased memo\u00adry band\u00adwidth with <span class=\"caps\">DDR5<\/span>, increased I\/O band\u00adwidth with PCIe5.0 and Com\u00adpu\u00adte Express Link (<span class=\"caps\">CXL<\/span>) 1.1 interconnect.&nbsp;<\/p>\n<p>At the foun\u00adda\u00adti\u00adon of it all is secu\u00adri\u00adty. With 4th Gen Xeon, Intel is deli\u00adve\u00adring the&nbsp;<a href=\"https:\/\/www.intel.com\/content\/www\/us\/en\/newsroom\/news\/state-confidential-computing.html\">most com\u00adpre\u00adhen\u00adsi\u00adve con\u00adfi\u00adden\u00adti\u00adal com\u00adpu\u00adting port\u00adfo\u00adlio<\/a>&nbsp;of any data cen\u00adter sili\u00adcon pro\u00advi\u00adder in the indus\u00adtry, enhan\u00adcing data secu\u00adri\u00adty, regu\u00adla\u00adto\u00adry com\u00adpli\u00adance and data sove\u00adreig\u00adn\u00adty. Intel remains the only sili\u00adcon pro\u00advi\u00adder to offer appli\u00adca\u00adti\u00adon iso\u00adla\u00adti\u00adon for data cen\u00adter com\u00adpu\u00adting with Intel\u00ae Soft\u00adware Guard Exten\u00adsi\u00adons (Intel\u00ae <span class=\"caps\">SGX<\/span>), which pro\u00advi\u00addes today\u2019s smal\u00adlest attack sur\u00adface for con\u00adfi\u00adden\u00adti\u00adal com\u00adpu\u00adting in pri\u00adva\u00adte, public and cloud-to-edge envi\u00adron\u00adments. Addi\u00adtio\u00adnal\u00adly, Intel\u2019s new vir\u00adtu\u00adal-machi\u00adne (<span class=\"caps\">VM<\/span>) iso\u00adla\u00adti\u00adon tech\u00adno\u00adlo\u00adgy, Intel\u00ae Trust Domain Exten\u00adsi\u00adons (Intel\u00ae <span class=\"caps\">TDX<\/span>), is ide\u00adal for port\u00ading exis\u00adting appli\u00adca\u00adti\u00adons into a con\u00adfi\u00adden\u00adti\u00adal envi\u00adron\u00adment and will debut with Micro\u00adsoft Azu\u00adre, Ali\u00adbaba Cloud, Goog\u00adle Cloud and <span class=\"caps\">IBM<\/span>&nbsp;Cloud.<\/p>\n<p>Final\u00adly, the modu\u00adlar archi\u00adtec\u00adtu\u00adre of 4th Gen Xeon allows Intel to offer a wide ran\u00adge of pro\u00adces\u00adsors across near\u00adly 50 tar\u00adge\u00adted SKUs for cus\u00adto\u00admer use cases or appli\u00adca\u00adti\u00adons, from main\u00adstream gene\u00adral-pur\u00adpo\u00adse SKUs to pur\u00adpo\u00adse-built SKUs for cloud, data\u00adba\u00adse and ana\u00adly\u00adtics, net\u00adwor\u00adking, sto\u00adrage, and sin\u00adgle-socket edge use cases.&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/4th-Gen-Xeon-On-Demand-Fact-Sheet.pdf\">The 4th Gen Xeon pro\u00adces\u00adsor fami\u00adly is On Demand-capa\u00adble<\/a>&nbsp;and varies in core count, fre\u00adquen\u00adcy, mix of acce\u00adle\u00adra\u00adtors, power enve\u00adlo\u00adpe and memo\u00adry through\u00adput as is appro\u00adpria\u00adte for tar\u00adget use cases and form fac\u00adtors addres\u00adsing cus\u00adto\u00admers\u2019 real-world requirements.<\/p>\n<p><strong><span class=\"caps\">SKU<\/span> <span class=\"caps\">TABLE<\/span>:<\/strong>&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/Intel-4th-Gen-Xeon_product_SKUs.jpg\">SKUs for 4th Gen Xeon and Intel Xeon <span class=\"caps\">CPU<\/span> Max Series<\/a><\/p>\n<p>&nbsp;<\/p>\n<p><a href=\"https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter wp-image-67518 size-large\" src=\"https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080-1024x576.jpg\" alt width=\"640\" height=\"360\" srcset=\"https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080-1024x576.jpg 1024w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080-300x169.jpg 300w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080-768x432.jpg 768w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080-1536x864.jpg 1536w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080-1568x882.jpg 1568w, https:\/\/www.planet3dnow.de\/cms\/wp-content\/uploads\/2023\/01\/newsroom-intel-4th-gen-intel-xeon-1.jpg.rendition.intel_.web_.1920.1080.jpg 1920w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\"><\/a><\/p>\n<p><sub>\u00b9 Geo\u00adme\u00adan of fol\u00adlo\u00adwing workloads:&nbsp;&nbsp;Rock\u00adsDB&nbsp;(<span class=\"caps\">IAA<\/span> vs <span class=\"caps\">ZTD<\/span>),&nbsp;Click\u00adHouse&nbsp;(<span class=\"caps\">IAA<\/span> vs <span class=\"caps\">ZTD<\/span>), <span class=\"caps\">SPDK<\/span> lar\u00adge media and data\u00adba\u00adse request pro\u00adxies (<span class=\"caps\">DSA<\/span> vs out of box), Image Clas\u00adsi\u00adfi\u00adca\u00adti\u00adon Res\u00adNet-50 (<span class=\"caps\">AMX<\/span> vs <span class=\"caps\">VNNI<\/span>), Object Detec\u00adtion SSD-Res\u00adNet-34 (<span class=\"caps\">AMX<\/span> vs <span class=\"caps\">VNNI<\/span>),&nbsp;QAT\u00adzip&nbsp;(<span class=\"caps\">QAT<\/span> vs&nbsp;zlib)<\/sub><\/p>\n<p><sub>\u00b2 1\u2011node, Intel Refe\u00adrence Vali\u00adda\u00adti\u00adon Plat\u00adform, 2x Intel\u00ae Xeon 8480+ (<span class=\"caps\">56C<\/span>,&nbsp;2GHz, <span class=\"caps\">350W<\/span> <span class=\"caps\">TDP<\/span>), <span class=\"caps\">HT<\/span> On,&nbsp; Tur\u00adbo <span class=\"caps\">ON<\/span>,&nbsp; Total Memo\u00adry: 1&nbsp;<span class=\"caps\">TB<\/span> (16 slots\/ <span class=\"caps\">64GB<\/span>\/ 4800 MHz), 1x <span class=\"caps\">P4510<\/span> 3.<span class=\"caps\">84TB<\/span>&nbsp;NVMe&nbsp;PCIe Gen4 dri\u00adve,&nbsp;<span class=\"caps\">BIOS<\/span>: 0091.<span class=\"caps\">D05<\/span>, (ucode:0x2b0000c0),&nbsp;&nbsp;Cent\u00adOS Stream 8, 5.15.0\u2011spr.bkc.pc.10.4.11.x86_64,&nbsp;Java Perf\/Watt&nbsp;w\/&nbsp;openjdk-11+28_linux-x64_bin, 112 ins\u00adtances, <span class=\"caps\">1550MB<\/span> Initial\/Max heap size,&nbsp;Tes\u00adted by Intel as of Oct&nbsp;2022.<\/sub><\/p>\n<p><sub>\u00b3 ResNet50 Image Classification<\/sub><\/p>\n<p><sub>New Con\u00adfi\u00adgu\u00adra\u00adti\u00adon: 1\u2011node, 2x pre-pro\u00adduc\u00adtion 4th Gen Intel\u00ae Xeon\u00ae Sca\u00adlable <span class=\"caps\">8490H<\/span> pro\u00adces\u00adsor (60 core) with Intel\u00ae Advan\u00adced Matrix Exten\u00adsi\u00adons (Intel <span class=\"caps\">AMX<\/span>),&nbsp; on pre-pro\u00adduc\u00adtion&nbsp;Super\u00adMi\u00adcro&nbsp;<span class=\"caps\">SYS-221H-TNR<\/span> with <span class=\"caps\">1024GB<\/span> <span class=\"caps\">DDR5<\/span> memo\u00adry (16x64 <span class=\"caps\">GB<\/span>), micro\u00adcode 0x2b0000c0, <span class=\"caps\">HT<\/span> On, Tur\u00adbo&nbsp;On, <span class=\"caps\">SNC<\/span> Off, Cent\u00adOS Stream 8, 5.19.16\u2013301.fc37.x86_64, 1x3.<span class=\"caps\">84TB<\/span> <span class=\"caps\">P5510<\/span>&nbsp;NVMe, 10GbE x540-AT2, Intel <span class=\"caps\">TF<\/span> 2.10, <span class=\"caps\">AI<\/span> Model=Resnet 50 v1_5, best scores achie\u00adved: <span class=\"caps\">BS1<\/span> <span class=\"caps\">AMX<\/span> 1 core\/instance (max. 15ms <span class=\"caps\">SLA<\/span>),&nbsp; using phy\u00adsi\u00adcal cores, tes\u00adted by Intel Novem\u00adber&nbsp; 2022. Base\u00adline: 1\u2011node, 2x&nbsp;pro\u00adduc\u00adtion 3rd Gen Intel Xeon Sca\u00adlable 8380 Pro\u00adces\u00adsor ( 40 cores) on&nbsp;Super\u00adMi\u00adcro&nbsp;<span class=\"caps\">SYS-220U-TNR<\/span> , <span class=\"caps\">DDR4<\/span> memo\u00adry total <span class=\"caps\">1024GB<\/span> (16x64 <span class=\"caps\">GB<\/span>), micro\u00adcode 0xd000375, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, <span class=\"caps\">SNC<\/span> Off, Cent\u00adOS Stream 8, 5.19.16\u2013301.fc37.x86_64, 1x3.<span class=\"caps\">84TB<\/span> <span class=\"caps\">P5510<\/span>&nbsp;NVMe, 10GbE x540-AT2, Intel <span class=\"caps\">TF<\/span>&nbsp;2.10, <span class=\"caps\">AI<\/span> Model=Resnet 50 v1_5, best scores achie\u00adved: <span class=\"caps\">BS1<\/span> <span class=\"caps\">INT8<\/span> 2 cores\/instance (max. 15ms <span class=\"caps\">SLA<\/span>), using phy\u00adsi\u00adcal cores, tes\u00adted by Intel Novem\u00adber&nbsp;2022.<\/sub><\/p>\n<p><sub>For a&nbsp;50 ser\u00adver&nbsp;fleet of 3rd Gen Xeon 8380 (<span class=\"caps\">RN50<\/span> w\/DLBoost), esti\u00adma\u00adted as of Novem\u00adber&nbsp;2022:&nbsp;<\/sub><\/p>\n<p><sub>CapEx&nbsp;cos\u00adts: $1.<span class=\"caps\">64M\u200b<\/span><\/sub><\/p>\n<p><sub>OpEx&nbsp;cos\u00adts (4 year, includes power and coo\u00adling uti\u00adli\u00adty cos\u00adts, infra\u00adstruc\u00adtu\u00adre and hard\u00adware main\u00adten\u00adan\u00adce cos\u00adts): $739.<span class=\"caps\">9K\u200b<\/span><\/sub><\/p>\n<p><sub>Ener\u00adgy use in kWh (4 year, per ser\u00adver): 44627, <span class=\"caps\">PUE<\/span> 1.6\u200b<\/sub><\/p>\n<p><sub>Other assump\u00adti\u00adons:&nbsp; uti\u00adli\u00adty cost $0.1\/kWh,&nbsp; kWh to kg <span class=\"caps\">CO2<\/span> fac\u00adtor 0.42394&nbsp;<\/sub><\/p>\n<p><sub>For a&nbsp;17 ser\u00adver&nbsp;fleet of 4th Gen Xeon <span class=\"caps\">8490H<\/span> (<span class=\"caps\">RN50<\/span> w\/<span class=\"caps\">AMX<\/span>), esti\u00adma\u00adted as of Novem\u00adber&nbsp;2022:&nbsp;<\/sub><\/p>\n<p><sub>CapEx&nbsp;cos\u00adts: $799.<span class=\"caps\">4K\u200b<\/span><\/sub><\/p>\n<p><sub>OpEx&nbsp;cos\u00adts (4 year, includes power and coo\u00adling uti\u00adli\u00adty cos\u00adts, infra\u00adstruc\u00adtu\u00adre and hard\u00adware main\u00adten\u00adan\u00adce cos\u00adts): $275.<span class=\"caps\">3K\u200b<\/span><\/sub><\/p>\n<p><sub>Ener\u00adgy use in kWh (4 year, per ser\u00adver): 58581, <span class=\"caps\">PUE<\/span> 1.6<\/sub><\/p>\n<p><sub><span class=\"caps\">AI<\/span> \u2014 55% lower <span class=\"caps\">TCO<\/span> by deploy\u00ading fewer 4th Gen Intel\u00ae Xeon\u00ae pro\u00adces\u00adsor-based ser\u00advers to meet the same per\u00adfor\u00admance requi\u00adre\u00adment. See [<span class=\"caps\">E7<\/span>] at intel.com\/processorclaims: 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors. Results may&nbsp;vary.<\/sub><\/p>\n<p><sub>Data\u00adba\u00adse \u2014 52% lower <span class=\"caps\">TCO<\/span> by deploy\u00ading fewer 4th Gen Intel\u00ae Xeon\u00ae pro\u00adces\u00adsor-based ser\u00advers to meet the same per\u00adfor\u00admance requi\u00adre\u00adment. See [<span class=\"caps\">E8<\/span>] at intel.com\/processorclaims: 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors. Results may&nbsp;vary.<\/sub><\/p>\n<p><sub><span class=\"caps\">HPC<\/span> \u2014 66% lower <span class=\"caps\">TCO<\/span> by deploy\u00ading fewer Intel\u00ae Xeon\u00ae <span class=\"caps\">CPU<\/span> Max pro\u00adces\u00adsor-based ser\u00advers to meet the same per\u00adfor\u00admance requi\u00adre\u00adment. See [<span class=\"caps\">E9<\/span>] at intel.com\/processorclaims: 4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors. Results may&nbsp;vary.<\/sub><\/p>\n<p><sub><sup>4<\/sup>&nbsp;Geo\u00adme\u00adan of <span class=\"caps\">HP<\/span>&nbsp;Lin\u00adpack, Stream Tri\u00adad, SPECrate2017_fp_base&nbsp;est, SPECrate2017_int_base est. See [<span class=\"caps\">G2<\/span>, <span class=\"caps\">G4<\/span>, <span class=\"caps\">G6<\/span>] at intel.com\/processorclaims: 4th Gen Intel Xeon Scalable.<\/sub><\/p>\n<p><sub><sup>5<\/sup>&nbsp;Up to 10x hig\u00adher&nbsp;PyTorch&nbsp;real-time infe\u00adrence per\u00adfor\u00admance with built-in Intel\u00ae Advan\u00adced Matrix Exten\u00adsi\u00adons (Intel\u00ae <span class=\"caps\">AMX<\/span>) (<span class=\"caps\">BF16<\/span>) vs. the pri\u00ador gene\u00adra\u00adti\u00adon (<span class=\"caps\">FP32<\/span>)\u200b<\/sub><\/p>\n<p><sub>PyTorch&nbsp;geo\u00adme\u00adan of ResNet50, Bert-Lar\u00adge,&nbsp;Mas\u00adkRCNN, SSD-Res\u00adNe\u00adt34, <span class=\"caps\">RNN<\/span>\u2011T, Resnext101.<\/sub><\/p>\n<p><sub><sup>6<\/sup>&nbsp;Up to 10x hig\u00adher&nbsp;PyTorch&nbsp;trai\u00adning per\u00adfor\u00admance with built-in Intel\u00ae Advan\u00adced Matrix Exten\u00adsi\u00adons (Intel\u00ae <span class=\"caps\">AMX<\/span>) (<span class=\"caps\">BF16<\/span>) vs. the pri\u00ador gene\u00adra\u00adti\u00adon (<span class=\"caps\">FP32<\/span>)\u200b<\/sub><\/p>\n<p><sub>PyTorch&nbsp;geo\u00adme\u00adan of ResNet50, Bert-Lar\u00adge, <span class=\"caps\">DLRM<\/span>,&nbsp;Mas\u00adkRCNN, SSD-Res\u00adNe\u00adt34, <span class=\"caps\">RNN<\/span>\u2011T.\u200b<\/sub><\/p>\n<p><sub><sup>7<\/sup>&nbsp;Esti\u00adma\u00adted as of 8\/30\/2022 based on 4th gene\u00adra\u00adti\u00adon Intel\u00ae Xeon\u00ae Sca\u00adlable pro\u00adces\u00adsor archi\u00adtec\u00adtu\u00adre impro\u00adve\u00adments vs 3rd gene\u00adra\u00adti\u00adon Intel\u00ae Xeon\u00ae Sca\u00adlable pro\u00adces\u00adsor at simi\u00adlar core count, socket power and fre\u00adquen\u00adcy on a test sce\u00adna\u00adrio&nbsp;using&nbsp;Flex\u00adRAN\u2122 soft\u00adware. Results may&nbsp;vary.<\/sub><\/p>\n<p><sub><sup>8<\/sup>&nbsp;Up to 95% fewer cores and 2x hig\u00adher level 1 com\u00adpres\u00adsi\u00adon through\u00adput with 4th Gen Intel Xeon Pla\u00adti\u00adnum <span class=\"caps\">8490H<\/span> using inte\u00adgra\u00adted Intel <span class=\"caps\">QAT<\/span> vs. pri\u00ador generation.<\/sub><\/p>\n<p><sub><span class=\"caps\">8490H<\/span>: 1\u2011node, pre-pro\u00adduc\u00adtion plat\u00adform with 2x 4th&nbsp; Gen Intel\u00ae Xeon Sca\u00adlable Pro\u00adces\u00adsor (60 core) with inte\u00adgra\u00adted Intel&nbsp;Quick\u00adAs\u00adsist&nbsp;Acce\u00adle\u00adra\u00adtor (Intel <span class=\"caps\">QAT<\/span>), <span class=\"caps\">QAT<\/span> device utilized=8(2 sockets acti\u00adve), with Total <span class=\"caps\">1024GB<\/span> (16x64 <span class=\"caps\">GB<\/span>) <span class=\"caps\">DDR5<\/span>&nbsp;memo\u00adry, micro\u00adcode 0xf000380, <span class=\"caps\">HT<\/span> On, Tur\u00adbo Off, <span class=\"caps\">SNC<\/span> Off, Ubun\u00adtu 22.04.1 <span class=\"caps\">LTS<\/span>, 5.15.0\u201347-generic, 1x 1.<span class=\"caps\">92TB<\/span> Intel\u00ae <span class=\"caps\">SSDSC2KG01<\/span>, <span class=\"caps\">QAT<\/span> v20.l.0.9.1 ,&nbsp;QAT\u00adzip&nbsp;v1.0.9 , <span class=\"caps\">ISA<\/span>\u2011L v2.3.0, tes\u00adted by Intel Sep\u00adtem\u00adber&nbsp;2022.&nbsp;<\/sub><\/p>\n<p><sub>8380: 1\u2011node, 2x 3rd Gen Intel Xeon Sca\u00adlable Pro\u00adces\u00adsors( 40 cores) on Coyo\u00adte Pass plat\u00adform, <span class=\"caps\">DDR4<\/span> memo\u00adry total <span class=\"caps\">1024GB<\/span> (16x64 <span class=\"caps\">GB<\/span>), micro\u00adcode 0xd000375, <span class=\"caps\">HT<\/span> On, Tur\u00adbo Off, <span class=\"caps\">SNC<\/span> Off, Ubun\u00adtu 22.04.1 <span class=\"caps\">LTS<\/span>, 5.15.0\u201347-generic, 1x 1.<span class=\"caps\">92TB<\/span>&nbsp;Intel <span class=\"caps\">SSDSC2KG01<\/span>,<span class=\"caps\">QAT<\/span> v1.7.l.4.16,&nbsp;&nbsp;QAT\u00adzip&nbsp;v1.0.9 , <span class=\"caps\">ISA<\/span>\u2011L v2.3.0, tes\u00adted by Intel Octo\u00adber&nbsp;2022.<\/sub><\/p>\n<p><sub><sup>9<\/sup>&nbsp;Up to 3x hig\u00adher&nbsp;Rock\u00adsDB&nbsp;per\u00adfor\u00admance with 4th Gen Intel Xeon Pla\u00adti\u00adnum <span class=\"caps\">8490H<\/span> using inte\u00adgra\u00adted Intel <span class=\"caps\">IAA<\/span> vs. pri\u00ador generation.<\/sub><\/p>\n<p><sub><span class=\"caps\">8490H<\/span>: 1\u2011node, pre-pro\u00adduc\u00adtion Intel plat\u00adform with 2x 4th Gen Intel Xeon Sca\u00adlable Pro\u00adces\u00adsor (60 cores) with inte\u00adgra\u00adted Intel In-Memo\u00adry Ana\u00adly\u00adtics Acce\u00adle\u00adra\u00adtor (Intel <span class=\"caps\">IAA<\/span>), <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, Total Memo\u00adry <span class=\"caps\">1024GB<\/span> (16x64GB <span class=\"caps\">DDR5<\/span>&nbsp;4800),&nbsp; micro\u00adcode 0xf000380,&nbsp; 1x 1.<span class=\"caps\">92TB<\/span> <span class=\"caps\">INTEL<\/span> <span class=\"caps\">SSDSC2KG01<\/span>, Ubun\u00adtu 22.04.1 <span class=\"caps\">LTS<\/span>, 5.18.12\u2013051812-generic, <span class=\"caps\">QPL<\/span> v0.1.21,accel-config-v3.4.6.4, <span class=\"caps\">ZSTD<\/span> v1.5.2,&nbsp;Rock\u00adsDB&nbsp;v6.4.6 (db_bench), tes\u00adted by Intel&nbsp; Sep\u00adtem\u00adber&nbsp;2022.<\/sub><\/p>\n<p><sub>8380: 1\u2011node, 2x&nbsp; 3rd Gen Intel Xeon Sca\u00adlable Pro\u00adces\u00adsors( 40 cores) on Coyo\u00adte Pass plat\u00adform, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, <span class=\"caps\">SNC<\/span> Off, Total Memo\u00adry <span class=\"caps\">1024GB<\/span> (16x64GB <span class=\"caps\">DDR4<\/span> 3200), micro\u00adcode 0xd000375, 1x 1.<span class=\"caps\">92TB<\/span> <span class=\"caps\">INTEL<\/span> <span class=\"caps\">SSDSC2KG01<\/span>, Ubun\u00adtu&nbsp;22.04.1 <span class=\"caps\">LTS<\/span>, 5.18.12\u2013051812-generic,&nbsp; <span class=\"caps\">ZSTD<\/span> v1.5.2,&nbsp;Rock\u00adsDB&nbsp;v6.4.6 (db_bench), tes\u00adted by Intel&nbsp; Octo\u00adber&nbsp;2022.<\/sub><\/p>\n<p><sub><sup>10<\/sup>&nbsp;Intel\u00ae Xeon\u00ae 8380: Test by Intel as of 10\/7\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae 8380 <span class=\"caps\">CPU<\/span>, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, Total Memo\u00adry 256 <span class=\"caps\">GB<\/span> (16x16GB <span class=\"caps\">3200MT<\/span>\/s <span class=\"caps\">DDR4<\/span>), <span class=\"caps\">BIOS<\/span> Ver\u00adsi\u00adon <span class=\"caps\">SE5C620<\/span>.<span class=\"caps\">86B<\/span>.01.01.0006.2207150335,&nbsp;ucode&nbsp;revision=0xd000375, Rocky Linux 8.6, Linux ver\u00adsi\u00adon 4.18.0\u2013372.26.1.el8_\u200b6.crt1.x86_\u200b64, <span class=\"caps\">YASK<\/span> v3.05.07\u200b<\/sub><\/p>\n<p><sub>Intel\u00ae Xeon\u00ae <span class=\"caps\">CPU<\/span> Max Series: Test by Intel as of ww36\u201922. 1\u2011node, 2x Intel\u00ae Xeon\u00ae <span class=\"caps\">CPU<\/span> Max&nbsp;SeriesHT&nbsp;On, Tur\u00adbo On, <span class=\"caps\">SNC4<\/span>, Total Memo\u00adry 128 <span class=\"caps\">GB<\/span> (8x16GB <span class=\"caps\">HBM2<\/span> <span class=\"caps\">3200MT<\/span>\/s), <span class=\"caps\">BIOS<\/span> Ver\u00adsi\u00adon <span class=\"caps\">SE5C7411<\/span>.<span class=\"caps\">86B<\/span>.8424.<span class=\"caps\">D03<\/span>.2208100444,&nbsp;ucode&nbsp;revision=0x2c000020, Cent\u00adOS Stream 8, Linux ver\u00adsi\u00adon 5.19.0\u2011rc6.0712.intel_\u200bnext.1.x86_\u200b64+server, <span class=\"caps\">YASK<\/span> v3.05.07.<\/sub><\/p>\n<p><sub><sup>11<\/sup>&nbsp;Up to 20% sys\u00adtem power savings uti\u00adli\u00adzing 4th Gen Xeon Sca\u00adlable with Opti\u00admi\u00adzed Power mode on vs&nbsp; off on sel\u00adect workloads inclu\u00adding&nbsp;Spec\u00adJBB, <span class=\"caps\">SPECINT<\/span> and <span class=\"caps\">NIGNX<\/span> key handshake.<\/sub><\/p>\n<p><sub><sup>12<\/sup>&nbsp;<span class=\"caps\">AMD<\/span> Milan: Tes\u00adted by Numen\u00adta as of 11\/28\/2022. 1\u2011node, 2x <span class=\"caps\">AMD<\/span> <span class=\"caps\">EPYC<\/span> <span class=\"caps\">7R13<\/span> on <span class=\"caps\">AWS<\/span> m6a.48xlarge, 768 <span class=\"caps\">GB<\/span> <span class=\"caps\">DDR4-3200<\/span>, Ubun\u00adtu 20.04 Ker\u00adnel 5.15, Open\u00adVI\u00adNO 2022.3, BERT-Lar\u00adge, Sequence Length 512, Batch Size&nbsp;1<\/sub><\/p>\n<p><sub>Intel\u00ae Xeon\u00ae 8480+: Tes\u00adted by Numen\u00adta as of 11\/28\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae 8480+, 512 <span class=\"caps\">GB<\/span> <span class=\"caps\">DDR5-4800<\/span>, Ubun\u00adtu 22.04 Ker\u00adnel 5.17, Open\u00adVI\u00adNO 2022.3, Numen\u00adta-Opti\u00admi\u00adzed BERT-Lar\u00adge, Sequence Length 512, Batch Size&nbsp;1<\/sub><\/p>\n<p><sub>Intel\u00ae Xeon\u00ae Max 9468: Tes\u00adted by Numen\u00adta as of 11\/30\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae Max 9468, 128 <span class=\"caps\">GB<\/span> HBM2e 3200 <span class=\"caps\">MT<\/span>\/s, Ubun\u00adtu 22.04 Ker\u00adnel 5.15, Open\u00adVI\u00adNO 2022.3, Numen\u00adta-Opti\u00admi\u00adzed BERT-Lar\u00adge, Sequence Length 512, Batch Size&nbsp;1<\/sub><\/p>\n<p><sub><sup>13<\/sup>&nbsp;Intel\u00ae Xeon\u00ae 8380: Test by Intel as of 10\/28\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae 8380 <span class=\"caps\">CPU<\/span>, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, Total Memo\u00adry 256 <span class=\"caps\">GB<\/span> (16x16GB <span class=\"caps\">3200MT<\/span>\/s, Dual-Rank), <span class=\"caps\">BIOS<\/span> Ver\u00adsi\u00adon <span class=\"caps\">SE5C6200<\/span>.<span class=\"caps\">86B<\/span>.0020.<span class=\"caps\">P23<\/span>.2103261309, ucode revision=0xd000270, Rocky Linux 8.6, Linux ver\u00adsi\u00adon 4.18.0\u2013372.19.1.el8_6.crt1.x86_64\u200b<\/sub><\/p>\n<p><sub>Intel\u00ae Xeon\u00ae <span class=\"caps\">CPU<\/span> Max Series <span class=\"caps\">HBM<\/span>: Test by Intel as of 10\/28\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae Max 9480, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, Total Memo\u00adry 128 <span class=\"caps\">GB<\/span> HBM2e, <span class=\"caps\">BIOS<\/span> <span class=\"caps\">EGSDCRB1<\/span>.<span class=\"caps\">DWR<\/span>.0085.<span class=\"caps\">D12<\/span>.2207281916, ucode 0xac000040, <span class=\"caps\">SUSE<\/span> Linux Enter\u00adpri\u00adse Ser\u00adver 15 <span class=\"caps\">SP3<\/span>, Ker\u00adnel 5.3.18, oneA\u00adPI 2022.3.0\u200b<\/sub><\/p>\n<p><sub>Intel\u00ae Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series with <span class=\"caps\">DDR<\/span> Host: Test by Intel as of 10\/28\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae Max 9480, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, Total Memo\u00adry 1024 <span class=\"caps\">GB<\/span> <span class=\"caps\">DDR5-4800<\/span> + 128 <span class=\"caps\">GB<\/span> HBM2e, Memo\u00adry Mode: Flat, HBM2e not used, 6x Intel\u00ae Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series, <span class=\"caps\">BIOS<\/span> <span class=\"caps\">EGSDCRB1<\/span>.<span class=\"caps\">DWR<\/span>.0085.<span class=\"caps\">D12<\/span>.2207281916, ucode 0xac000040, Aga\u00adma pvc-prq-54, <span class=\"caps\">SUSE<\/span> Linux Enter\u00adpri\u00adse Ser\u00adver 15 <span class=\"caps\">SP3<\/span>, Ker\u00adnel 5.3.18, oneA\u00adPI 2022.3.0\u200b<\/sub><\/p>\n<p><sub>Intel\u00ae Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series with <span class=\"caps\">HBM<\/span> Host: Test by Intel as of 10\/28\/2022. 1\u2011node, 2x Intel\u00ae Xeon\u00ae Max 9480, <span class=\"caps\">HT<\/span> On, Tur\u00adbo On, Total Memo\u00adry 128 <span class=\"caps\">GB<\/span> HBM2e, 6x Intel\u00ae Data Cen\u00adter <span class=\"caps\">GPU<\/span> Max Series, <span class=\"caps\">BIOS<\/span> <span class=\"caps\">EGSDCRB1<\/span>.<span class=\"caps\">DWR<\/span>.0085.<span class=\"caps\">D12<\/span>.2207281916, ucode 0xac000040, Aga\u00adma pvc-prq-54, <span class=\"caps\">SUSE<\/span> Linux Enter\u00adpri\u00adse Ser\u00adver 15 <span class=\"caps\">SP3<\/span>, Ker\u00adnel 5.3.18, oneA\u00adPI 2022.3.0\u200b<\/sub><\/p>\n<p>&nbsp;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Intel high\u00adlights broad indus\u00adtry adop\u00adti\u00adon across all major CSPs, OEMs, ODMs and ISVs, and show\u00adca\u00adses increased per\u00adfor\u00admance in <span class=\"caps\">AI<\/span>, net\u00adwor\u00adking and high per\u00adfor\u00admance computing.<\/p>\n<p><strong><span class=\"caps\">NEWS<\/span> <span class=\"caps\">HIGHLIGHTS<\/span><\/strong><br>\n&nbsp;<\/p>\n<ul>\n<li>Expan\u00adsi\u00adve&nbsp;<a href=\"https:\/\/download.intel.com\/newsroom\/2023\/data-center-hpc\/4th-Gen-Xeon-Customer-Fact-Sheet.pdf\">cus\u00adto\u00admer and part\u00adner adop\u00adti\u00adon<\/a>&nbsp;from <span class=\"caps\">AWS<\/span>, Cis\u00adco, Clou\u00adde\u00adra, Core\u00adWea\u00adve, Dell Tech\u00adno\u00adlo\u00adgies, Drop\u00adbox, Erics\u00adson, Fuji\u00adtsu, Goog\u00adle Cloud, Hew\u00adlett Packard Enter\u00adpri\u00adse, <span class=\"caps\">IBM<\/span> Cloud, Inspur Infor\u00adma\u00adti\u00adon, <span class=\"caps\">IONOS<\/span>, Leno\u00advo, Los Ala\u00admos Natio\u00adnal Labo\u00adra\u00adto\u00adry, Micro\u00adsoft Azu\u00adre, <span class=\"caps\">NVIDIA<\/span>, Ora\u00adcle Cloud, OVH\u00adcloud, phoe\u00adnix\u00adNAP, Red\u00adHat, <span class=\"caps\">SAP<\/span>, Super\u00adMi\u00adcro, Tele\u00adfo\u00adni\u00adca and VMware, among others.<\/li>\n<li>With the most built-in acce\u00adle\u00adra\u00adtors of any <span class=\"caps\">CPU<\/span> in the world for key workloads such as <span class=\"caps\">AI<\/span>, ana\u00adly\u00adtics, net\u00adwor\u00adking, secu\u00adri\u00adty, sto\u00adrage and high per\u00adfor\u00admance com\u00adpu\u00adting (<span class=\"caps\">HPC<\/span>), 4th Gen Intel Xeon Sca\u00adlable and Intel Max Series fami\u00adlies deli\u00adver lea\u00adder\u00adship per\u00adfor\u00admance in a pur\u00adpo\u00adse-built workload-first approach.<\/li>\n<li>4th Gen Intel Xeon Sca\u00adlable pro\u00adces\u00adsors are Intel\u2019s most sus\u00adtainable data cen\u00adter pro\u00adces\u00adsors, deli\u00adve\u00adring a ran\u00adge of fea\u00adtures for opti\u00admi\u00adzing power and per\u00adfor\u00admance, making opti\u00admal use of <span class=\"caps\">CPU<\/span> resour\u00adces to help achie\u00adve cus\u00adto\u00admers\u2019 sus\u00adtaina\u00adbi\u00adli\u00adty&nbsp;goals.<\/li>\n<li>When com\u00adpared with pri\u00ador gene\u00adra\u00adti\u00adons, 4th Gen Xeon cus\u00adto\u00admers can expect a 2.9x<sup>1<\/sup>&nbsp;avera\u00adge per\u00adfor\u00admance per watt effi\u00adci\u00aden\u00adcy impro\u00adve\u00adment for tar\u00adge\u00adted workloads when uti\u00adli\u00adzing built-in acce\u00adle\u00adra\u00adtors, up to 70-watt<sup>2<\/sup>&nbsp;power savings per <span class=\"caps\">CPU<\/span> in opti\u00admi\u00adzed power mode with mini\u00admal per\u00adfor\u00admance loss for sel\u00adect workloads and a 52% to 66% lower total cost of owner\u00adship (<span class=\"caps\">TCO<\/span>)<sup>3<\/sup>.<\/li>\n<\/ul>\n<p> (\u2026) <a class=\"moretag\" href=\"https:\/\/www.planet3dnow.de\/cms\/67516-intel-launches-4th-gen-xeon-scalable-processors-max-series-cpus\/\">Wei\u00adter\u00adle\u00adsen&nbsp;\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":[2821,1004,2820,2819],"class_list":["post-67516","post","type-post","status-publish","format-standard","hentry","category-pressemitteilungen","tag-4th-gen-intel-xeon-scalable","tag-intel","tag-ponte-vecchio","tag-sapphire-rapids","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/67516","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=67516"}],"version-history":[{"count":1,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/67516\/revisions"}],"predecessor-version":[{"id":67519,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/67516\/revisions\/67519"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=67516"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=67516"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=67516"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}