{"id":43187,"date":"2019-01-11T08:41:19","date_gmt":"2019-01-11T07:41:19","guid":{"rendered":"https:\/\/www.planet3dnow.de\/cms\/?p=43187"},"modified":"2019-01-13T00:34:16","modified_gmt":"2019-01-12T23:34:16","slug":"amd-veroeffentlicht-26-benchmarks-zur-radeon-vii-29-prozent-mehrleistung","status":"publish","type":"post","link":"https:\/\/www.planet3dnow.de\/cms\/43187-amd-veroeffentlicht-26-benchmarks-zur-radeon-vii-29-prozent-mehrleistung\/","title":{"rendered":"AMD<\/span> ver\u00f6ffentlicht 26 Benchmarks zur Radeon VII<\/span> \u2013 29 % Mehrleistung"},"content":{"rendered":"

Bereits vor\u00adges\u00adtern haben wir \u00fcber AMDs Vor\u00adstel\u00adlung der neu\u00aden High-End-Gra\u00adfik\u00adkar\u00adte AMD<\/span> Rade\u00adon VII<\/span> berich\u00adtet<\/a>. Es han\u00addelt sich dabei um einen \u201cSal\u00adva\u00adge-Able\u00adger\u201d der Rade\u00adon Instinct MI60<\/span><\/a> aus dem HPC-Markt, also einen Vega-20-Chip mit teil\u00adwei\u00adse deak\u00adti\u00advier\u00adten Ein\u00adhei\u00adten (60 statt akti\u00adve 64 CUs).<\/p>\n

W\u00e4h\u00adrend wir sowohl \u00fcber die tech\u00adni\u00adschen Daten als auch die zuge\u00adh\u00f6\u00adri\u00adge Pres\u00adse\u00admit\u00adtei\u00adlung bereits berich\u00adtet<\/a> haben, war uns gar nicht auf\u00adge\u00adfal\u00adlen, dass AMD<\/span> in den Fu\u00df\u00adno\u00adten der Pres\u00adse\u00ader\u00adkl\u00e4\u00adrung<\/a> neben dem Mit\u00adtel wei\u00adte\u00adre 25 kon\u00adkre\u00adte Bench\u00admark-Wer\u00adte ver\u00adsteckt hat\u00adte (alle Wer\u00adte @4K<\/span> mit max. Settings):<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Titel<\/th>\nRade\u00adon VII<\/span><\/th>\nRade\u00adon RX<\/span> Vega 64<\/th>\nMehr\u00adleis\u00adtung<\/th>\n<\/tr>\n<\/thead>\n
Assassin\u2019s Creed Odyssey<\/td>\n36,0 fps<\/td>\n28,0 fps<\/td>\n28,6 %<\/td>\n<\/tr>\n
Batt\u00adle\u00adfield 1 (DX12<\/span>)<\/td>\n80,5 fps<\/td>\n59,2 fps<\/td>\n36,0 %<\/td>\n<\/tr>\n
Batt\u00adle\u00adfield 5 (DX12<\/span>)<\/td>\n62,2 fps<\/td>\n46,7 fps<\/td>\n33,2 %<\/td>\n<\/tr>\n
Call of Duty: Black Ops 4<\/td>\n82,3 fps<\/td>\n68,0 fps<\/td>\n21,0 %<\/td>\n<\/tr>\n
Desti\u00adny 2<\/td>\n65,1 fps<\/td>\n50,9 fps<\/td>\n27,9 %<\/td>\n<\/tr>\n
Deus Ex: Man\u00adkind Divi\u00added (DX12<\/span>)<\/td>\n53,4 fps<\/td>\n40,2 fps<\/td>\n32,8 %<\/td>\n<\/tr>\n
Doom (2016)<\/td>\n89,5 fps<\/td>\n67,2 fps<\/td>\n33,2 %<\/td>\n<\/tr>\n
F1<\/span> 2018 (DX12<\/span>)<\/td>\n78,0 fps<\/td>\n61,0 fps<\/td>\n27,9 %<\/td>\n<\/tr>\n
Fall\u00adout 76<\/td>\n76,6 fps<\/td>\n45,5 fps<\/td>\n68,4 %<\/td>\n<\/tr>\n
Far Cry 5<\/td>\n62,0 fps<\/td>\n49,0 fps<\/td>\n26,5 %<\/td>\n<\/tr>\n
For\u00adza Hori\u00adzon 4 (DX12<\/span>)<\/td>\n72,8 fps<\/td>\n62,8 fps<\/td>\n15,9 %<\/td>\n<\/tr>\n
Grand Theft Auto V<\/td>\n76,2 fps<\/td>\n60,1 fps<\/td>\n26,8 %<\/td>\n<\/tr>\n
Hit\u00adman 2<\/td>\n53,3 fps<\/td>\n49,6 fps<\/td>\n7,5 %<\/td>\n<\/tr>\n
Just Cau\u00adse 4<\/td>\n50,8 fps<\/td>\n42,6 fps<\/td>\n19,2 %<\/td>\n<\/tr>\n
Midd\u00adle-Earth: Shadow Of War<\/td>\n54,3 fps<\/td>\n41,6 fps<\/td>\n30,5 %<\/td>\n<\/tr>\n
Mons\u00adter Hun\u00adter World<\/td>\n35,4 fps<\/td>\n29,4 fps<\/td>\n20,4 %<\/td>\n<\/tr>\n
Rise of the Tomb Rai\u00adder (DX12<\/span>)<\/td>\n58,3 fps<\/td>\n46,0 fps<\/td>\n26,7 %<\/td>\n<\/tr>\n
Shadow of the Tomb Rai\u00adder (DX12<\/span>)<\/td>\n47,5 fps<\/td>\n36,3 fps<\/td>\n30,9 %<\/td>\n<\/tr>\n
Sid Meier\u2019s Civi\u00adliza\u00adti\u00adon VI<\/span> (DX12<\/span>)<\/td>\n97,1 fps<\/td>\n78,1 fps<\/td>\n24,3 %<\/td>\n<\/tr>\n
Star Con\u00adtrol: Origins<\/td>\n88,9 fps<\/td>\n69,2 fps<\/td>\n28,5 %<\/td>\n<\/tr>\n
Stran\u00adge Bri\u00adga\u00adde Vulkan<\/td>\n86,7 fps<\/td>\n60,9 fps<\/td>\n42,4 %<\/td>\n<\/tr>\n
The Wit\u00adcher 3<\/td>\n55,4 fps<\/td>\n41,4 fps<\/td>\n33,8 %<\/td>\n<\/tr>\n
Ghost Recon Wildlands<\/td>\n36,3 fps<\/td>\n29,2 fps<\/td>\n24,3 %<\/td>\n<\/tr>\n
Total War: War\u00adham\u00admer 2<\/td>\n34,6 fps<\/td>\n28,3 fps<\/td>\n22,3 %<\/td>\n<\/tr>\n
Wol\u00adfen\u00adstein 2<\/td>\n93,4 fps<\/td>\n74,2 fps<\/td>\n25,9 %<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Inter\u00ades\u00adsan\u00adter\u00adwei\u00adse wur\u00adden die Tests von AMD<\/span> nicht mit einem Ryzen-Sys\u00adtem ange\u00adfer\u00adtigt, son\u00addern mit einem Intel Core i7-7700K. Selbst\u00adver\u00adst\u00e4nd\u00adlich muss man wie \u00fcblich vor\u00adsich\u00adtig sein mit Her\u00adstel\u00adler-Bench\u00admarks, schlie\u00df\u00adlich k\u00f6n\u00adnen Klei\u00adnig\u00adkei\u00adten bei der Kon\u00adfi\u00adgu\u00adra\u00adti\u00adon die Ergeb\u00adnis\u00adse in die ein oder ande\u00adre Rich\u00adtung beein\u00adflus\u00adsen; wie man m\u00f6ch\u00adte. Da hier aber kei\u00adne Pro\u00adduk\u00adte des Mit\u00adbe\u00adwerbs getes\u00adtet wur\u00adden, son\u00addern aus\u00adschlie\u00df\u00adlich haus\u00adei\u00adge\u00adne Gra\u00adfik\u00adkar\u00adten, las\u00adsen wir das mal so ste\u00adhen und tun f\u00fcr einen Moment so, als w\u00e4ren es unab\u00adh\u00e4n\u00adgi\u00adge Tests.<\/p>\n

Sofern die\u00adse Wer\u00adte tat\u00ads\u00e4ch\u00adlich die Rea\u00adli\u00adt\u00e4t abbil\u00adden, w\u00e4re der Per\u00adfor\u00admance-Gewinn durch Vega II<\/span> \u2013 par\u00addon, Rade\u00adon VII<\/span> \u2013 respek\u00adta\u00adbel. Zwi\u00adschen 7,5 % und 68,4 % liegt der Leis\u00adtungs\u00adge\u00adwinn, im Mit\u00adtel bei etwa 30 %. Daf\u00fcr, dass es sich \u2013 anders als bei der Vega 64 \u2013 nicht um einen Voll\u00adaus\u00adbau des Chips han\u00addelt, ist das bemer\u00adkens\u00adwert. Doch woher k\u00f6nn\u00adte der Leis\u00adtungs\u00adge\u00adwinn kom\u00admen? Zun\u00e4chst ein\u00admal nat\u00fcr\u00adlich vom Shrink auf 7 nm, des\u00adsen Effi\u00adzi\u00adenz\u00adge\u00adwinn man ent\u00adwe\u00adder mit\u00adneh\u00admen oder in h\u00f6he\u00adre Takt\u00adra\u00adten inves\u00adtie\u00adren kann, wobei hier offen\u00adbar Letz\u00adte\u00adres gesche\u00adhen ist. Fer\u00adner wur\u00adde die Spei\u00adcher\u00adband\u00adbrei\u00adte gegen\u00ad\u00fcber Vega 64 mehr als ver\u00addop\u00adpelt; es ste\u00adhen nun bei\u00adna\u00adhe absurd hohe 1 TB<\/span>\/s zur Ver\u00adf\u00fc\u00adgung. Da Vega 64 je nach Test oft im Fla\u00adschen\u00adhals \u201cSpei\u00adcher\u00adband\u00adbrei\u00adte\u201d fest\u00adhing, wirkt die erh\u00f6h\u00adte Spei\u00adcher\u00adleis\u00adtung f\u00fcr die Kar\u00adte wie ein Befrei\u00adungs\u00adschlag. Und last but not least: der Vega-20-Chip ist kein rei\u00adner Die-Shrink von Vega 10, wie man in der dama\u00adli\u00adgen Ver\u00ad\u00f6f\u00adfent\u00adli\u00adchung<\/a> nach\u00adle\u00adsen kann:<\/p>\n

Key fea\u00adtures of the AMD<\/span> Rade\u00adon Instinct\u2122 MI60<\/span> and MI50<\/span> acce\u00adle\u00adra\u00adtors include:<\/strong><\/p>\n

Opti\u00admi\u00adzed Deep Lear\u00adning Ope\u00adra\u00adti\u00adons:<\/strong> Pro\u00advi\u00addes fle\u00adxi\u00adble mixed-pre\u00adcis\u00adi\u00adon FP16<\/span>, FP32<\/span> and INT4<\/span>\/INT8<\/span> capa\u00adbi\u00adli\u00adties to meet gro\u00adwing demand for dyna\u00admic and ever-chan\u00adging workloads, from trai\u00adning com\u00adplex neu\u00adral net\u00adworks to run\u00adning infe\u00adrence against tho\u00adse trai\u00adned networks.
\nWorld\u2019s Fas\u00adtest Dou\u00adble Pre\u00adcis\u00adi\u00adon PCIe\u00ae2 Accelerator5:<\/strong> The AMD<\/span> Rade\u00adon Instinct\u2122 MI60<\/span> is the world\u2019s fas\u00adtest dou\u00adble pre\u00adcis\u00adi\u00adon PCIe 4.0 capa\u00adble acce\u00adle\u00adra\u00adtor, deli\u00adve\u00adring up to 7.4 TFLOPS<\/span> peak FP64<\/span> performance5 allo\u00adwing sci\u00aden\u00adtists and rese\u00adar\u00adchers to more effi\u00adci\u00adent\u00adly pro\u00adcess HPC<\/span> appli\u00adca\u00adti\u00adons across a ran\u00adge of indus\u00adtries inclu\u00adding life sci\u00aden\u00adces, ener\u00adgy, finan\u00adce, auto\u00admo\u00adti\u00adve, aero\u00adspace, aca\u00adde\u00admics, govern\u00adment, defen\u00adse and more. The AMD<\/span> Rade\u00adon Instinct\u2122 MI50<\/span> deli\u00advers up to 6.7 TFLOPS<\/span> FP64<\/span> peak performance1, while pro\u00advi\u00adding an effi\u00adci\u00adent, cost-effec\u00adti\u00adve solu\u00adti\u00adon for a varie\u00adty of deep lear\u00adning workloads, as well as enab\u00adling high reu\u00adse in Vir\u00adtu\u00adal Desk\u00adtop Infra\u00adstruc\u00adtu\u00adre (VDI<\/span>), Desk\u00adtop-as-a-Ser\u00advice (DaaS) and cloud environments.
\nUp to 6X<\/span> Fas\u00adter Data Trans\u00adfer:<\/strong> Two Infi\u00adni\u00adty Fabric\u2122 Links per GPU<\/span> deli\u00adver up to 200 GB<\/span>\/s of peer-to-peer band\u00adwidth \u2013 up to 6X<\/span> fas\u00adter than PCIe 3.0 alone4 \u2013 and enable the con\u00adnec\u00adtion of up to 4 GPUs in a hive ring con\u00adfi\u00adgu\u00adra\u00adti\u00adon (2 hives in 8 GPU<\/span> servers).
\nUltra-Fast HBM2<\/span> Memo\u00adry:<\/strong> The AMD<\/span> Rade\u00adon Instinct\u2122 MI60<\/span> pro\u00advi\u00addes 32GB<\/span> of HBM2<\/span> Error-cor\u00adrec\u00adting code (ECC<\/span>) memory6, and the Rade\u00adon Instinct\u2122 MI50<\/span> pro\u00advi\u00addes 16GB<\/span> of HBM2<\/span> ECC<\/span> memo\u00adry. Both GPUs pro\u00advi\u00adde full-chip ECC<\/span> and Relia\u00adbi\u00adli\u00adty, Acces\u00adsi\u00adbi\u00adli\u00adty and Ser\u00advicea\u00adbi\u00adli\u00adty (RAS<\/span>)7 tech\u00adno\u00adlo\u00adgies, which are cri\u00adti\u00adcal to deli\u00adver more accu\u00adra\u00adte com\u00adpu\u00adte results for lar\u00adge-sca\u00adle HPC<\/span> deployments.
\nSecu\u00adre Vir\u00adtua\u00adli\u00adzed Workload Sup\u00adport:<\/strong> AMD<\/span> MxG\u00adPU Tech\u00adno\u00adlo\u00adgy, the industry\u2019s only hard\u00adware-based GPU<\/span> vir\u00adtua\u00adliza\u00adti\u00adon solu\u00adti\u00adon, which is based on the indus\u00adtry-stan\u00addard SR-IOV<\/span> (Sin\u00adgle Root I\/O Vir\u00adtua\u00adliza\u00adti\u00adon) tech\u00adno\u00adlo\u00adgy, makes it dif\u00adfi\u00adcult for hackers to attack at the hard\u00adware level, hel\u00adping pro\u00advi\u00adde secu\u00adri\u00adty for vir\u00adtua\u00adli\u00adzed cloud deployments.<\/p><\/blockquote>\n

Ob alles davon auch bei der End\u00adnut\u00adzer-Vari\u00adan\u00adte Rade\u00adon VII<\/span> frei\u00adge\u00adschal\u00adtet ist, darf zwar bezwei\u00adfelt wer\u00adden, aber es zeigt: Vega 20 ist mehr als nur eine Vega 10 in 7 nm. So wur\u00adde, um ein f\u00fcr Gam\u00ading rele\u00advan\u00adtes kon\u00adkre\u00adtes Bei\u00adspiel zu nen\u00adnen, auch die Anzahl der Ras\u00adter-Ope\u00adra\u00adti\u00adons-Pipe\u00adlines (ROPs) auf 128 ver\u00addop\u00adpelt, was ins\u00adbe\u00adson\u00adde\u00adre bei Spie\u00adlen in \u201cschar\u00adfen\u201d Ein\u00adstel\u00adlun\u00adgen (hohe Anti-Ali\u00ada\u00adsing-Ein\u00adstel\u00adlung, etc.) einen Boost bewirk\u00adten sollte.<\/p>\n","protected":false},"excerpt":{"rendered":"

Bereits vor\u00adges\u00adtern haben wir \u00fcber AMDs Vor\u00adstel\u00adlung der neu\u00aden High-End-Gra\u00adfik\u00adkar\u00adte AMD<\/span> Rade\u00adon VII<\/span> berich\u00adtet. W\u00e4h\u00adrend wir sowohl \u00fcber die tech\u00adni\u00adschen Daten als auch die zuge\u00adh\u00f6\u00adri\u00adge Pres\u00adse\u00admit\u00adtei\u00adlung bereits berich\u00adtet haben, war uns gar nicht auf\u00adge\u00adfal\u00adlen, dass AMD<\/span> in den Fu\u00df\u00adno\u00adten der Pres\u00adse\u00ader\u00adkl\u00e4\u00adrung wei\u00adte\u00adre 25 kon\u00adkre\u00adte Bench\u00admark-Wer\u00adte ver\u00adsteckt hat\u00adte. (\u2026) Wei\u00adter\u00adle\u00adsen \u00bb<\/a><\/p>\n","protected":false},"author":2,"featured_media":43160,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"wp_typography_post_enhancements_disabled":false,"ngg_post_thumbnail":0,"footnotes":""},"categories":[12],"tags":[966,1209,1620,1619,1709,1102],"class_list":["post-43187","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aktuelles","tag-amd","tag-instinct","tag-mi50","tag-mi60","tag-radeon-vii","tag-vega","entry"],"share_on_mastodon":{"url":"","error":""},"_links":{"self":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/43187","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\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/comments?post=43187"}],"version-history":[{"count":16,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/43187\/revisions"}],"predecessor-version":[{"id":43346,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/posts\/43187\/revisions\/43346"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media\/43160"}],"wp:attachment":[{"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/media?parent=43187"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/categories?post=43187"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.planet3dnow.de\/cms\/wp-json\/wp\/v2\/tags?post=43187"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}