{"id":63961,"date":"2024-07-22T12:36:05","date_gmt":"2024-07-22T12:36:05","guid":{"rendered":"https:\/\/ekamu.net\/?p=63961"},"modified":"2024-07-22T12:36:05","modified_gmt":"2024-07-22T12:36:05","slug":"amdye-gun-dogdu-nvidianin-cuda-hakimiyet-scale-ile-bitecek","status":"publish","type":"post","link":"https:\/\/ekamu.net\/index.php\/2024\/07\/22\/amdye-gun-dogdu-nvidianin-cuda-hakimiyet-scale-ile-bitecek\/","title":{"rendered":"AMD\u2019ye g\u00fcn do\u011fdu: Nvidia\u2019n\u0131n CUDA hakimiyet SCALE ile bitecek"},"content":{"rendered":"

<\/span> \u0130ngiliz giri\u015fim Spectral Compute<\/strong>, NVIDIA CUDA uygulamalar\u0131n\u0131n ek bir \u00e7al\u0131\u015fma gerektirmeden AMD ekran kartlar\u0131yla yerel olarak derlenmesini sa\u011flayan yeni bir GPGPU ara\u00e7 zinciri seti olan SCALE<\/strong>\u2019i tan\u0131tt\u0131. Nvidia\u2019n\u0131n CUDA hakimiyeti firmay\u0131 do\u011frudan yapay zeka d\u00fcnyas\u0131n\u0131n tek hakimi haline getirmi\u015fti. CUDA\u2019n\u0131n AMD GPU\u2019larla kullan\u0131labilmesi ise Nvidia\u2019n\u0131n hakimiyetini sarsabilir. <\/figure>\n

CUDA, AMD GPU\u2019lar\u0131yla da \u00e7al\u0131\u015facak<\/b><\/p>\n

G\u00f6r\u00fcn\u00fc\u015fe g\u00f6re end\u00fcstri Nvidia\u2019n\u0131n yaz\u0131l\u0131m y\u0131\u011f\u0131n\u0131 hakimiyetini k\u0131rmay\u0131 ba\u015fard\u0131. Daha \u00f6nce CUDA k\u00fct\u00fcphanelerinin AMD’nin ROCm<\/strong>‘si ile \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flayan a\u00e7\u0131k kaynakl\u0131 bir ta\u015f\u0131ma projesi olan ZLUDA’n\u0131n ortaya \u00e7\u0131k\u0131\u015f\u0131n\u0131 g\u00f6rm\u00fc\u015ft\u00fck. Ve \u015fimdi Nvidia\u2019n\u0131n CUDA yeteneklerinden yararlanarak \u00fcst d\u00fczey hibrit model olu\u015fturulmas\u0131na olanak tan\u0131yan SCALE<\/strong> adl\u0131 yeni bir ara\u00e7 sahneye giriyor.<\/p>\n

CUDA, geli\u015fmi\u015f sim\u00fclasyon, fizik hesaplamalar\u0131, bilimsel modelleme, kripto madencili\u011fi ve evet, makine \u00f6\u011frenimi i\u00e7in GPU’lar\u0131n sundu\u011fu g\u00fc\u00e7ten yararlanmak i\u00e7in kullan\u0131l\u0131yor. Makine \u00f6\u011frenimindeki g\u00fc\u00e7l\u00fc durum nedeniyle yapay zeka patlamas\u0131n\u0131n ortas\u0131nda firmalar, Nvidia\u2019n\u0131n GPU\u2019lar\u0131n\u0131 tercih etti. Yayg\u0131n olarak end\u00fcstri standard\u0131 olarak g\u00f6r\u00fclen bu durum, AMD gibi rakiplerin GPU’lar\u0131n\u0131n yaz\u0131l\u0131m i\u00e7in ba\u015fka yerlere bakmas\u0131na neden olmu\u015ftu.<\/p>\n

Aktar\u0131lanlara g\u00f6re SCALE, CUDA program\u0131n\u0131n veya yap\u0131 sisteminin de\u011fi\u015ftirilmesini gerektirmiyor. Bu konuda a\u00e7\u0131klama yapan Spectral Compute CEO’su Michael Sondergaard, \u201cKodu bir kez yazman\u0131n ve herhangi bir donan\u0131m platformunda olu\u015fturman\u0131n\/\u00e7al\u0131\u015ft\u0131rman\u0131n m\u00fcmk\u00fcn olmas\u0131 gerekti\u011fine inan\u0131yoruz. Bu, CPU kodu i\u00e7in uzun y\u0131llard\u0131r bir ger\u00e7eklikti, \u00f6yleyse neden GPU’lar i\u00e7in de olmas\u0131n?\u201d<\/em> ifadelerini kulland\u0131.<\/p>\n

Spectral Compute’a g\u00f6re SCALE yedi y\u0131ld\u0131r geli\u015ftiriliyor<\/strong> ve Nvidia\u2019n\u0131n koduna g\u00fcvenmiyor ancak CUDA uyumlu bir ara\u00e7 zincirini olu\u015fturuyor, bu da SCALE’i AMD’nin RDNA GPU’lar\u0131 gibi birden fazla platform aras\u0131nda son derece uyarlanabilir hale getiriyor. Dolay\u0131s\u0131yla SCALE, CUDA Toolkit’e benzer \u015fekilde \u00e7al\u0131\u015f\u0131yor ve ara\u00e7 setini basit\u00e7e taklit ediyor. B\u00f6ylelikle ara\u00e7lar ve komut dosyalar\u0131 do\u011frudan \u00e7al\u0131\u015ft\u0131r\u0131labiliyor. Spectral Compute ayr\u0131ca Blender, Llama-cpp, XGboost, FAISS, GOMC ve daha fazlas\u0131 gibi, t\u00fcm\u00fc AMD GPU’larda \u00e7al\u0131\u015fan uygulamalarla SCALE’i ba\u015far\u0131yla test ettiklerini de s\u00f6yl\u00fcyor.<\/p>\n\n

<\/span><\/p>\n

Kaynak :\u00a0https:\/\/www.donanimhaber.com\/amd-ye-gun-dogdu-nvidia-nin-cuda-hakimiyet-scale-ile-bitecek–179739<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

\u0130ngiliz giri\u015fim Spectral Compute, NVIDIA CUDA uygulamalar\u0131n\u0131n ek bir \u00e7al\u0131\u015fma gerektirmeden AMD ekran kartlar\u0131yla yerel olarak derlenmesini sa\u011flayan yeni bir GPGPU ara\u00e7 zinciri seti olan SCALE\u2019i tan\u0131tt\u0131. Nvidia\u2019n\u0131n CUDA hakimiyeti firmay\u0131 do\u011frudan …<\/p>\n","protected":false},"author":1,"featured_media":63962,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[122,256,2682,3198],"class_list":["post-63961","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-teknoloji","tag-amd","tag-arac","tag-cuda","tag-scale"],"_links":{"self":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts\/63961","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/comments?post=63961"}],"version-history":[{"count":1,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts\/63961\/revisions"}],"predecessor-version":[{"id":63964,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/posts\/63961\/revisions\/63964"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/media\/63962"}],"wp:attachment":[{"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/media?parent=63961"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/categories?post=63961"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ekamu.net\/index.php\/wp-json\/wp\/v2\/tags?post=63961"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}