Kamar yadda ya zo daga bayanin da Jim Keller, babban mataimakin shugaban fasaha da tsarin gine-gine a Intel, ya shaida wa duniya, kamfaninsa a halin yanzu yana aiki don ƙirƙirar sabon microarchitecture, wanda ya kamata ya zama "mafi girma da kusanci ga dogaro na layi na aiki. akan adadin transistor,” fiye da ƙirar zamani na Sunny Cove. A bayyane yake, ya kamata a fassara wannan ta hanyar da a cikin ƴan shekaru za mu sami na'ura mai sarrafawa wanda zai fi rikitarwa da mahimmanci fiye da CPUs wanda giant microprocessor ke bayarwa a halin yanzu.
Sabon microarchitecture na Sunny Cove, wanda Intel ke amfani da shi a cikin sabbin na'urori masu sarrafa Ice Lake, ya zama babban ci gaba, tunda bayan dogon hutu ya ƙara haɓaka IPC (yawan umarnin da ake aiwatar da kowane agogo). Amma guru mai sarrafawa Jim Keller, wanda a halin yanzu yana aiki da Intel, ya ce wannan yayi nisa daga ƙarshen batu. Yanzu yana aiki akan ƙarni na gaba na microarchitecture, wanda zai iya yin cikakken amfani da haɓaka da yawa a cikin kasafin kuɗin transistor da ake sa ran nan da shekaru masu zuwa.
Dangane da ƙididdigar Intel, fa'idar muryoyin Sunny Cove idan aka kwatanta da kogin Coffee Lake a takamaiman aikin ya kai 15-18% (a cikin saurin agogo ɗaya). Koyaya, kasafin kudin transistor na Sunny Cove ya zarce kasafin wanda ya gabace shi da wani adadi mai mahimmanci - kusan kashi 38%. Dangane da bayanin da Keller ya bayyana, ainihin da ke da microarchitecture na Sunny Cove ya ƙunshi transistors kusan miliyan 300 10-nm, yayin da babban tafkin Coffee ya haɗa da transistors kusan miliyan 217 14-nm. Ya bayyana cewa haɓakar haɓakawa a cikin Sunny Cove bai kai ga dogaro na layi ba akan girman kasafin kuɗi na transistor: ci gaban yawan aiki ya zama kusan rabin sauri kamar haɓakar rikitaccen kristal na semiconductor. A cewar Keller, bai kamata ya kasance haka ba.
Da yake ba da lacca a Jami'ar Berkeley, babban ƙwararre daga Intel ya tabo batun juyin halitta na microarchitectures na masu sarrafa Intel kuma a cikin labarin bai tsaya a kan Sunny Cove ba, amma ya ambaci mai yiwuwa magajin wannan microarchitecture: “Sunny Cove yana aiki tare da Umarni 800 a lokaci guda, aiwatarwa daga 3 zuwa 6 x86- umarni a kowane lokaci Amma muna aiki akan ƙarni na microarchitecture wanda ya fi girma, kuma ka'idar haɓaka aikin ta fi kusa da layi. Gaskiya babban sauyi ne a tunani."
Matsayin injiniyan tauraro shine fasahar sarrafawa har yanzu ba ta kai ga kowane iyaka ba. A cewar Keller, Intel yana da kyawawan tsare-tsare na gaba, waɗanda suka haɗa da haɓakar adadin transistor sau 50 a cikin na'urori masu sarrafawa da manyan haɓakawa a kusan kowane rukunin aiki. Kuma babu wani abu mai wuya game da wannan. Kamar yadda Keller ya yi bayani: “Mutane da yawa ne ke ƙirƙira na’urori, amma a zahiri ɗimbin ƙananan ƙungiyoyi ne. Kuna iya inganta hasashen reshe, saitin koyarwa, gine-gine, yin haɓakawa, amfani da kayan aikin ƙira mafi kyau da mafi kyawun ɗakunan karatu. Adadin wuraren aikace-aikacen daban-daban inda akwai dakin ƙirƙira yana da girma da yawa sosai.
Tsare-tsaren jama'a na Intel na yanzu sun haɗa da nau'ikan kayan haɓaka microarchitecture guda biyu bayan Sunny Cove. An yi alƙawarin ƙira na Willow Cove na gaba don nuna canje-canje ga tsarin tsarin cache da ƙaura zuwa sabuwar fasahar semiconductor (wataƙila 7nm). Golden Cove za ta ƙara yin aiki mai zare guda ɗaya da mai da hankali kan ayyukan AI, tare da haɓakawa da ake buƙata don ingantaccen aiki akan hanyoyin sadarwar XNUMXG. Wataƙila Jim Keller yana da Golden Cove a cikin rahotonsa, kodayake ba a faɗi komai ba game da wannan.
source: 3dnews.ru