Me ya faru
Masu bincike daga Jami'ar Fasaha ta Sydney, Jami'ar ShanghaiTech, da Jami'ar Fasaha ta Munich sun fito da IC-ThermBench, sabon ma'auni na buɗaɗɗen ma'auni don kimanta ƙirar ƙirar thermal na AI-kore a cikin manyan marufi na semiconductor. Alamar ma'auni ta haɗa da bayanan samfuran samfuran 50,000 musamman waɗanda ke yin niyya na gine-ginen chiplet na 2.5D, tare da ayyukan da ake da su don tsayuwar 3D-IC da kuma nazarin yanayin zafi na wucin gadi.
Aikin IC-ThermBench, dalla-dalla a cikin takarda ta Hang et al., yayi bayani game da haɓakar haɓakar ƙirar ƙirar zafi a ƙirar guntu na zamani. Hanyoyin kwaikwaiyo na al'ada galibi suna da tsadar ƙididdiga, wanda ke haifar da ɗaukar ƙwararrun masu maye gurbin AI. Koyaya, waɗannan samfuran galibi suna kokawa da yanayin 'babu rarrabawa' (OOD) inda ƙirar guntu ta karkata daga bayanan horo.
An tsara ma'auni don zama mai ci gaba, ma'ana yana gwada samfura akan ayyuka masu wahala. Ya haɗa da tsawo na 50,000-samfurin musamman don 2.5D chiplets, waɗanda ke ƙara zama ruwan dare a cikin AI accelerators da manyan na'urori masu sarrafawa. Manufar ita ce kimanta yadda samfuran AI za su iya canja wurin ilimin ilimin kimiyyar zafin jiki da suka koya a cikin nau'ikan fakiti daban-daban da daidaitawar jiki.
Bayanan tushe: semiengineering.com ↗
Me ya sa yake da mahimmanci
Kamar yadda ƙirar semiconductor ke motsawa zuwa 2.5D da 3D gine-gine don shawo kan ƙulla-ƙulla, sarrafa zafin jiki ya zama ƙalubale na injiniya mai mahimmanci. Ana ƙara amfani da samfuran AI don hasashen rarraba zafi a cikin waɗannan ƙananan kwakwalwan kwamfuta masu launuka iri-iri. IC-ThermBench yana ba da ingantacciyar hanya don auna yadda waɗannan samfuran ke gabaɗaya gabaɗayan bambance-bambancen jiki daban-daban da daidaitawar fakitin giciye, wanda ke da mahimmanci don tabbatar da amincin kayan aikin kwamfuta mai girma na gaba. Ta hanyar ba da tsarin ƙima na ci gaba, maƙasudin yana taimaka wa masu bincike gano inda samfuran AI suka kasa yin lissafi don haɗaɗɗun ƙarfin zafi a cikin shimfidar guntu marasa al'ada, mai yuwuwar haɓaka haɓaka ingantaccen kayan aikin siminti na thermal don masana'antar semiconductor.
Gudanar da thermal shine babban hani a cikin ƙirar 3D-ICs, inda yawan zafi ya fi girma fiye da guntuwar 2D na gargajiya. Samar da yanayin zafi mara daidai zai iya haifar da maƙarƙashiya ko gazawar hardware.
Madaidaitan ma'auni kamar IC-ThermBench wajibi ne don matsar da ƙirar thermal AI daga binciken ilimi zuwa ingantaccen aikin masana'antu. Ta hanyar samar da ma'auni na gama-gari, masu binciken suna nufin haɓaka ingantacciyar ƙira a cikin ƙira, tabbatar da cewa za su iya sarrafa sassa daban-daban da haɓakar ƙirar tushen chiplet.
Ingantacciyar hanyar sadarwa: Yadda A zahiri yake Aiki
Bincika fasahar da ke bayan wannan ci gaban ta hanyar mu'amala.
Which component of an AI application is the machine-learning model itself?
Abin kallo na gaba
Ƙungiyar binciken ta buga maƙasudin a matsayin bugu na arXiv (arXiv:2608.23977). Ci gaba da gaba za su iya mayar da hankali kan ko masu samar da kayan aikin EDA (Electronic Design Automation) na masana'antu sun ɗauki waɗannan ma'auni don tabbatar da nasu fasalin siminti na thermal na AI. Ya rage a ga yadda yadda waɗannan samfuran ke aiki yadda ya kamata idan aka yi amfani da su zuwa ga ainihin duniya, ƙirar guntu na mallakar mallaka waɗanda za su iya bambanta sosai da madaidaitan bayanan roba ko na ilimi.
A halin yanzu ana samun ma'auni a matsayin tushen tushen albarkatu don al'ummar bincike. Ya kamata masu sa ido su saka idanu ko wannan ma'auni ya zama madaidaicin tunani don takaddun simintin thermal da AI ke tukawa a cikin semiconductor da taron koyon injin mai zuwa.
Wani mahimmin abin da ba a san shi ba shine iyakar abin da waɗannan ma'auni na ilimi suka daidaita tare da aikin software na simintin simintin simintin gyare-gyaren da manyan masana'antun ke amfani da su.