Izwe lobuhlakani bokwenziwa libhekene noshintsho olungalindelekile. Ngokusho kombiko we-The Information, i-OpenAI ithengile amashumi ezinkulungwane zama-Mac minis nama-Mac Studios ezinyangeni zamuva nje. Umgomo uwukuqeqesha ama-ejenti ayo e-AI ukuthi asebenzise ikhompyutha njengomuntu: ukuchofoza, ukuzulazula kumamenyu, ukugcwalisa amafomu, nokwenza imisebenzi eyinkimbinkimbi. Leli cebo, elimangaze abaningi, libonisa ukuthi i-hardware ye-Apple ayiseyona eyabathengi nabadali kuphela, kodwa isibe yingxenye ebalulekile engqalasizinda yamalebhu amakhulu e-AI.
Lezi zindaba, ezibikwe futhi yizitolo ezifana neCult of Mac, Moneycontrol, kanye neWall St. engama-24/7, zibonisa ukuthi lawa makhompyutha asetshenziswa esigabeni esithile sokuqeqeshwa: ukufunda okuqiniswayo. Kule nqubo, amamodeli e-AI asebenzisana nezixhumi zangempela zezithombe, ahlola izenzo futhi athola imivuzo noma izinhlawulo ngokusekelwe emphumeleni. Le ndlela idinga inani elikhulu lenkumbulo kanye nekhono lokusebenzisa izinkulungwane zokulingisa ngesikhathi esifanayo, into ama-Mac anama-chip e-Apple Silicon abonakala efaneleka kakhulu kuyo.
Kungani i-Apple Silicon ibalulekile ekufundeni ukuqinisa
Ukuqeqesha i-ejenti ye-AI okumele isebenzise ikhompyutha kuhluke kakhulu ekuqeqesheni imodeli yolimi yendabuko. Esikhundleni sokuthola umbhalo omningi, i-ejenti kufanele chaza lokho okubonayo esikrinini, unqume ukuthi yisiphi isinyathelo okufanele usithathe, bese uhlola ukuthi siyasebenza yini.Umzamo ngamunye uyahlolwa, futhi imodeli ilungisa isu layo ngokusekelwe kuleyo mpendulo. Ukuze lokhu kwenzeke, kudingeka izinkulungwane zezikhathi ezizimele, ngasinye sisebenza ngesikhathi esisodwa ngesimo saso kanye nomongo waso.
Yilapho i-Apple's Unified Memory Architecture (UMA) idlala khona indima. Ngokungafani nezinhlelo zendabuko ezisekelwe ku-GPU, lapho imemori yevidiyo ne-RAM kuhlukene khona futhi idatha kufanele ihambe iye phambili naphambili, kuma-Mac i-CPU, i-GPU, kanye ne-Neural Engine babelana nge-memory block efanayo. Lokhu kunciphisa izithiyo futhi kuvumela imodeli ukuthi ifinyelele ngokushesha ulwazi eludingayo nganoma yisiphi isikhathi. Njengoba ukuhlaziya okuningana kubonisa, lesi sici sibalulekile ku-ejenti ukuthi igcine ukumelwa okulinganayo kwesikrini, imiyalelo, kanye nemiphumela yezenzo zayo.
Enye inzuzo ye-Mac mini kanye ne-Mac Studio uhlelo lwazo lokupholisa. Ngokungafani nama-MacBook, amancane futhi alula, lawa makhompyutha edeskithophu anabalandeli abazinikele abawavumela ukuthi alondoloze ukusebenza okuzinzile amahora noma ngisho nezinsuku zokuqeqeshwa okujulile. Lokhu kubalulekile ekufundeni okuqinisiwe, okungadinga izikhathi ezinde kakhulu ngaphandle kokushisa kakhulu kwekhompyutha kanye nokunciphisa ijubane.
Ukuthenga kwe-OpenAI, intela ye-Anthropic: amasu amabili okufeza okufanayo
Lezi zindaba ziphinde zembule ukuthi i-OpenAI akuyona yodwa inkampani enentshisekelo kule ndlela. I-Anthropic, esekelwa yi-Google kanye ne-Amazon, kubikwa ukuthi ikhethe isu elihlukile: esikhundleni sokuthenga imishini, iqasha i-Mac mini capacity nge-Amazon Web Services (AWS)Lo mehluko ubalulekile ngoba ubonisa ukuthi awudingi ukutshala imali eningi kwihadiwe ukuze uhlole ukuthi lo mklamo uyasebenza yini. Ungaqasha umthamo, ulinganise ukusebenza, bese unquma ukuthi kufanelekile yini ukuwuthenga noma ukuqhubeka nokuqasha.
Intshisekelo yala ma-lab kuma-Mac akusho ukuthi azothatha indawo yamaqoqo amakhulu e-GPU e-Nvidia. Ukuqeqeshwa kwemodeli yolimi olukhulu kusathembele kulawo ma-accelerator akhethekile. Kodwa-ke, emisebenzini ejulile yokuqonda kanye nama-ejenti okuqeqesha asebenzisana nama-interface, ama-Mac anikeza enye indlela engabizi kakhulu nephumelelayo. Njengoba umhlaziyi uShay Boloor echaza, amakhompyutha e-Apple asetshenziselwa isigaba sokufunda sokuqinisa, kuyilapho ama-GPU ehlala ebalulekile ekuqeqeshweni kwangaphambi kokuqeqeshwa.
Isidingo sibeka ingcindezi ochungechungeni lokuhlinzeka futhi i-Apple iyasabela
Lo mkhuba we-Mac awuzange ubonakale ohlwini lokuthengwa kwempahla. Ngokusho kwemibiko eminingana, Ukucushwa kwe-Mac mini kanye ne-Mac Studio okunememori eyengeziwe sekuphele izinyanga eziningi. Futhi izikhathi zokulethwa kwezinguqulo ezenziwe ngokwezifiso sezithathe amasonto amaningi. Ukuntuleka komhlaba wonke kwememori ye-DRAM ne-NAND, okubangelwa ukufunwa okuvela ezikhungweni zedatha, kuye kwenza isimo saba sibi kakhulu. Abaphakeli babeka phambili ama-hyperscalers amakhulu, kanti ama-oda e-Apple ayancintisana emgqeni ofanayo.
Ukuze kuxazululwe lesi simo, i-Apple yasheshisa isimiso sayo sokuvuselela. Ngo-Agasti 25, 2026, inkampani yethule i-Mac mini entsha enama-chip e-M6 kanye ne-M5 Pro, kanye ne-Mac Studio ene-M5 Max kanye ne-M5 Ultra. Lesi sinyathelo sokuqala esingavamile siphendula isidingo sokugcwalisa imakethe ngama-chip anamandla futhi asebenzayo. Amamodeli amasha afaka phakathi ukuthuthukiswa okukhulu ekusebenzeni kanye nememori, ngezinketho ezifinyelela ku-512GB yesitoreji esihlanganisiwe ku-Mac Studio M5 Ultra.
Isinqumo se-Apple sokunyusa ukwethulwa kwale nkampani sihunyushwe njengempendulo ekucindezelweni kwesidingo. Le nkampani ayikaqinisekisi ngokusemthethweni ukuthengwa kwe-OpenAI, kodwa izikhathi zokulethwa kanye nokuntuleka kwesitoko kusikisela ukuthi isidingo sezikhungo singokoqobo. Eqinisweni, imali engenayo evela esigabeni se-Mac ikhule ngo-29% unyaka nonyaka kwikota edlule, yafinyelela ku-$10.400 billion, okwenza kwaba umugqa webhizinisi okhula ngokushesha we-Apple.
I-Nvidia ibukele eduze njengoba i-Apple izibeka emakethe yebhizinisi
Ukwanda kwama-Mac emkhakheni we-AI yendawo akuqashelwanga yiNvidia. Ngokusho komthombo oseduze nenkampani, I-Nvidia isivele ibheka i-Apple njengomncintiswano wayo omkhulu ekucutshungulweni kwe-AI yendawoUkuze kulwiwe nalolu songo, iNvidia yethule i-DGX Spark ngasekupheleni konyaka odlule, ikhompyutha yedeskithophu enomklamo ofana ne-Mac mini, efakwe i-chip ye-Grace Blackwell kanye nememori ehlanganisiwe efinyelela ku-128 GB. Lo mkhiqizo uhloselwe ngqo emakethe ye-desktop AI inference.
I-Apple, yona, izama ukusebenzisa lesi sikhathi ukuqinisa isikhundla sayo emakethe yebhizinisi. NgoJuni, inkampani ibambe umcimbi wangasese e-Apple Park obizwa ngokuthi "iBusiness at the Park," owawuhanjelwe abaphathi abavela ku-Disney, Ford, kanye nomsunguli we-Anthropic uJared Kaplan. Kulo mcimbi, i-Apple igcizelele izinzuzo zehadiwe yayo yokucubungula imisebenzi ye-AI endaweni, kanti i-Mac mini ithatha indawo ephambili. Kodwa-ke, inkampani isenalo ithimba lokuthengisa elizinikele lesigaba se-AI yebhizinisi, into abanye abaphathi bangaphambilini abaye bayikhomba njengethuba elilahlekile.
Ukuntuleka kokunikezwa kwempahla nakho kudala amamodeli ebhizinisi amasha. UPeter Voell, owayeyisisebenzi se-OpenAI, usungule iMount Thor, inkampani yokubala amafu esekelwe kwihadiwe ye-Apple esasebenza ngendlela eyimfihlo. Ezinye izinkampani, njengeNamespace Labs, ziye zasebenzisa i-racking MacBook Pros ukuze zihlangabezane nesidingo samakhasimende. Le ndlela entsha yokusebenza ibonisa ukuthi intshisekelo kuma-Mac njengengqalasizinda ye-AI idlulela ngale kwama-laboratories amakhulu.
Ngamafuphi, izindaba zokuthi i-OpenAI ithengile amashumi ezinkulungwane zama-Mac minis nama-Mac Studios ukuze iqeqeshe ama-ejenti ayo e-AI ziphawula ushintsho olukhulu embonini. Inkumbulo ehlanganisiwe ye-Apple Silicon, ukusebenza kahle kwamandla, kanye nekhono lokuqhuba izinkulungwane zezikhathi ngasikhathi sinye Lawa madivayisi abe yithuluzi eliwusizo lokufunda ukuqinisa. Nakuba i-Apple noma i-OpenAI ingazange iqinisekise ngokusemthethweni lesi sivumelwano, izibalo zokuthengisa, ukuntuleka kwamasheya, kanye nokukhishwa kwasekuqaleni kusikisela ukuthi lo mkhuba ungokoqobo. Kusazobonakala ukuthi i-Nvidia izosabela kanjani nokuthi i-Apple izolisebenzisa yini leli thuba ukuqinisa isikhundla sayo emakethe ye-AI yebhizinisi.
