I-Rack Yeseva Entsha kaNvidia Izosebenzisa I-AI Chips Eyenziwe Izimbangi
I-Rack Yeseva Entsha kaNvidia Izosebenzisa I-AI Chips Eyenziwe Izimbangi Ku-pivot yamasu anesibindi, i-Nvidia ikhuluma ngokuncintisana okukhulayo emakethe ye-chip ye-AI ngokwembula i-server rack entsha eklanyelwe ukusebenzisa hhayi i-hardware yayo kuphela kodwa nama-chips enziwe izimbangi zayo. Lesi sinyathelo sibeka i-Nvidia ukuthi izuze kulo lonke uhlelo lwengqalasizinda ye-AI, ngisho nabaqhudelana nabo abafana ne-AMD kanye ne-Intel inselele ukubusa kwayo. Isimemezelo, esenziwe engqungqutheleni yaminyaka yonke yonjiniyela be-GTC, sibonisa ushintsho olubalulekile ekuthengiseni ama-chips nje kuphela ukuya ekunikezeni izicubu ezibalulekile ezixhumene nezikhungo zedatha zesimanje.
I-Nvidia's Strategic Pivot Emakethe Yokuncintisana Imakethe yokusheshisa i-AI ishisa ngokushesha. Sekuyiminyaka, ama-GPU kaNvidia bekuyinjini engaphikiswa ye-boom yobuhlakani bokwenziwa. Kodwa-ke, isimo siyashintsha njengoba izinkampani ezinkulu zobuchwepheshe kanye nezimbangi ze-semiconductor zitshala imali eningi ekwakheni ama-chips azo e-AI, avame ukubizwa ngokuthi ama-ASIC. Lo mkhuba, owaziwa ngokuthi "i-silicon diversification," usongela isabelo semakethe sikaNvidia. Izinkampani ezifana ne-Google, i-Amazon, ne-Microsoft zidizayina ama-chips enzelwe imithwalo yawo ethile yomsebenzi. Ngaleso sikhathi, i-AMD ne-Intel bethula ezinye izindlela zokuncintisana ze-GPU. Irekhi yeseva entsha ye-Nvidia iyimpendulo eqondile kule ngcindezi yokuncintisana, iqinisekisa ukuthi ubuchwepheshe bayo buhlala bumaphakathi kungakhathaliseki ukuthi i-silicon kabani engaphakathi kweseva.
Kusukela ku-Chip Vendor kuya ku-Platform Provider Leli su likhuphula i-Nvidia kusuka kumnikezeli wengxenye kuya kumhlinzeki wenkundla obalulekile. Ngokwenza uhlelo lwe-rack olungakholelwa kumthengisi we-chip, i-Nvidia ibheja kunethiwekhi yayo kanye nesitaki sesofthiwe njengomsele wayo oyinhloko. Ubuchwepheshe benethiwekhi yenkampani, ikakhulukazi izinkundla zayo ze-InfiniBand kanye ne-Spectrum-X Ethernet, buthathwa njengesezingeni eliphezulu kakhulu lamaqoqo e-AI asebenza kahle kakhulu. I-rack entsha iguqula i-Nvidia ibe "isistimu yezinzwa ephakathi" yesikhungo sedatha ye-AI. Ngisho noma ikhasimende lisebenzisa chip ye-AI eliqhudelana nalo, lingase lithembele ku-Nvidia ukuze liqinisekise ukuthi lawo ma-chips angakwazi ukuxhumana ngesivinini nangesilinganiso esidingekayo ukuze kuqeqeshwe amamodeli ezilimi ezinkulu. Lokhu kudala ukusakazwa kwemali engenayo okunamandla, okuphindelelayo okungasengozini kakhulu emqhudelwaneni we-chip.
Ngaphakathi kwe-New Server Rack Architecture Ngokusho kwemithombo ejwayelene nephrojekthi, ukuqamba okusha okubalulekile kwe-server rack kulele kwendwangu yayo yokuxhumana ehlanganisiwe. Lobu buchwepheshe yisoso eyimfihlo exhuma okungenzeka ukuthi amakhulukhulu ama-I chips—kungaba avela ku-Nvidia noma omunye umkhiqizi—kuyise ikhompuyutha eyodwa, ehambisanayo. I-rack yakhelwe ukuxazulula ibhodlela elikhulu ku-AI enkulukazi: ukuxhumana kwama-chip. Ukuqeqesha amamodeli asezingeni eliphezulu kudinga izinkulungwane zama-chips ukuze asebenze ngokuzwana okuphelele. Ukudluliswa kwedatha okuhamba kancane noma okungathembeki phakathi kwama-chip kungakhubaza ukusebenza nokusebenza kahle, okwenza ungqimba lwenethiwekhi lubaluleke njengekhomputha ye-silicon ngokwayo.
Izingxenye Ezibalulekile Nobuchwepheshe I-rack isebenzisa ubuchwepheshe obujulile buka-Nvidia kukhompyutha esebenza kahle kakhulu. Ngenkathi amagama emikhiqizo ethile engazange adalulwe yimithombo, isistimu kungenzeka ihlanganise ubuchwepheshe obuningana obuyisisekelo be-Nvidia: Ukushintsha Kwenethiwekhi Yomkhawulokudonsa Ophakeme: Ukushintsha ngokwezifiso okusekelwe ekwakhiweni kwe-Nvidia's Spectrum noma i-InfiniBand ukuze ulawule ukugeleza kwedatha okukhulu. I-NVLink Technology: Ilungiselwe ukuhlinzeka ukuxhumana okusheshayo, okuqondile phakathi kwama-GPU kanye namanye ama-accelerator e-AI. I-Universal Chassis Design: Ifomu le-rack elimisiwe elingafaka amathreyi e-OEM ahlukahlukene aqukethe ama-chips ahlukene. Isofthiwe Yokuphatha Ehlanganisiwe: Isofthiwe yesistimu yokuhlela umthwalo wokusebenza, ukuqapha impilo, nokuthuthukisa ukusebenza kuyo yonke imiksi ye-silicon. Le ndlela inika opharetha besikhungo sedatha ukuguquguquka okungakaze kubonwe. Bangakwazi ukuhlanganisa futhi baqondanise ihadiwe ngokusekelwe ezindlekweni, ekusebenzeni, noma ezidingweni ezithile zomsebenzi, konke kuyilapho begcina ingqalasizinda ebumbene ephethwe inkundla ye-Nvidia.
Imithelela Ebanzi Yomkhakha we-AI Ukuhamba kuka-Nvidia kunemiphumela ebalulekile ye-ripple kuwo wonke umkhakha wezobuchwepheshe. Ivuma futhi isheshise umkhuba obheke ezindaweni zehadiwe ze-AI zabathengisi abaningi. Kumakhasimende ebhizinisi, lezi yizindaba ezinhle kakhulu, njengoba zithembisa ukukhetha okwengeziwe, izindleko ezingase zibe phansi, kanye nokunciphisa ukukhiya kwabathengisi. Kodwa-ke, iphinde ihlanganise amandla kaNvidia kwesendlalelo esihlukile sesitaki. Nguukulawula inkundla yokuxhumana nokuhlanganiswa, i-Nvidia ibeka izindinganiso zokusebenza nokusebenzisana. Abenzi bama-chip abaqhudelanayo manje kufanele baqinisekise ukuthi imikhiqizo yabo isebenza ngaphandle komthungo ngaphakathi kohlaka lwengqalasizinda okungenzeka luchazwe imbangi yabo enkulu.
Inkathi Entsha Yokubambisana Isimo sidala "ukubambisana" okuyinkimbinkimbi. Izimbangi ezifana ne-AMD zizoqhudelana kanzima no-Nvidia ukuze bathengise i-MI300X yabo noma ama-chips e-AI azayo. Kodwa-ke, bangathola nama-chips abo efakwa ngaphakathi kwezinhlelo zokuqopha zikaNvidia. Lokhu kukhombisa amasu ezibuko abonwa kweminye imikhakha yezobuchwepheshe, lapho izinkampani ziqhudelana ngomkhiqizo owodwa ngenkathi zisebenzisana epulatifomu eyenza konke kusebenze. Kubasebenzisi bokugcina, inzuzo yokugcina kufanele ibe ukusungulwa okusheshisiwe kanye nekhompuyutha ye-AI enamandla kakhulu, esebenza kahle. Njengoba nje i-SEO ye-Instagram esebenzayo idinga ukuqonda ama-algorithms weplatifomu ukuze itholakale, ukuphumelela engqalasizinda ye-AI kuzodinga ukuqonda ukuthi ungayilungiselela kanjani i-paradigm entsha yenethiwekhi ye-Nvidia, kungakhathalekile ukwakheka kwe-chip eyisisekelo.
Isiphetho: I-Masterstroke ku-Ecosystem Control Irekhi yeseva entsha kaNvidia iyi-masterstroke ekumiseni kwamasu. Esikhundleni sokulwa nempi yokuzivikela ngokumelene nawo wonke umuntu omusha ongenayo i-chip AI, inkampani yakha umgwaqo omkhulu obalulekile okufanele uhambe kuwo wonke amathrafikhi e-AI. Lokhu kuqinisekisa ukubaluleka kwayo kanye nokukhula kwemali engenayo ngisho nasemakethe ye-silicon ehlukene. Impumelelo yalelisu izoncika ekwamukelweni kwemboni okwandile nokuthi izimbangi ziyayemukela yini inkundla ye-Nvidia njengezinga elingathathi hlangothi. Into eyodwa icacile: impi yokuphakama kwe-AI ayisenandaba nje yokuthi ubani one-chip esheshayo, kodwa olawula indwangu yokuxhuma yesikhungo sedatha. Kumabhizinisi abheke ukuthuthukisa i-AI, ukuhlala phambili kusho ukuqonda lezi zinguquko zengqalasizinda. Ukuze uthole imininingwane eyengeziwe mayelana nokuzulazula kwezindawo zobuchwepheshe ezincintisanayo, hlola ukuhlaziya kwethu ekutholeni ukubonakala ezinkundleni ezithuthukayo. Ulungele ukuthuthukisa isu lakho ledijithali? Vumela i-Seemless ikusize wakhe icebo lobufakazi besikhathi esizayo. Xhumana nethimba lethu namuhla ukuze uthole ukubonisana.
