Takaddunin AI sun kashe kafin zamanin sauyin yin sauyin gaba: Me ta hanyar sauyin gaba na yaki mai ƙarfi da sauyin gaba ba tare mai yaki ba sun zama maimakon

Cibiyoyin Bayanai na AI Suna Shigar da Sabuwar Zamanin sanyaya: Me yasa sanyaya ruwa mai zafi da sanyaya bushe yana zama mahimmanci

Saurin ci gaban ilimin kere kere yana sake fasalin yadda ake sanyaya cibiyoyin bayanai. Yayin da nauyin aikin AI ke ƙaruwa, ƙarfin zafi da sabbin sabbin sabbin sabobin ke samarwa yana kaiwa matakan da tsarin sanyaya iska na gargajiya da tsarin tururi ba za su iya sarrafawa ba.

Wannan canjin ya zama tabbatacce a tsakiyar 2026, lokacin da NVIDIA ta buga ƙirar kwatancen Rubin-generation tsarin sanyaya ruwa mai cikakken ruwa wanda aka gina a kusa da madafan glycol masu zafin jiki masu haɗuwa tare da sanyaya bushe na waje maimakon ruwan sanyi ko hasken sanyaya hasumiya. Yana daya daga cikin alamun da suka fi bayyanawa cewa masu sanyaya bushe suna motsawa daga matsayin tallafi zuwa ainihin ƙin ƙin zafi na cibiyar bayanan AI.

Ƙarfin Kwamfuta na AI Ya Ƙirƙiri Sabuwar Kalubalen Cooling

Babban kalubalen cibiyoyin bayanan AI ba kawai ƙarfin sarrafawa bane yana sarrafa zafi.

Kamar yadda GPU performance ya tattara, kuma yadda rack power density ya tattara da kyau. Yadda single‑GPU thermal design power ta tattara zuwa 1,800–2,300W a cikin madaidaici na farko, kuma yadda peak rack densities a AI clusters zasu tattara wajen kuma daga cikin wadanda data centers na farko suka raba. data centers na farko ba su da abu ne da suka raba.

Air cooling na farko ta gaskiya ba ta iya tattara da wannan canji saboda:

• Air yana da abu ne da yawa a cikin heat transfer capacity kamar yadda liquid
• Yadda airflow ya fi yawa ya zama da amfani da energy da fans
• Abubuwan da suka yi amfani da air a cikin wani lokaci ya bukata wani space mai yawa
• Yadda temperature uniformity ya zama da matsala a cikin high densities

Don halin wannan matsala, madaidaici na farko na AI suna canza zuwa direct liquid cooling, inda heat ta kawo daga chips ta go zuwa liquid coolant loop baya don kawo wajen gida.

High‑Temperature Glycol Cooling Ya Zama Wani Madaidaici na Data Center Architecture

Wani daga cikin mahimmanci na advanced AI cooling shine amfani da high‑temperature closed‑loop glycol systems.

A cikin rarraba ta amfani da mazanin gilakol ko kuma tashin yadda na kariya, rarrabbar NVIDIA na Rubin ya sa gilakol ya tsaya a yadda mai ƙarfi da yawa wajen yadda na kariya na gilakol — gilakol ya shiga daga kofin kofin a yadda ta 45°C kuma ya fito a yadda ta 55°C, wanda ke nufin yadda mai ƙarfi da yawa wajen yadda na kariya wanda ya zama daidai a cikin yadda na gida, ta hanyar kofin kofin na gida, kuma zai iya sanya yadda a cikin kofin kofin ba tare da amfani da mazanin kariya mai ƙarfi ba a cikin karamin yadda.

Tashin rarraba: AI server rack zuwa heat exchanger zuwa outdoor dry cooler a cikin sistema na glycol na kofin kofin

Kasuwancin fahimtar yadda ke daidai: kariya ya zama daidai a cikin 40% na karamin yadda na data center , kuma yadda na kariya mai ƙarfi da yawa a cikin kofin kofin na kariya ta ƙara yadda ta ƙara da ƙarshen yadda ta ƙara da 4%. Amfani da kofin kofin mai ƙarfi da yawa kuma yana nufin yadda na gilakol mai ƙarfi da yawa, yana nufin yadda mai ƙarfi da yawa wajen kofin kofin na kariya ba tare da kariya mai ƙarfi ba.

Karamin Kulluwa ta Maiman Zin Daga Wani Hanyar

Babu kuma kamar tura na yankin gudunmawa, sistan na glycol na ƙasa-baƙi ba su da muhimmiyar kulluwa ta maiman zin daga wani hanyar. NVIDIA ya bincike cewa tsarin data center na Rubin zai iya kaddamarwa maiman zin daga wani hanyar kamar 2.6 million gallons of water per megawatt annually ta hanyar kuskure na gudunmawa — wanda ya cire maiman zin daga wani hanyar a matsayin abin da ke haifar da kuskure a cikin saitun da aka zaune. Mu kawo shari’in kulluwa ta maiman zin daga wani hanyar a cikin shirinmu na dry coolers and data center water consumption.

Wannan ya sa sistan na glycol na ƙasa-baƙi suka zama mafi kyau don:

• Yankin kura da maiman zin daga wani hanyar da ke kusa
• Yankin kusurta da kuskure na bayanin al'adu na gudunmawa
• Saitun AI na data center a wajen kafin da maiman zin daga wani hanyar ke zama abin da ke haifar da kuskure a cikin saitun da aka zaune

Dry Coolers Sun Zama Abin Da Ke Cikin Gudunmawa Na Bayanin Tukun

Kamar AI ya zama mai yawa, coolers na suka ƙare kuma suna zama babu wani heat exchanger na farko — suna zama babu wani hanyar muhimmi na cooling architecture dukkan, ba wani component na musamman da aka saba shi bayan sa.

V‑Type Modular Dry Coolers

Don AI applications na yawa, V‑type modular dry coolers suna da fahimtar gaba daya: suna damuwa a cikin surface area na heat exchange don hana amfani da kariya, suna ba da high cooling capacity da installation na musamman, suna gyara yadda ake bukatar waje, kuma suna ba da izini don gyara capacity ta hanyar modular configuration yayin da compute demand ya zama mai yawa.

Matsayin system na traditional cooling tower da system na next‑generation closed‑loop glycol da dry cooler, da heat rejection flow diagram

An rufe shi don Extreme Climate Conditions

Cibiyoyin bayanan AI na gaba ba za su kasance a cikin cibiyoyin fasaha na gargajiya kawai ba ana haɓaka manyan cibiyoyin AI da yawa a yankuna masu ƙalubalen yanayi, don haka tsarin sanyaya yana buƙatar tallafawa kewayon zafin jiki na aiki.

Tsarin sanyaya na zamani suliɗa an tsara su don aiki mai dorewa daga kimanin 20 °C (tare da kariya mai kyau daga daskarewa) har zuwa kusan 40 °C na yanayi, tare da dogaro na dogon lokaci a cikin yanayin waje mai tsanani. Ga yankunan bakin teku da masana'antu, kariyar lalata ruwa kamar murfin matakin C5M an ƙayyade shi sosai don kayan aiki da aka fallasa zuwa gishiri da iska mai tsanani duba maganin karewa na bushewa mai sanyi don yanayin haɗari mai yawa don zaɓin kayan aiki a cikin waɗannan muhallin.

Tsarin sanyaya na zamani zai tallafawa fadada cibiyoyin samar da AI a nan gaba

Ci gaban AI yana tafiya da sauri fiye da tsarin tsara cibiyar bayanai na gargajiya, don haka tsarin sanyaya yana buƙatar zama mai sassauƙa don faɗaɗa tare da ƙaruwar buƙatar ƙididdiga.

Sistemu na dry cooler na modulari suna ba da wani wannan ƙarfin. Bayan amfani da wani abu mai yawa na cooling, masu amfani zasu iya saita kuma adda capacity ta hanyar yin kudin da kudin bayan AI workloads suka ziyarce — wanda ya ba da ƙarfin yin transport kuma installation, yin project deployment a lokaci mafi kyau, yin kara risk na maintenance (saboda wani module zai iya aikata kafin ina shut down wani system daga kowane), kuma cooling capacity na scalable wanda ya tsaya da compute growth ba tare da yin forecast a lokacin da suka gabata a shekara manya.

Alama na suka run da dry coolers a wannan scale suna kara su da continuous condition monitoring — our guide on smart dry coolers and IoT remote monitoring yana bayyana yadda IoT sensors suka gane fan, coil, ko airflow issues kafin su iya fuskanta uptime.

Masu Sabon Gaskiya

Shin high‑temperature liquid cooling yana nufin cewa dry coolers suna canza mechanical chillers daga kowane?

Ba kusan, amma wani maqasun shine in gudanen yadda za a yi amfani da su. Saboda alama na Rubin ta hana coolant mai ƙarfi kafin wani abu na gaba daya na chilled-water, dry coolers zasu iya gudanen wannan ƙarfi don samun hawaye a cikin ƙarfi mai yawa na gaba daya na climates da musamman — kuma mechanical cooling zai kasance kamar tushen bayanin lokacin da aka fi kara ƙarfi kafin wani abu na gaba daya ba ne kamar shirin da ke gudanen ƙarfi.

Shin za a iya saita dry coolers a cikin data centers na gaba daya don high-temperature liquid cooling, ko ya bukatar rarraba ta farko?

Kowane irin zai iya aiki, amma yadda zai iya aiki yana taka kafin infrastructure na gaba daya na cooling. Wadanda suka yi amfani da liquid cooling loop da heat exchanger da outdoor heat rejection equipment suna da hanyar da ke yawa; wadanda suka yi amfani da air cooling ko evaporative towers kuma suna bukatar rarraba mai yawa na infrastructure.

Yadda high-temperature glycol cooling zai iya gyara sizing na dry cooler don standard applications?

Yadda da ayyukan kuma yadda da ƙarfi na gaba daya (wani farko a tsakiya na yadda da ƙarfi na gaba daya da yadda da ƙarfi na hawaye) ya sa aiki da kariya mai sauƙi don samun kariya ta hanyar dry cooler, saboda farko na yadda da ƙarfi ya kashe. Amma, kuma yadda da ƙarfi na gaba daya zai bukatar fahimtar kuma yadda da glycol, wanda ya dace da yadda da ƙarfi na gaba daya — duba mu gabatarin kariya na dry cooler don bayanin hanyar aiki.

Zaman Lissafi na AI Ya Kashe Kuma Yadda da Kariya Mai Sauƙi

Kullum kuma yadda da AI na zamani zai bukatar kuma yadda da ƙarfi mai sauƙi — zai bukatar kuma yadda da kariya mai sauƙi. Kariya mai sauƙi na glycol na yadda da ƙarfi mai sauƙi ta hanyar V-type modular dry coolers yana nufin kariya mai sauƙi ga kariya na AI na zamani, kuma NVIDIA's Rubin reference design ya sa kariya ta zuwa a cikin kariya na zamani.

SINRUI Radiator yana ƙirƙirawa da ayyukan kuma bayanin taimaka na kula da hankali wanda ke cikin gaba daya, domin suka yi amfani da su, kamar dry coolers, liquid cooling heat exchangers, da systems na radiator na kasuwanci. A lokacin da AI computing ya zama mai yawa, kula da hankali daidai zai zama babbar abu ne wanda ke shirya daga cikin abubuwan da suka shirya mafi kyawun kariya, kuma kuma kariya na data centers na gaba.

Gaba daya na AI ba ta hanyar chip-pi mafi kyau kuma — amma ta hanyar kula da hankali mafi kyau.