Ngesikhathi sengqungquthela ye-GTC 2019, isikhulu se-NVIDIA uJen-Hsun Huang utshele abezindaba ukuthi inkampani izofaka ama-oda amaningi esizukulwane esilandelayo se-7nm GPUs nge-TSMC, kuyilapho abakwaSamsung bezothola isabelo esincane kakhulu.
Esikhathini esedlule, kwavela amahlebezi okuthi i-Samsung cishe izoba ngumkhiqizi oyinhloko wama-GPU azayo avela kwa-NVIDIA. Kuyasolakala ukuthi, inqubo yakwaSamsung ye-7nm deep ultraviolet lithography (7nm EUV) izosetshenziswa ukukhiqiza isizukulwane esilandelayo sama-NVIDIA GPU. Kepha manje inhloko ye-NVIDIA iwachithile la mahemuhemu.
UJensen Huang uthe inkampani yakhe inobudlelwano obusondelene ne-TSMC, ekhiqize ama-Pascal GPUs ayo adlule angu-16nm futhi manje isikhiqiza i-12nm Volta neTuring yamanje. Ngokushesha, inhloko ye-NVIDIA ayizange yehluleke ukuqaphela isakhiwo se-Turing, esakhiwe kubuchwepheshe benqubo ye-TSMC ye-12-nm futhi, ngokusho kwakhe, inikeza ukusebenza kahle nokusebenza okungcono kunemikhiqizo yokuncintisana ngokusekelwe kubuchwepheshe benqubo ye-7-nm. Kuphinde kwaphawulwa ukuthi ngaphandle kwe-TSMC kanye nobuchwepheshe bayo benqubo ethuthukisiwe, ama-NVIDIA GPUs angeke aphumelele njengoba enza, yingakho ukubambisana ne-TSMC kubaluleke kakhulu ku-NVIDIA.
Kodwa-ke, ngokusho kuka-Huang, i-Samsung isazothola ama-oda avela ku-NVIDIA, kodwa ngevolumu encane kune-TSMC. Njengoba kusanda kwaziwa, i-Samsung izokhiqiza ama-processor amasha , eyenzelwe izimoto ezizimele. Kungenzeka futhi ukuthi, ngaphezu kwalokhu, i-NVIDIA izofaka ama-oda ne-Samsung okukhiqizwa kwamanye ama-chip. Lawa, ngendlela, angaba amanye ama-GPU azayo. Masikhumbule ukuthi emndenini wakwaPascal i-chip encane kunazo zonke i-GP107 yakhiqizwa yi-Samsung, kanti enye ikhiqizwa yi-TSMC.
Ekugcineni, uJensen Huang wabuzwa mayelana nesikhathi sokwethulwa kwe-NVIDIA yesizukulwane esilandelayo se-7nm GPUs, kodwa uvele waphendula ngokuthi manje akusona isikhathi sabo noma sokuveza noma yiluphi usuku. Kusukela kungxoxo yakamuva ne-NVIDIA CFO, uColette Kress, ukuthi i-NVIDIA ifuna ukumangaza wonke umuntu ngesimemezelo se-7nm GPU, kodwa ilinde isikhathi esifanele salokhu.
Source: 3dnews.ru