According to industry sources, NVIDIA’s next-generation GPU products will be built using TSMC’s 5nm process, and the release date will be about later in 2022, and the price will be higher.
Last week, TSMC announced an undifferentiated price increase for foundry. Among them, mature processes above 16nm/12nm are 15 to 20%, and advanced processes such as 5nm/7nm are 7% to 10%. If NVIDIA chooses TSMC’s foundry graphics card, There is a high probability of being affected.
Public information shows that NVIDIA 5nm GPUs may be divided into RTX 40 series game graphics cards, code-named Ada Lovelace, and professional users-oriented graphics cards, code-named Hopper. Both code-names take the names of female scientists.
Among them, AD102 has a core area of 600mm², 144 sets of SM units (18342 CUDA), a core frequency of 2.2GHz, matching 384bit GDDR6X video memory, and the power consumption of the entire card is 400 to 450W.
News/Image Source: cnBeta
Further reading: