Yeah. So its not like they let the die shrink save costs by having smaller die sizes on say 7nm. So had to do something or have much lower margins.The 16nm and 12nm nodes cost basically the same. These were the nodes that Pascal and Turing were on respectively, yet NVIDIA wanted to add around 60% more transistors between the GP102 (1080 Ti) and TU102 (2080 Ti).
Where do you think NVIDIA was going to recoup the cost of making Turing?