Problem with the unified approach is that a lot of die space may be used for architecture which could be useless for gaming and unnecessarily increases the cost of each chip that is destined for gaming.
No, if you read the article they explain their thinking, currently Nvidia use a unified architecture, so the gaming GPU's have the same capabilities as the workstation GPU's and with that the they can utilise the full feature set of CUDA, that's how you get DLSS and Ray Reconstruction ecte...
For a couple of year now AMD have been building their equivalent or rival to CUDA, ROCm, now the want to unify the architecture for the same reason Nvidia already do, this is how AMD get to DLSS and Ray Reconstruction equivalents.
Last edited: