DLSS is based on Deep Learning. Super sampling works by making images sharper by rendering them in a higher resolution and then take that more precise information and use it to draw a sharper image. In DLSS you use an AI to learn how a game is supposed to look like in a higher resolution and then apply the resulting algorithm to the picture. This way the algorithm can seemingly create information out of nowhere, thanks to past training. That process is way less demanding on hardware than having to calculate a whole picture in a high resolution. It's so effective that you can basically do the opposite of regular super sampling, which downscales a high resolution to a lower one. DLSS instead uses a lower resolution and then upscales it with additional information from the deep learning algorithm.
That way you can have a better looking picture while still using way less compute power. A technique like that is great for consoles and especially for Nintendo, since they like to underpower their hardware. If the game is built with DLSS in mind it could actually be an amazing innovation and deliver high fps in high resolutions and highly detailed textures on comparably weak machines.
1. So hypothetically, DLSS could make Switch 2 just as good looking as an Xbox Series X, if they could make it efficient enough? Or perhaps by Switch 3?
2. Why would Microsoft and Sony not abandon current methods and embrace DLSS themselves, if this is more efficient?
3. If the other two continue in the more in-efficient methods, then could we see Nintendo catch up to the power levels of the other two in a gen or two, by means of more efficient and cheaper technology?Last edited by Dulfite - on 09 October 2020