Presentation of Nvidia's DLSS 2.0
-
The GPU developer has announced a new version of its artificial intelligence system for graphics sampling. The concept involves rendering images at a smaller size and then using AI to reconstruct it and create a higher definition image. The process is carried out frame by frame.
-
It seems like a much smarter system than the previous version, as it doesn't need to train for each game and each resolution. It requires motion vector information every frame from the game engine, so it needs the developer to facilitate it, but most modern games already use that information for temporal antialiasing systems, so it should be much more common than the previous version of DLSS.
It's a shame that only the 2xxx series will benefit from it.
-
I bet it has more of the sharpening filter than of refinement.
