The whole point in DLSS is to use the physical Tensor cores to prevent image quality loss from downscaling and upscaling. Because if I wanted performance and didnt care about image quality, I would just use the resolution scale and get the performance back.
Nvidia says it themselves..
1.
2.
3.
(generate pixels it has never seen before, aka fill in the gaps in a 1440p img and make it 4k)
4.
5.
( but in BFV DLSS looks worse than 1440p on a 4k screen ?)
STOP saying stuff like " what did you expect", "DLSS is going to be blurry" etc. AI is like a human (except its a lot faster and can correct stuff in real time when you play the game) that is supposed to fill in the incorrect stuff on the fly and make it not look like a straight up 1440p upscaling.
imagine if you were printing some photos and u find mistakes in the photos, you go back and fix them and then reprint them. But right now DLSS isnt doing that, it just seems to be doing the printing without correcting the images.
But I am disappointed that DLSS so far is not doing what it was hyped up as or what its actually supposed to do..