Originally Posted by Nagorak
DLSS was a classic example of over(software)engineering, IMO. It was an interesting idea that might have turned out to have merit. But instead it turned out to not doing anything more than a standard dumb filter could do, all the while being 1000 times more complicated. It's possible to come up with very intricate ways of solving simple problems, and that's what Nvidia managed to do with DLSS. The result simply isn't that great, but meanwhile it's very impractical to implement since it requires machine learning.
I also suspect it will quietly be dropped in the next year or so, and we just won't hear anything more about it. So, I certainly wouldn't buy a 2000 series card based on DLSS.
Nobody should buy a 2000 series card based on DLSS. But when it's implemented properly (Metro Exodus), its superior to any sharpening filter. Because it isn't a sharpening filter. A sharpening filter won't magically add in texture detail because all it's doing is sharpening. The idea behind super resolution image processing is to take a lower resolution image and produce a near identical higher resolution image. This isn't new, this kind of image processing existed for years (medical, satellite, photography), requiring machine learning as well. Nvidia calls their version DLSS, they're the first to bring it to gaming. Metro Exodus does this well and even Hardware Unboxed, who love to crap on Nvidia, had to admit the DLSS image looked better vs RIS but then went on to complain about DLSS performance (which doesn't matter with a 2080Ti).
Lastly, If you really want to discuss this further there's another thread for that in the other graphics card section.