davis.anthony
Veteran
I think FSR2 and DLSS2 are proving that resolution is a waste of rendering power these days.
Unreal Engine 5, [UE5 Developer Availability 2022-04-05]
- Thread starter mpg1
- Start date
I think FSR2 and DLSS2 are proving that resolution is a waste of rendering power these days.
Silent_Buddha
Legend
Except that while they are good enough for some people they are also not good enough for some people. Even on this forum there are people that love it, there are ones that find it situationally decent and there are ones that find them in general to be worse than not having them on in many cases.
Temporal reconstruction is needed, yes, and is likely the future, but there's still a long way to go before it is universally better than "native" resolution.
Of course, if all anyone looks at are screenshots then DLSS, FSR and most temporal reconstruction techniques will yield pretty darn good results. However, for me and many others, temporal reconstruction artifacts and instability are still more distracting than any performance benefits they might bring to the table when the game is in motion.
Again, keep in mind that I'm not saying it's universally bad either. Plenty of people find it completely acceptable or even an improvement. However, while it isn't universally bad it's also not universally good. Thus depending on the person it's either almost always better or almost always worse with a gamut of people in between who will feel it's better or worse depending on the game and implementation.
Regards,
SB
Inuhanyou
Veteran
Regardless of if one thinks DLSS and FSR2 have flaws or not, saying arbitrary resolutions are worth much on fixed hardware or lower end pcs is silly. Especially the idea that if something is not native 4k is not "next gen".
Resolution means almost nothing at this point as long as it's suitably high enough to be reconstructed, and we've seen even series S to down in res pretty far but get reconstructed very well with TSR, to say nothing of the higher end consoles reconstructing from a higher res by default
Resolution means almost nothing at this point as long as it's suitably high enough to be reconstructed, and we've seen even series S to down in res pretty far but get reconstructed very well with TSR, to say nothing of the higher end consoles reconstructing from a higher res by default
davis.anthony
Veteran
Upscaling will allow a set piece of hardware to reach a level of graphical output that native will never match.
Gamers will always chose Half Life 2 graphics running at 4k via DLSS over Half Life 1 graphics running at native 4k.
FSR 2.0 is comfortably in the higher end of console upscaling solutions.
Inuhanyou
Veteran
No. FSR2 implementation atleast on console so far has been pretty good. Atleast outside of forspoken.
Dead space, cyberpunk, Witcher all look much better with it than without, by a notable degree.
It's second only to insomniacs custom temporal injection(well maybe third if we count TSR. That gives great results, but we only have one game so far to fully test it)
I don't think we'll ever hit a point where rendering is "solved" for real-time, so there will always be a use for spacial and temporal upscaling to allow for more complex or more intensive algorithms. Maybe the consoles should offer three modes: TSR@60 fps (ultra/high), TSR@120 fps (medium) and 4k native @ 30 fps (medium). I have a feeling the latter one would barely get picked.
Many argue the importance of RT performance, however i dunno if the choice was there, ai/ml or RT in these consoles to be more performant or accelerated i think a xess or dlss would actually be more of a welcoming thing. In special regarding the weaker cpu’s limits to be overcome.
Is it though? On PC where the two solutions can be compared head to head in Spiderman, the general consensus is that FSR comes out on top.
cheapchips
Veteran
Is that a consensus?
I thought Insomniac's solution looked better than FSR2, even if it's a little softer. Which probably is what would put off the typical PC crowd...
FSR 2 adds extremely obvious artifacts in way too many areas. I'd much rather have a slightly softer image than one full of speckling, fizzling and other artifacts around every single aspect of the image that exhibits any movement. The ghosting is also tremendous and much worse than good TAA implementations.
I remember reading a hand ful of different comparisons at the time including Alex's from Digital Foundry which all concluded FSR produced a better image.
Even NXG agreed but naturally he put it down to the PC implementation of TAUU being inferior to the PS5 one (seemingly while using different base resolutions).
Inuhanyou
Veteran
Well either way. Console has good solutions that don't require the need for native res or a raw image. We are heading into a situation where TAA isn't just AA anymore but actually resolving the image in a greater way and that's awesome 
DavidGraham
Veteran
I think that was way in the beginning, Nixxes improved their DLSS implementation considerably after that, and added DLSS3 which was also excellent in Miles Morales.
Well that's exactly right, isn't it? Which means that 1440p is a 'target' resolution of the past, and was hardly ever a thing in the first place really given there are no such thing as 1440p TV's. No reason not to target 4k nowadays in the days of reconstruction.
In terms of this argument over whether FSR2 is good enough - reconstruction techniques will get better. It will be a prime point of development going forward. Not that FSR2 even as of now is really some massive detriment unless you're a big pixel peeper type. Consoles have always lived in the realm of 'good enough'.
arandomguy
Veteran
The additional consideration I'd wonder what the test conditions and criteria are? Something in general I've noticed with a lot of these comparisons and solutions is that often have relative weaknesses and strengths, that may not be weighted the same depending on usage conditions.
Typically how comparisons are done of the PC tends to be via monitors and often with basically pixel peeping at close range, including via static shots vs in motion. Would the same hold up if tested from living room seating differences on TVs and say in motion?
Obviously resource intensive and so unlikely to be ever done but I'd be interested in an actual blind test of all these various solutions under different usage conditions and what the results would be.
Indeed - the OG version was still WIP and changed considerably post-release (aka check their patch notes). Heck, it changed even after just now as 2.5.1 releases hah
But I think you would have to be a bit ??? to think the IGTI looks superior to DLSS or even FSR 2 in nearly all cases.
Turn down all sharpening to 0 for everything and do the comps in motion and in stills.
Have someone even attempted to compare PS5 custom IGTI to DLSS or FRS2 (or even IGTI PC)? That should be quite easy to do using Spider-man. I still think PS5 IGTI is way better than FRS2 which has plenty of problems last time I looked.
By the way nobody should be surprised to learn PS5 IGTI is custom as it's only logical on a console.
Similar threads
