What DLSS/FSR Upscaling Mode do you use?

[AusWolf]

Also agree.

For me it's:
1. Set maximum graphics at native res. If it runs well, great.
2. If it doesn't run well, decrease graphics options to find a balance between looks and performance.
3. If it still isn't good, set everything to low.
4. If still no results, turn on FSR. Then, see whether low graphics, or FSR, or a combination of both gives the (best) least bad results.

Obviously, if I'm at step 4, it's probably time for a GPU upgrade soon.

 

[wheresmycar]

For me it's:
1. Set maximum graphics at native res. If it runs well, great.
2. If it doesn't run well, decrease graphics options to find a balance between looks and performance.
3. If it still isn't good, set everything to low.
4. If still no results, turn on FSR. Then, see whether low graphics, or FSR, or a combination of both gives the (best) least bad results.

Obviously, if I'm at step 4, it's probably time for a GPU upgrade soon.

This is where I am too - except No. 3. (me, fo-sho, no go low)

Its down to individual preferences and the 2 games i regularly play at 1440p just run better at native. It really is as simple as that. Perhaps other titles are better optimised with DLSS/FSR where upscaling deserves merit (esp. 4K). When supported, I always test the available features in the games I play but ultimately, the IQ results speak for themselves. RT is too taxing for my 3080. Upscaling is a hit and miss. A 'hit' suggests if native performance is unpleasantly tanked, to the extent i'm pushed to settling with noticeably poor quality lower settings, I would absolutely put DLSS to work and then when the time is right, UPGRADE my GPU.

Fingers crossed - moving forward, image reconstruction optimizations will hopefully yield better results at lower resolutions. I’m an optimist and theres plenty of room to keep experimenting, learning and when implemented more effectively (1080p>1440p), i'm diving right in!

The curious cat: Do you guys sit at an arms length from your display while gaming? I’m using a 27" 1440p panel and I wonder if some of us perceive differences in image quality (DLSS vs. Native) differently based on factors like pixel density, viewing distance, screen size, personal sensitivities to artefacts/whatnot and even the type of game being played (eg. fast-paced action 'vs' slower detail oriented titles)?

 

[lexluthermiester]

The curious cat: Do you guys sit at an arms length from your display while gaming?

Yes. My screens are both at finger tip distance. I was using 1440p displays but went back to my 1080p screens due to difficulties with scaling on some programs.

I wonder if some of us perceive differences in image quality (DLSS vs. Native) differently based on factors like pixel density, viewing distance, screen size, personal sensitivities to artefacts/whatnot and even the type of game being played (eg. fast-paced action 'vs' slower detail oriented titles)?

There's very likely some of that going on. Settings are a very personal and subjective things.

 

[wolf]

Utter Moose Muffins

Excellent counter point, I never thought of it like that.

Because I expect excellence

No AA ain't it mate. But hey, you do you

 

[JustBenching]

People prefer setting everything to low - and decreasing the resolution - before using upscaling. This is not a conversation based on facts.

Its a mix bag. If we're going by still images, if possible i'd tweak the settings to find the middle ground. Top image looks smoother but the plants/veg look a little mushy/blurred. Bottom image carries minor artefacts and the plants/veg look over-sharpened. Both to my distaste.

I got an idea... if you're willing to aid and assist. How about 2 short video samples using 1440p. Create a new thread with a poll. This could be valuable for those unfamiliar with DLSS and might even spark their interest. Starting a new thread will also help attract a broader audience to pitch in.

Problem with the video is youtube will compress the details. I don't mind doing it but I think it's useless, I've already posted some comparisons and nobody could tell what is what. So it's evident that DLSS = Native in a blind test. Only when you know which is which people seem to prefer native. In medicine we call that placebo.

PS1. The above screenshots were DLSS Q vs NATIVE MSAA 4x (!!!). The performance gap was vast, for what? The native one doesn't even look universally better and it costs 40% performance.

 

[AusWolf]

The curious cat: Do you guys sit at an arms length from your display while gaming? I’m using a 27" 1440p panel and I wonder if some of us perceive differences in image quality (DLSS vs. Native) differently based on factors like pixel density, viewing distance, screen size, personal sensitivities to artefacts/whatnot and even the type of game being played (eg. fast-paced action 'vs' slower detail oriented titles)?

Yes, about arms length from a 1440 UW 34". I'm a slow gamer, always looking for story and atmosphere in a game. 50-60 FPS and I'm good.

This is not a conversation based on facts.

No, it's not, and it never has been. It's a conversation based on personal preferences - hence the poll.

 

[JustBenching]

No, it's not, and it never has been. It's a conversation based on personal preferences - hence the poll.

You think it's impossible for image quality to be "measured"? Sure some people will disagree with the measurements and still think that X looks better than Y etc. but im sure there is an objective way of doing it. Like, objectively, 8k looks better than 720p, right?

 

[Raffles]

I go with DLSS Quality wherever possible. In current versions of DLSS, it looks incredible, and offers a huge boost in performance with no drawbacks.

 

[AusWolf]

You think it's impossible for image quality to be "measured"? Sure some people will disagree with the measurements and still think that X looks better than Y etc. but im sure there is an objective way of doing it. Like, objectively, 8k looks better than 720p, right?

You can measure pixels, but you can't measure what your eyes find more pleasing.

 

[JustBenching]

You can measure pixels, but you can't measure what your eyes find more pleasing.

But you are only using the subjective argument when it fits your argument. For example yesterday you said you don't understand why people buy Nvidia for dlss since it's of questionable use. Well if it's all subjective then it's obvious why people prefer dlss, cause it looks better to them right?

You cant argue that image quality is subjective and then don't get why people prefer dlss. The 2 points go against each other.

 

[AusWolf]

But you are only using the subjective argument when it fits your argument. For example yesterday you said you don't understand why people buy Nvidia for dlss since it's of questionable use. Well if it's all subjective then it's obvious why people prefer dlss, cause it looks better to them right?

You cant argue that image quality is subjective and then don't get why people prefer dlss. The 2 points go against each other.

Well, if there are people who think a soft, fuzzy, blurry image looks better than a sharp, crisp one, then I guess you have a point, however weird I think that is.

 

[JustBenching]

Well, if there are people who think a soft, fuzzy, blurry image looks better than a sharp, crisp one, then I guess you have a point, however weird I think that is.

Now you are back to claiming it's not subjective.

If it's soft, fuzzy and blurry then obviously you'd tell them apart in a blind test. Do it and I'll absolutely change my mind, I'll throw dlss under the bus.

 

[AusWolf]

Now you are back to claiming it's not subjective.

If it's soft, fuzzy and blurry then obviously you'd tell them apart in a blind test. Do it and I'll absolutely change my mind, I'll throw dlss under the bus.

I'm not trying to change your mind. All I'm saying is that I see no reason to prefer it over native if you don't count the extra performance. If you do, be my guest.

 

[JustBenching]

I'm not trying to change your mind. All I'm saying is that I see no reason to prefer it over native if you don't count the extra performance. If you do, be my guest.

I am trying to change my mind. I mean, if I'm wrong and native looks better, I'd like to know so I'll stop promoting dlss like an idiot. I dont want to give people the wrong suggestions.

And again, it's not about performance. You can target the same performance between dlss and native. The question is which one looks better, since framerate will be similar

 

[AusWolf]

I am trying to change my mind. I mean, if I'm wrong and native looks better, I'd like to know so I'll stop promoting dlss like an idiot. I dont want to give people the wrong suggestions.

In my opinion, native always looks better because AI can never guess the missing pixels with 100% accuracy. How much better, it depends on the game and its implementation.

Sure, it's only an opinion, but I'm not alone with it, which proves that DLSS isn't absolutely better for everyone. If it was, no one would have voted for native in the poll.

Personally, I don't like promoting anything that only I consider better for myself. I'm not a paid advertiser. I only promote things that might be beneficial for the person I'm talking to.

And again, it's not about performance. You can target the same performance between dlss and native. The question is which one looks better, since framerate will be similar

How would it be similar? DLSS always gives you more performance at the cost of image quality. That's what it's for.

 

[AVATARAT]

I am trying to change my mind. I mean, if I'm wrong and native looks better, I'd like to know so I'll stop promoting dlss like an idiot. I dont want to give people the wrong suggestions.

And again, it's not about performance. You can target the same performance between dlss and native. The question is which one looks better, since framerate will be similar

In some games DLSS can give a quality very close to native, in others it's just crap and it depends on its implementation in the game.
The main goal of upscalers is to provide quality as close to native as possible, since when upscaling they have to fill in missing information.

As an example, you can try Throne and Liberty (F2P), which has been a title for a few months, and check how "well" the DLSS work (and this is probably normal for MMORPGs).

 

[JustBenching]

In my opinion, native always looks better because AI can never guess the missing pixels with 100% accuracy. How much better, it depends on the game and its implementation.

Sure, it's only an opinion, but I'm not alone with it, which proves that DLSS isn't absolutely better for everyone. If it was, no one would have voted for native in the poll.

Personally, I don't like promoting anything that only I consider better for myself. I'm not a paid advertiser. I only promote things that might be beneficial for the person I'm talking to.


How would it be similar? DLSS always gives you more performance at the cost of image quality. That's what it's for.

There are options like dlaa and supersampling that land you at similar performance to native (dldsr + dlss for example). Or as I've repeatedly said before, choosing a 4k monitor to use with dlss or a 1440p monitor to play natively. Which one gets me better image since performance is the same.

Your opinion is by definition wrong since we have examples like image restoration, where the original native picture is destroyed in parts and the ai algorithm guesses. An algorithm trained by looking at billions of faces, how is it at all unlikely that it can guess correctly what a tiny tiny spot on a face should look like? Native is an approximation anyways since it's trying to approximate a face for example with a limited amount of pixels.

In some games DLSS can give a quality very close to native, in others it's just crap and it depends on its implementation in the game.
The main goal of upscalers is to provide quality as close to native as possible, since when upscaling they have to fill in missing information.

As an example, you can try Throne and Liberty (F2P), which has been a title for a few months, and check how "well" the DLSS work (and this is probably normal for MMORPGs).

Surely there are games that dlss looks like crap (rdr2 too), but there are as many that native TAA looks crap too.

 

[AVATARAT]

Surely there are games that dlss looks like crap (rdr2 too), but there are as many that native TAA looks crap too.

TAA is in the same situation - if implemented badly, it will look bad.
And yes, promoting something that works from time to time, some of the time - as a must have thing is not ok.

 

[AusWolf]

There are options like dlaa and supersampling that land you at similar performance to native (dldsr + dlss for example).

But that's not native or simply "DLSS Q" anymore, is it?

Or as I've repeatedly said before, choosing a 4k monitor to use with dlss or a 1440p monitor to play natively. Which one gets me better image since performance is the same.

Or as I've said before: can my PC run a 4K monitor in my games natively? Or do I want to spend money to upgrade it? If the answer is no, I won't buy a 4K monitor.

Your opinion is by definition wrong since we have examples like image restoration, where the original native picture is destroyed in parts and the ai algorithm guesses. An algorithm trained by looking at billions of faces, how is it at all unlikely that it can guess correctly what a tiny tiny spot on a face should look like? Native is an approximation anyways since it's trying to approximate a face for example with a limited amount of pixels.

Can you show me how you can restore an image (every image) at 100% accuracy? Until you can, I won't believe it.

Surely there are games that dlss looks like crap (rdr2 too), but there are as many that native TAA looks crap too.

That's more like an argument against both DLSS and TAA rather than an argument for either.

 

[Onasi]

But that's not native or simply "DLSS Q" anymore, is it?

DLAA is, indeed, just applying the AA algorithm of DLSS to the native frame. So it’s no different in practice to any other AA method at native. In fact, out of all temporal AA methods it’s arguably the best one to use.

 

[JustBenching]

But that's not native or simply "DLSS Q" anymore, is it?

Does it matter? Dlss is still part of the process.

Or as I've said before: can my PC run a 4K monitor in my games natively? Or do I want to spend money to upgrade it? If the answer is no, I won't buy a 4K monitor.

That's irrelevant. Nobody said anything about upgrading. People buy pcs and monitors. Or even better, laptops. So the question of should you go with a lower res monitor or a higher res and use dlss is very. The only way to answer that is to... well test it.

Can you show me how you can restore an image (every image) at 100% accuracy? Until you can, I won't believe it.

You can just search for image restoration man.

 

[AusWolf]

DLAA is, indeed, just applying the AA algorithm of DLSS to the native frame. So it’s no different in practice to any other AA method at native. In fact, out of all temporal AA methods it’s arguably the best one to use.

Does it matter? Dlss is still part of the process.

It does - you can't say how brilliant DLSS looks if you're looking at a DLAA augmented image. The praise goes to DLAA, not DLSS.

That's irrelevant. Nobody said anything about upgrading. People buy pcs and monitors. Or even better, laptops. So the question of should you go with a lower res monitor or a higher res and use dlss is very. The only way to answer that is to... well test it.

I've tested it (albeit back on my 2070 before it died) and I think you know my answer.

You can just search for image restoration man.

I just did... I see what you mean now. Well, that seems like a cool technology because it restores part of the picture that are not there.

DLSS, however, guesses parts of your image that could be there if you were playing at your monitors's native res. Apples to oranges.

 

[JustBenching]

DLSS, however, guesses parts of your image that could be there if you were playing at your monitors's native res. Apples to oranges.

No, not really. We have no idea what training algo dlss is using. Most likely, a much higher resolution that whatever you are trying to run.

But as I've said, blind test. Since you've tried it on your pc and you obviously could tell a difference, what would stop you from telling the difference in a blind test?

We don't need to spam the thread anymore, say yes you'll do it and we open a thread to do so, or no I won't do it and I'll stop asking.

 

[AusWolf]

No, not really. We have no idea what training algo dlss is using. Most likely, a much higher resolution that whatever you are trying to run.

But as I've said, blind test. Since you've tried it on your pc and you obviously could tell a difference, what would stop you from telling the difference in a blind test?

We don't need to spam the thread anymore, say yes you'll do it and we open a thread to do so, or no I won't do it and I'll stop asking.

Sure, do it, but no DLAA magic, just pure no-AA native vs DLSS.

But like you, I'm also afraid that we won't see much because of Youtube's compression.

 

[lexluthermiester]

Excellent counter point, I never thought of it like that.

Well, if you're going to shovel it, I'm going to call it out. A native rendered image is always the the highest quality one can achieve. This is not subjective.

No AA ain't it mate. But hey, you do you

That's YOUR opinion and exactly MY point. Personal preference is subjective.