Support NeoGAF

[BluRayHiDef]

The following are screen captures that I've made using the print-screen function. Every image is 4K but differently so. Some are upscaled to 4K via regular upscaling since that's the native resolution of my display, other are upscaled to 4K via Nvidia's DLSS (Deep Learning Super Sampling), and some are native 4K. The ones that are upscaled to 4K via regular upscaling due to 4K being the native resolution of my display are simply labeled as the resolution from which they've been upscaled since they do not benefit from any enhancements to their quality.

Spoiler

1080p:

1080p to 4K via DLSS:

1440p:

1440p to 4K via DLSS:

Native 4K

_______________________________________________________

1080p:

1080p to 4K via DLSS:

1440p:

1440p to 4K via DLSS:

Native 4K

 

[Dibils2k]

they literally look the same on my phone, this is why we need a 800% zoom version

 

[BluRayHiDef]

they literally look the same on my phone, this is why we need a 800% zoom version

You can see the difference on your phone if you zoom in to an extent such that the portions of the images that are visible are as large as they'd be on a desktop-sized display.

 

[Kuranghi]

they literally look the same on my phone, this is why we need a 800% zoom version

I have a 65" screen thats 8 feet away and I have to display them fullscreen to fully see the differences, its just how you compare images. You show them 1:1 on your screen, you don't need to zoom in.

 

[Dibils2k]

You can see the difference on your phone if you zoom in to an extent such that the portions of the images that are visible are as large as they'd be on a desktop-sized display.

doesnt work since i have to drag the shit around, if we had small zoomed versions of each focusing on certain part, would work great but hey

 

[BluRayHiDef]

I have a 65" screen thats 8 feet away and I have to display them fullscreen to fully see the differences, its just how you compare images. You show them 1:1 on your screen, you don't need to zoom in.

What do you think about the differences? Does DLSS appear to make a difference? In my opinion, the DLSS images are sharper than their native-counterparts (e.g. 1080p to 4K via DLSS looks sharper to me than 1080p). Also, I think that the DLSS images look as good as the native 4K images, even the 1080p ones.

 

[BluRayHiDef]

doesnt work since i have to drag the shit around, if we had small zoomed versions of each focusing on certain part, would work great but hey

When you select an image on NeoGAF and it subsequently becomes full-screen, hit the arrow button in the upper right hand corner to view the image outside of NeoGAF's domain, where it should be easier to zoom on it and move it around.

 

[Kuranghi]

What do you think about the differences? Does DLSS appear to make a difference? In my opinion, the DLSS images are sharper than their native-counterparts (e.g. 1080p to 4K via DLSS looks sharper to me than 1080p). Also, I think that the DLSS images look as good as the native 4K images, even the 1080p ones.

Everytime I say my piece on this people get pissed lol.

I think the 1080p and 1440p DLSS'd to 4K images definitely have a good reduction in aliasing over the native 1080p and 1440p images, and have less arifacts in the hair as well.

As far as the DLSS images looking as good as the native 4K one I can't agree. I think the 1440p DLSS images look similar to 1800p native. So for me its similar to PS4 Pro checkerboarding (analysis I've seen says a 1440p native CB'd game ends up having a similar effective resolution as 1800p) except without as many artifacts so thats definitely an improvement over CB.

The test on the sign is garbled and not as clear on the 1440p DLSS image, compared to the native 4K one. The texture detail on the wall is better in the native 4K image too.

I think maybe a DLSS mode with 1800p as the starting res would look much closer to native 4K. For me, on a 65" 4K TV I can clearly see a noticeable sharpness increase when I go from even 3712x2088 to native 4K. So even though DLSS is better I can still tell its not native straight away.

I am going to wait for a few more versions of it until I care about DLSS tbh. I'd prefer to optimise my settings to keep a locked 30/60 at native 4K than use this.

 

[Kuranghi]

doesnt work since i have to drag the shit around, if we had small zoomed versions of each focusing on certain part, would work great but hey

Open the images in their own tabs separately, make them full-size, then swipe between your tabs to compare.

 

[BluRayHiDef]

Everytime I say my piece on this people get pissed lol.

I think the 1080p and 1440p DLSS'd to 4K images definitely have a good reduction in aliasing over the native 1080p and 1440p images, and have less arifacts in the hair as well.

As far as the DLSS images looking as good as the native 4K one I can't agree. I think the 1440p DLSS images look similar to 1800p native. So for me its similar to PS4 Pro checkerboarding (analysis I've seen says a 1440p native CB'd game ends up having a similar effective resolution as 1800p) except without as many artifacts so thats definitely an improvement over CB.

The test on the sign is garbled and not as clear on the 1440p DLSS image, compared to the native 4K one. The texture detail on the wall is better in the native 4K image too.

I think maybe a DLSS mode with 1800p as the starting res would look much closer to native 4K. For me, on a 65" 4K TV I can clearly see a noticeable sharpness increase when I go from even 3712x2088 to native 4K. So even though DLSS is better I can still tell its not native straight away.

I am going to wait for a few more versions of it until I care about DLSS tbh. I'd prefer to optimise my settings to keep a locked 30/60 at native 4K than use this.

I'll have to examine the images again, because I mainly focused on the character.

 

[nani17]

OK looks good what's your average FPS?

 

[BluRayHiDef]

OK looks good what's your average FPS?

With maximum settings (including ray tracing), I get the following.

Native 4K: Mid twenties to thirties.

1440p to 4K via DLSS: Eighties to nineties.

1080p to 4K via DLSS: nineties to one hundred thirties.

 

[Kuranghi]

I'll have to examine the images again, because I mainly focused on the character.

Another thing to look for is how DLSS'd images alter the contrast of the image and sometimes make textures too defined when on the native image its more subtle. Grain is affected as well so the image takes on a certain "thickness" I'm not a fan of.

TBH though if you think it looks as good then just enjoy your awesome framerate increase, all listening to me/others will do is train your eyes to see the differences and ruin it for you.

What size of screen are you on and how far away do you sit from it?

 

[Kuranghi]

I could do a "spotting shit" session if you guys want to come round to my house (Wait, not a good time for that...), but you'll just leave feeling sadder than when you entered my house lol.

 

[BluRayHiDef]

Another thing to look for is how DLSS'd images alter the contrast of the image and sometimes make textures too defined when on the native image its more subtle. Grain is affected as well so the image takes on a certain "thickness" I'm not a fan of.

TBH though if you think it looks as good then just enjoy your awesome framerate increase, all listening to me/others will do is train your eyes to see the differences and ruin it for you.

What size of screen are you on and how far away do you sit from it?

55" screen. I sit two to three feet away from it.

Spoiler

 

[Kuranghi]

55" screen. I sit two to three feet away from it.

Spoiler

Alright, thats definitely within the distance where you could see what I'm seeing. Just making sure you weren't like 12 feet from a 43 inch screen or something like that.

 

[BluRayHiDef]

Alright, thats definitely within the distance where you could see what I'm seeing. Just making sure you weren't like 12 feet from a 43 inch screen or something like that.

I guess I'm just not as observant; I focus on the character and the rendering quality of her looks consistent between 1080p to 4K via DLSS, 1440p to 4K via DLSS, and Native 4K. However, due to you pointing out the signage, I have noticed the inferior clarity of the text in the DLSS images. However, the DLSS images still look as clean overall as the native 4K image, in my opinion (i.e. equal levels or lack thereof of aliasing, clarity of character models, etc).

 

[Rikkori]

In my opinion, the DLSS images are sharper than their native-counterparts

Yes, that's called sharpening. In Control in particular they add shitloads of sharpening with DLSS, as well as contrast modifiers. You should add CAS to native (50-60%) for comparisons. Doesn't change the conclusion though, in Control it works very well, but it has a shit TAA to begin with.

 

[BluRayHiDef]

Yes, that's called sharpening. In Control in particular they add shitloads of sharpening with DLSS, as well as contrast modifiers. You should add CAS to native (50-60%) for comparisons. Doesn't change the conclusion though, in Control it works very well, but it has a shit TAA to begin with.

LOL.

 

[Kuranghi]

I guess I'm just not as observant; I focus on the character and the rendering quality of her looks consistent between 1080p to 4K via DLSS, 1440p to 4K via DLSS, and Native 4K. However, due to you pointing out the signage, I have noticed the inferior clarity of the text in the DLSS images. However, the overall image still looks as clean as the native 4K image, in my opinion.

It could be that you have settings activated on your TV that make it harder to notice the stuff I see on my calibrated display. Like dynamic colour and contrast settings, because they remove fine detail, make gradients harsher and blow out colours if turned up too high.

 

[M1chl]

I have a 65" screen thats 8 feet away and I have to display them fullscreen to fully see the differences, its just how you compare images. You show them 1:1 on your screen, you don't need to zoom in.

Damn I came here for simping on Control yet again, but now I am intrigued about what kind of screen you have. Seems pretty massive...

 

[EverydayBeast]

Gamers have shown they will spend $ on 4K

 

[RedVIper]

DLSS comparisons without comparing performance are beyond retarded.

The best way to comparate it (IMO) would be this.

Set the game to a certain resolution, let's say 1080p and see what performance you get.

Then push dlss as much as you can until you match that performance.

Compare the resulting images.

 

[ZywyPL]

Does Control also have forced TAA like QB had? If so, calling it "native 4K" is stretching it a bit. But anyway, there's no reason to not to use DLSS in this game, the results are just too good.

 

[DonJuanSchlong]

 

[Kuranghi]

Damn I came here for simping on Control yet again, but now I am intrigued about what kind of screen you have. Seems pretty massive...

Its a Sony 65" ZD9, lovely TV, heres it is compared to a 2017 OLED, most stunning HDR I've seen overall but I haven't seen a ZG9. Was £4000 in 2016 but I got it the last open-box set in my country for £1300 in 2018:




The ZD9 is on the left but you shouldn't judge the picture of either set from the video.

 

[Soulblighter31]

Everytime I say my piece on this people get pissed lol.

I think the 1080p and 1440p DLSS'd to 4K images definitely have a good reduction in aliasing over the native 1080p and 1440p images, and have less arifacts in the hair as well.

As far as the DLSS images looking as good as the native 4K one I can't agree. I think the 1440p DLSS images look similar to 1800p native. So for me its similar to PS4 Pro checkerboarding (analysis I've seen says a 1440p native CB'd game ends up having a similar effective resolution as 1800p) except without as many artifacts so thats definitely an improvement over CB.

The test on the sign is garbled and not as clear on the 1440p DLSS image, compared to the native 4K one. The texture detail on the wall is better in the native 4K image too.

I think maybe a DLSS mode with 1800p as the starting res would look much closer to native 4K. For me, on a 65" 4K TV I can clearly see a noticeable sharpness increase when I go from even 3712x2088 to native 4K. So even though DLSS is better I can still tell its not native straight away.

I am going to wait for a few more versions of it until I care about DLSS tbh. I'd prefer to optimise my settings to keep a locked 30/60 at native 4K than use this.

It depends on the games. You are right regarding the text on the board. But look, in Death Stranding, its the opposite. The text is unreadable on native 4k and gets clearer with dlss. Checkerboarding is the shittiest solution that produces artifacts in the image. It doesnt have any place being compared to dlss. Its a comparison between the absolute worst reconstruction technique on the market and the best.

Have a look at these pics for the recently release game Pumpkin Jack. The one in the left is native 4k, looking at materials in the distance and the one in the right is DLSS. Not only does dlss offer an immensely better quality, you can literally see how the AI adds details that it deems correct for the scene. Its quite amazing in that game

 

[Kuranghi]

It depends on the games. You are right regarding the text on the board. But look, in Death Stranding, its the opposite. The text is unreadable on native 4k and gets clearer with dlss. Checkerboarding is the shittiest solution that produces artifacts in the image. It doesnt have any place being compared to dlss. Its a comparison between the absolute worst reconstruction technique on the market and the best.

Have a look at these pics for the recently release game Pumpkin Jack. The one in the left is native 4k, looking at materials in the distance and the one in the right is DLSS. Not only does dlss offer an immensely better quality, you can literally see how the AI adds details that it deems correct for the scene. Its quite amazing in that game

I'd need to see Pumpkin Jack running in native 4K and then the same frame but DLSS'd from 1440p or another res. Or a link for the article at least?

 

[Kuranghi]

But look, in Death Stranding, its the opposite.

I've seen it, I don't disagree that the text is sharper but I already said why I don't like it compared to native res. Its the same in DS as in Control in that regard for me, I made a post about that video as well, saying how I see it altering gradients, contrast and colour for the worse.

 

[Soulblighter31]

Those are my pictures, they're not from an article. First one is native 4k, no rtx, no dlss. Second one is rtx on, dlss on

 

[Kuranghi]

I don't think it looks better overall . I've weighed up all the pro and cons and overall DLSS looks too much like contrast adaptive sharpening to me. I actually prefer AMDs CAS though ironically, but thats because I was using it with native res.

It flattens the image overall for me, really look at the super fine detail on her cheeks, you can see how DLSS smooths it out. If you don't notice then great for you, you get "better than native res IQ" and higher fps.

Just please stop saying everyone who disagrees is mental or something lol

 

[Kuranghi]

Those are my pictures, they're not from an article. First one is native 4k, no rtx, no dlss. Second one is rtx on, dlss on

Sorry to tell you mate, but I think its just the RTX changing the definition of the stuff in the distance. Its RTX changing how the depth of field, blur or lighting is being rendered, making it way sharper.

Post an image where RTX is off for both images, make the only difference be that one is native 4K and the other is DLSS on (From whatever source res you choose, just tell us what it is, if you don't get a choice then just turn it on when the game is set to 4K), same settings, etc.

I could tell something is up because the wooden boards and other things aren't gaining any detail at all.

 

[Dave_at_Home]

I could do a "spotting shit" session if you guys want to come round to my house (Wait, not a good time for that...), but you'll just leave feeling sadder than when you entered my house lol.

 

[Kuranghi]

"spotting shit" = Helping people see the differences in IQ between two different images of a game with different settings. Or differences in multiplatform game's versions.

Off topic: What the fuck has happened to David's face, its disturbing.

 

[Dave_at_Home]

"spotting shit" = Helping people see the differences in IQ between two different images of a game with different settings. Or differences in multiplatform game's versions.

Off topic: What the fuck has happened to David's face, its disturbing.

Also, could be the Gif, but probably Botox.

 

[Kuranghi]

Also, could be the Gif, but probably Botox.

I think its a combo of both problems lol

I forgot to say after the IQ comparisons we have sex using our armpits, obviously.

 

[RedVIper]

Those are my pictures, they're not from an article. First one is native 4k, no rtx, no dlss. Second one is rtx on, dlss on

I don't think comparing it with rtx on and off is fair, and I'm a dlss fan.

 

[thelastword]

It depends on the games. You are right regarding the text on the board. But look, in Death Stranding, its the opposite. The text is unreadable on native 4k and gets clearer with dlss. Checkerboarding is the shittiest solution that produces artifacts in the image. It doesnt have any place being compared to dlss. Its a comparison between the absolute worst reconstruction technique on the market and the best.

Have a look at these pics for the recently release game Pumpkin Jack. The one in the left is native 4k, looking at materials in the distance and the one in the right is DLSS. Not only does dlss offer an immensely better quality, you can literally see how the AI adds details that it deems correct for the scene. Its quite amazing in that game

Yes Sharper on the right, but rest assured the entire image looks different as opposed to the artist's vision.

What I'm seeing here is heavy sharpening applied to the DLSS image with TAA on, but look how the image on the right has lost it's composition. On the left we have DOF, the water displays a level of thickness with the proper caustics and an accurate look, where on the right it looks like some blue paint was thrown in the water. looks flatter without the appropriate water undulations and detail. The piece of wood submerged lacks the depth and detail shown on the left, it just looks super-imposed on a flat 2d surface. The entire visual makeup is lost on the right just going for a sharper approximated picture......


Now if you just want a sharper picture on native 4K. A proper comparison would be to sharpen a native 4K image with no TAA, perhaps even no DOF because DOF is now lost in the DLSS image, so 4k native, no TAA, one pic with DOF and one without DOF. Another image to present would be a 4k image with TAA with sharpening, with DOF and one without DOF, for better comparisons vs your DLSS 1080p or 1440p DLSS images.

 

[The Cockatrice]

We already discussed this a million times when Control launched. Why are we discussing it again? It looks good and the performance gain outweighs the artifacts. End of discussion.

 

[BluRayHiDef]

Does Control also have forced TAA like QB had? If so, calling it "native 4K" is stretching it a bit. But anyway, there's no reason to not to use DLSS in this game, the results are just too good.

I had been using DLSS Quality Mode (1440p to 4K) up until today, but I decided to sample DLSS Performance Mode (1080p to 4K). I think that it looks just as sharp and smooth as the former and love the additional frames per second; the game runs at 80 to over one hundred frames per second now.

 

[metareferential]

It even looks better than native resolution, resolving more detail in certain areas.

And yes, anyone who disagrees is mental.

 

[Diddy X]

Doesn't look better than native at all. Also @ those other pics from the pumpkin game, they have different settings and those plants on the background look just like the devs want them to look, if you make them look different you are basically altering/messing up the image.

 

[Polygonal_Sprite]

It’s a crazy technology. Really hope it’s implemented into the next Switch. The idea it could be running games at native 480p and be displaying it as if it was native 1080p is amazing.

 

[BluRayHiDef]

Doesn't look better than native at all. Also @ those pics, they have different settings and those plants on the background look just like the devs want them to look, if you make them look different you are basically altering/messing up the image.

Wrong. All of the images are of the game with the same settings: maximum settings (including maximum ray tracing settings).

 

[Diddy X]

Wrong. All of the images are of the game with the same settings: maximum settings (including maximum ray tracing settings).

I was referring to the other pics, sorry.

 

[fersnake]

With maximum settings (including ray tracing), I get the following.

Native 4K: Mid twenties to thirties.

1440p to 4K via DLSS: Eighties to nineties.

1080p to 4K via DLSS: nineties to one hundred thirties.

what? you can do 1080p to 4k? i though i just worked with 2k and forward?

 

[DonJuanSchlong]

what? you can do 1080p to 4k? i though i just worked with 2k and forward?

You can do something like 960p IIRC

 

[BluRayHiDef]

what? you can do 1080p to 4k? i though i just worked with 2k and forward?

There's even an option below 1080p, though I can't recall what it is.

 

[sertopico]

I've been playing Control @2880p DSR (smoothness 15%) with DLSS set @1440p, which is my native resolution: overall the image looks indeed sharper and cleaner than native, but there are still some issues: many surfaces like metal gratings and such show a quite noticeable shimmering when moving the camera, something that the DLSS isn't apparently able to "resolve" succesfully. If you stand still the image is perfect of course, but as soon as the character starts moving the temporal aliasing kicks in on these selected surfaces. I hope that the next iteration of this technique will correct these problems.

 

[BluRayHiDef]

I've been playing Control @2880p DSR (smoothness 15%) with DLSS set @1440p, which is my native resolution: overall the image looks indeed sharper and cleaner than native, but there are still some issues: many surfaces like metal gratings and such show a quite noticeable shimmering when moving the camera, something that the DLSS isn't apparently able to "resolve" succesfully. If you stand still the image is perfect of course, but as soon as the character starts moving the temporal aliasing kicks in on these selected surfaces. I hope that the next iteration of this technique will correct these problems.

I'm a bit confused about your combined usage of DSR and DLSS.

DSR renders the game at 2880p but then downscales it to 1440p, afterwhich DLSS renders it at a resolution that's lower than 1440p but then upscales it back to 1440p?

I'm pretty sure I'm wrong. Please explain how these two technologies work together.