Support NeoGAF

[Bojji]

PSSR is comparable to DLSS3 as it works on same basis. Not as refined though as it's relatively new and require time for tweaking


FSR4 is a transformer (cnn-transformer hybrid) - of course it's more advanced. It's a level of DLSS4 and Pro is waiting for backporting to int8 that is known to be on the way.


PSSR as a model is DLSS2-3 (CNN).

Model may be similar but end results aren't. PSSR is completely incompatible with some games/engines unlike other ML reconstruction techniques.

XeSS is also CNN model, older than PSSR. There is also leaked FSR4 Int 8 model that is close in quality to FP8 model and runs on shader cores:

 

[Karim]

The original comment by Bojji mentioned DLSS 4. But the DLSS 3 CNN model the Switch 2 uses is still mostly superior to PSSR, while the lite model is obviously not. This will likely change next year, and the new PSSR should be superior to DLSS 3.


Yes, as resolution decreases then DLSS starts struggling as well. Although DLSS does offer a superior image to PSSR at any rendering resolution. Next year things will hopefully be different and the gap should shrink significantly.

Superior based on what? Forgive me but I'm leaning to not believe to a couple of unknown people on youtube who always leaning to be DLSS fans for such comparison.

 

[Karim]

It's not comparable to DLSS3 or DLSS4 or FSR4. It's just as simple as that.

I tested vast majority of Pro games, when I bought it I had one month of PS Plus so games were on the service in full or in a form of Trials. And unlike you I could compare end results with FSR3, DLSS, TAAU or TSR found in PC versions.

If PSSR was as good as you or some other folks claim, then your lord and savior Cerny woudn't say this:

"The neural network (and training recipe) in FSR 4's upscaler are the first results of the Amethyst collaboration," Cerny told us. "And results are excellent, it's a more advanced approach that can exceed the crispness of PSSR. I'm very proud of the work of the joint team!"

Link.



PSSR is close to DLSS3 only in few games, The Last of Us 2 looks excellent with it. Apparently KCD2 and Yotei also have good implementations. Rest of games have more or less issues.



That's not true at all. Show examples.

oh God. here we go. It's not comparable for you lol. I'd like to see AC Shadow at 800p scaled to 4k in motion whatever DLSS you want to see how much is so superior to PSSR. But sure if you want to use the worst cases how not.

 

[FireFly]

I'm talking about running off battery. Obviously if you're charging the unit at the same time it's going to increase the power consumption since that power doesn't come from nowhere.

You simply can't escape the laws of physics. A 20Wh battery can't supply more than 10W of power for 2+ hours! If it could then it wouldn't be a 20Wh battery.

 

[Bojji]

Superior based on what? Forgive me but I'm leaning to not believe to a couple of unknown people on youtube who always leaning to be DLSS fans for such comparison.

oh God. here we go. It's not comparable for you lol. I'd like to see AC Shadow at 800p scaled to 4k in motion whatever DLSS you want to see how much is so superior.

And you are basing this on what? Show me some comparison to prove your words.

 

[Karim]

And you are basing this on what? Show me some comparison to prove your words.

What I'm basing? I seen DLSS in action is not that change of life at lower resolution than 1080p. But if you are able to make a clip comparison about AC Shadows, I'm interested. Personally I haven't the time neither the hardware.

 

[Deerock71]

It's my opinion? Lol there is a video analysis which prove they downgraded the poly counts in the environment on switch 2

This is the opinion part, "but very apparent in a big screen." As a counter-opinion...no it's not. I don't have a PS5 and a Switch 2 on my wall side-by-side, and I'm not playing Star Wars: Outlaws simultaneously on both systems. It's only final-form nerds that give a shit.

 

[rodrigolfp]

I'm talking about running off battery.

That is exactly what the guy is measuring at the time I stamped.

 

[Karim]

This is the opinion part, "but very apparent in a big screen." As a counter-opinion...no it's not. I don't have a PS5 and a Switch 2 on my wall side-by-side, and I'm not playing Star Wars: Outlaws simultaneously on both systems. It's only final-form nerds that give a shit.

It's not an opinion. There a lot less NPCs, lower LOD and raw edges around the poly which are "rounded" on base console. Yes on handheld mode is barely noticeable, but in a big screen is quite apparent. But it's fine, in the end it's an handheld console, docked mode is just a nice bonus.

 

[daclynk]

So this confirms it will be shit on Switch 2 then ?

Or is it a Nintendo DLSSlite like version Nintendo Patented for lightweight DLSS on Switch 2 a while back.

 

[FireFly]

That is exactly what the guy is measuring at the time I stamped.

"We have the USB-C plugged into the USB-C on the Bluetti power station. [...] I'm going to have to do more extensive tests to see what kind of charging we're getting while we're playing. "

And you can literally see the cable plugged into the Switch 2's top USB port in the video.

 

[rodrigolfp]

"We have the USB-C plugged into the USB-C on the Bluetti power station. [...] I'm going to have to do more extensive tests to see what kind of charging we're getting while we're playing. "

And you can literally see the cable plugged into the Switch 2's top USB port in the video.

So? Is it inaccurate?

 

[Bojji]

What I'm basing? I seen DLSS in action is not that change of life at lower resolution than 1080p. But if you are able to make a clip comparison about AC Shadows, I'm interested. Personally I haven't the time.

Why only I have to show proofs, you and cireza like to talk about stuff but provide no evidence. I don't have access to ACS but two other games, H:FW and RDR2:

H:FW, 4k:

4K with DLSS UP (1280x720 -> 3840x2160):

Video:

Spoiler: 1440p and 1080p screens

RDR2, native 4k vs. 1280x720 -> 3840x2160:

 

[FireFly]

So? Is it inaccurate?

It's measuring the power consumed by the Switch 2 unit itself + the power allocated for charging. Hence it fails to tell us what the Switch 2 alone is consuming.

 

[Karim]

Why only I have to show proofs, you and cireza like to talk about stuff but provide no evidence. I don't have access to ACS but two other games, H:FW and RDR2:

H:FW, 4k:

4K with DLSS UP (1280x720 -> 3840x2160):

Video:

Spoiler: 1440p and 1080p screens

RDR2, native 4k vs. 1280x720 -> 3840x2160:

Nothing in motion?

 

[Bojji]

Nothing in motion?

Video is there.

 

[Karim]

Video is there.

Anyway I already noticed the noise around the foliage just in the capture.

 

[daclynk]

here found it.



Alex explained it

 

[rodrigolfp]

+ the power allocated for charging.

Hmm. Makes sense.

 

[Bojji]

Anyway I already noticed the noise around the foliage just in the capture.

Now show me that PSSR provides results as good as this.

Edit: I forgot that this comparison was already done...

 

[kevboard]

And yet the pixel salad is super obvious at the top of the right pole, and definitely visible in other places. And this is on a static image. You can be sure that it will shimmer in movement.

if it shimmers with DLSS, it will 99% of the time also shimmer with TAA. often more with TAA.

 

[yamaci17]

Now show me that PSSR provides results as good as this.

Edit: I forgot that this comparison was already done...

here's a funny comparison we made with my friend in ac shadows

https://imgsli.com/MzcxMjQz/1/2

quality mode with PSSR runs above 1440p from I recall

not exactly fits the narrative here but it's just impressive what it can do with 1080p

I also think PSSR here uses sharpening but that is up for debate

 

[cinnamonandgravy]

it's a meme that their community started and which they now use in videos.

"full fat" as an expression's been around for decades

 

[kevboard]

"full fat" as an expression's been around for decades

yes, but the comments started referring to stuff as "full fat raytraying", "full fat console experience" or "full fat DLSS" to the point where it now became a recurring meme for them.

 

[cireza]

if it shimmers with DLSS, it will 99% of the time also shimmer with TAA. often more with TAA.

Back to native it is then !

 

[kevboard]

Back to native it is then !

you can't play without TAA in most modern games. because sadly most modern engines go down the same shitty route that Unreal Engine went down, where everything is running at low fidelity with a dithering effect, and TAA is used to smear it into a coherent image.

here's what hair looks like in modern games if you run without TAA:

in many games the same dithering pattern is used for shadows, AO, foliage and of course SSR.

 

[Bojji]

you can't play without TAA in most modern games. because sadly most modern engines go down the same shitty route that Unreal Engine went down, where everything is running at low fidelity with a dithering effect, and TAA is used to smear it into a coherent image.

here's what hair looks like in modern games if you run without TAA:

RDR2 with MSAAx4:

Many edges are not anti aliased at all (most even). In motion... it hurts eyes.

here's a funny comparison we made with my friend in ac shadows

https://imgsli.com/MzcxMjQz/1/2

quality mode with PSSR runs above 1440p from I recall

not exactly fits the narrative here but it's just impressive what it can do with 1080p

I also think PSSR here uses sharpening but that is up for debate

Shadows has one of the better PSSR implementations. That's for sure.

 

[kevboard]

RDR2 with MSAAx4:

Many edges are not anti aliased at all (most even). In motion... it hurts eyes.

MSAA only covers polygon edges. some MSAA implementations also render foliage at higher res, but that can become insanely taxing in modern games.

MSAA is like a cheaper form of SSAA. SSAA renders the entire image at a higer resolution and then downsamples it.
MSAA renders all the polygon meshes at a higher resolution and downsamples them.
however, every other element of the image is rendered, basically fully untreated, at the output resolution.
and untreated edges in combination with the high frequency detail of a game like RDR2 will indeed probably look awful lol.

what you can try however is using Nvida's DLDSR in combination with MSAA. that should result in a very clean and sharp image. (but could he very taxing)

 

[Bojji]

MSAA only covers polygon edges. some MSAA implementations also render foliage at higher res, but that can become insanely taxing in modern games.

MSAA is like a cheaper form of SSAA. SSAA renders the entire image at a higer resolution and then downsamples it.
MSAA renders all the polygon meshes at a higher resolution and downsamples them.
however, every other element of the image is rendered, basically fully untreated, at the output resolution.
and untreated edges in combination with the high frequency detail of a game like RDR2 will indeed probably look awful lol.

what you can try however is using Nvida's DLDSR in combination with MSAA. that should result in a very clean and sharp image. (but could he very taxing)

I think since deffered rendering appread MSAA stopped being vialable solution. In older games it provides almost perfect image quality but since PS3 gen is started to look like shit and just kill FPS. FXAA and SMAA tried to anti alias games but those techniques are dogshit.

TAA came and finally fixed image quality but introduced issues like ghosting and overall blur. DLSS fixes some of them.

 

[Karim]

here's a funny comparison we made with my friend in ac shadows

https://imgsli.com/MzcxMjQz/1/2

quality mode with PSSR runs above 1440p from I recall

not exactly fits the narrative here but it's just impressive what it can do with 1080p

I also think PSSR here uses sharpening but that is up for debate

It's 800p at 60 fps. Probably in balance mode too.

 

[Zathalus]

Superior based on what? Forgive me but I'm leaning to not believe to a couple of unknown people on youtube who always leaning to be DLSS fans for such comparison.

Superior based on various videos online and my own comparisons. Most recently I checked the difference with Star Wars Outlaws and the stability on the grass was night and day when switching between my Pro and PC. If PSSR didn't have these issues we wouldn't have Sony stating they will be upgrading it with the work done on FSR4.

 

[Karim]

Now show me that PSSR provides results as good as this.

Edit: I forgot that this comparison was already done...

Alan wake 2 isn't hasn't exactly the best PSSR implementation...

 

[Bojji]

Alan wake 2 isn't hasn't exactly the best PSSR implementation...

They used PSSR.dll provided by Sony. They did exactly same thing with FSR2 and DLSS.

This comparison is also vs. "old" DLSS3.

 

[Karim]

Superior based on various videos online and my own comparisons. Most recently I checked the difference with Star Wars Outlaws and the stability on the grass was night and day when switching between my Pro and PC. If PSSR didn't have these issues we wouldn't have Sony stating they will be upgrading it with the work done on FSR4.

This is a nonsense. Sony shouldn't upgrade PSSR with FSR4 algorithm when they collaborate with AMD? This is pure console warring lol. AI logic need to be update steadily.

 

[Zathalus]

This is a nonsense. Sony shouldn't upgrade PSSR with FSR4 algorithm when they collaborate with AMD? This is pure console warring lol.

I think you misread my statement. PSSR is being upgraded with FSR4 because FSR4 is better. If PSSR was equal to DLSS then that upgrade would not be needed. I never said Sony shouldn't upgrade, they obviously should, because by their own admission FSR4 is better.

 

[Karim]

I think you misread my statement. PSSR is being upgraded with FSR4 because FSR4 is better. If PSSR was equal to DLSS then that upgrade would not be needed. I never said Sony shouldn't upgrade, they obviously should, because by their own admission FSR4 is better.

It's not upgraded to FSR4. It adds FSR4 logic to the existing AI algorithm.

 

[Bojji]

I think you misread my statement. PSSR is being upgraded with FSR4 because FSR4 is better. If PSSR was equal to DLSS then that upgrade would not be needed. I never said Sony shouldn't upgrade, they obviously should, because by their own admission FSR4 is better.

Cerny even stated that AMD started developing FSR4 after PSSR was done (and AMD said that they used some Sony things to train it) and that FSR4 provides better results. Just like you said, if PSSR was as good as DLSS or FSR4 there would be no need for total change, just slight upgrades here and there.

 

[Karim]

Cerny even stated that AMD started developing FSR4 after PSSR was done (and AMD said that they used some Sony things to train it) and that FSR4 provides better results. Just like you said, if PSSR was as good as DLSS or FSR4 there would be no need for total change, just slight upgrades here and there.

So you don't understand how AI work. More logics you have, better the AI works. PSSR is very young, it has less than an year of informations and surely hasn't all the experience of FSR.

 

[Zathalus]

It's not upgraded to FSR4. It adds FSR4 logic to the existing AI algorithm.

That's not how it works. You can't simply mesh two algorithms together like that. Especially since PSSR doesn't do a single thing better than FSR4 which is itself superior to DLSS3. The work AMD is doing on INT8 is obviously targeted at the Pro.

 

[Karim]

That's not how it works. You can't simply mesh two algorithms together like that. Especially since PSSR doesn't do a single thing better than FSR4 which is itself superior to DLSS3. The work AMD is doing on INT8 is obviously targeted at the Pro.

It's exactly how works the AI logic lol. More informations you give the better it works. And none program existent using just a single algorithm.

 

[Bojji]

So you don't understand how AI work. More logics you have, better the AI works. PSSR is very young, it has less than an year of informations and surely hasn't all the experience of FSR.

FSR4 is much younger than PSSR and much better. How you explain that?

FSR2 and 3 don't have any Machine Learning in them.

 

[kevboard]

That's not how it works. You can't simply mesh two algorithms together like that. Especially since PSSR doesn't do a single thing better than FSR4 which is itself superior to DLSS3. The work AMD is doing on INT8 is obviously targeted at the Pro.

and, fingers crossed, hopefully also for the Series X.

 

[Karim]

FSR4 is much younger than PSSR and much better. How you explain that?

FSR2 and 3 don't have any Machine Learning in them.

They have years of FSR1, FSR2 and FSR3 experiences counting all the pc games with them. They probably refine the past logics with the AI learning.

 

[yamaci17]

They have the FSR2 and FSR3 experiences and all the pc games with them.

 

[Bojji]

They have the FSR2 and FSR3 experiences and all the pc games with them.

FSR 2 and FSR 3 have nothing to do with FSR4, they only share name.

FSR2/3 is in the same group as TSR and TAAU, FSR4 is ML upscaler like PSSR, DLSS, XeSS and Apple Metal SR.

 

[Karim]

Do you really think they rewrite from zero the FSR4? The things they have learned with the previouses FSR are used to achieve FSR4.

 

[Zathalus]

It's exactly how works the AI logic lol. More informations you give the better it works.

Each model is trained on its own data and interprets data per the models they are trained on. They are also executed on very different hardware, PSSR runs on INT8 compute while FSR4 runs on FP8. Since FSR4 literally does everything better than PSSR what is the point of keeping the old PSSR around? You're the only one defending this, PSSR is getting upgraded next year anyways, and you're the only one who thinks it currently matches DLSS.

Have you actually tested this yourself? Do you have a PC with DLSS to compare? Because even someone half blind can choose between them.

 

[Karim]

Each model is trained on its own data and interprets data per the models they are trained on. They are also executed on very different hardware, PSSR runs on INT8 compute while FSR4 runs on FP8. Since FSR4 literally does everything better than PSSR what is the point of keeping the old PSSR around? You're the only one defending this, PSSR is getting upgraded next year anyways, and you're the only one who thinks it currently matches DLSS.

Have you actually tested this yourself? Do you have a PC with DLSS to compare? Because even someone half blind can choose between them.

Again you clearly haven't a clue how AI training works or how programming work.

 

[yamaci17]

Do you really think they rewrite from zero the FSR4?

honestly, yeah, and I really hope so
i'd even assume dlss 4 has nothing to do with dlss 3, as it has big problems that dlss 3.7 didn't have

 

[Karim]

honestly, yeah, and I really hope so
i'd even assume dlss 4 has nothing to do with dlss 3, as it has big problems that dlss 3.7 didn't have

It's not a bad thing if it uses also the DLSS3. AI can discern or mix the better of the logics of both version to give the best result.