Support NeoGAF

[jaaz]

Those who want to see and show reality use an image with light.



It's amazing, to hate Sony the guy even uses images where nobody can see anything to say that there's nothing to see.

Yep, this is what it looks like on my Pro and LG OLED, captured using the in game screenshot feature an hour ago.

 

[Shmunter]

The Pro just doubled in value

 

[ShakenG]

The improvement in quality and system wide use of it is huge.
Seriously cant wait to use this on GT7 and space marine 2.

 

[Gp1]

Seeing the comparison, it look like they left the sharpness at default value. Every FSR4 game that i tested you can adjust the sharpness with FSR4.

I would like to see it against the FSR 4.1 (i tested it on Cyberpunk, its even sharper than 4.0)

 

[Three]

My @onQ123 moment:

Pro has int8 version of FSR4.1 according to Kepler

But CrustyBritches and Bojji you were both suggesting that FSR4 and Sony's efforts parted ways and did their own thing just this week:

Look who liked the above post saying Pro couldn't handle FSR4 and how they're not on the same path due to Cerny's mistake or something.
From this thread.

 

[Topher]

But CrustyBritches and Bojji you were both suggesting that FSR4 and Sony's efforts parted ways and did their own thing just this week:

Look who liked the above post saying Pro couldn't handle FSR4 and how they're not on the same path due to Cerny's mistake or something.

 

[poppabk]

Talk about still not understanding how Amethyst works. You just don't get it. PSSR (1or2) have different priorities to FSR4. One is biased for fixed framerate variable resolution, the other is the opposite. One is targetted at large distant screens, the other close monitors. One is float the other is int. Not to mention a fixed hardware target vs multiple GPU'S.

The underlying techniques and models are shared but individually tuned. There is no merge, these are forks feeding back into a common codebase where it makes sense.

Your comment 'there is no point in separate upscalers' is utterly at odds with what is happening and what AMD and Cerny have decided is the best path.

We've seen on PC how the different upscalers can basically be injected in to games, hence why Sony can have a system wide toggle for PSSR2. Yeah they might change some parameters but the underlying upscaler is the same. I mean why do you think Int8 version of FSR4 exists if it isn't officially supported by any AMD GPU?

 

[ShakenG]

Exactly. Just knowing I don't have to wait for every dev to get around to patching the games is pretty huge for me.

So simply put -
With the option enabled, it will use the latest version kept on the firmware and resort to the default game included verion when switched off?
And when Sony updates the firmware version it holds, thats the version that you'll always be able to override with?

 

[Topher]

So simply put -
With the option enabled, it will use the latest version kept on the firmware and resort to the default game included verion when switched off?
And when Sony updates the firmware version it holds, thats the version that you'll always be able to override with?

That's my understanding, yes

 

[Inviusx]

I'm not sure what's confusing. Just leave the toggle on if you want games that use PSSR 1 to use PSSR 2 instead. Games like Resi 9 (developed with PSSR 2) will continue to use PSSR 2 whether the toggle is on or off I'd imagine.

That's not the confusing part, for me its this: does RE9 still work with PSSR2 if the toggle is turned off? Im assuming yes because it has been specifically designed with PSSR2 which is a separate implementation to the toggle. So is the quality of the upscaling going to be different for games going PSSR1 to PSSR2 using the toggle, compared to games specifically patched or developed with PSSR2 in mind?

I feel like the brute force toggle is still going to be incredibly hit and miss for current PSSR1 games and we will still all be wondering if future games have been specifically designed around PSSR2 or if devs will just rely on whatever brute force method Sony have developed.

Lots of ambiguity until the March firmware update.

 

[ChiefDada]

But CrustyBritches and Bojji you were both suggesting that FSR4 and Sony's efforts parted ways and did their own thing just this week:

Look who liked the above post saying Pro couldn't handle FSR4 and how they're not on the same path due to Cerny's mistake or something.
From this thread.

Bojji

 

[poppabk]

That's not the confusing part, for me its this: does RE9 still work with PSSR2 if the toggle is turned off? Im assuming yes because it has been specifically designed with PSSR2 which is a separate implementation to the toggle. So is the quality of the upscaling going to be different for games going PSSR1 to PSSR2 using the toggle, compared to games specifically patched or developed with PSSR2 in mind?

I feel like the brute force toggle is still going to be incredibly hit and miss for current PSSR1 games and we will still all be wondering if future games have been specifically designed around PSSR2 or if devs will just rely on whatever brute force method Sony have developed.

Lots of ambiguity until the March firmware update.

I think you have to get past the idea that any of this is optimized for a specific game.

 

[Topher]

That's not the confusing part, for me its this: does RE9 still work with PSSR2 if the toggle is turned off? Im assuming yes because it has been specifically designed with PSSR2 which is a separate implementation to the toggle. So is the quality of the upscaling going to be different for games going PSSR1 to PSSR2 using the toggle, compared to games specifically patched or developed with PSSR2 in mind?

I feel like the brute force toggle is still going to be incredibly hit and miss for current PSSR1 games and we will still all be wondering if future games have been specifically designed around PSSR2 or if devs will just rely on whatever brute force method Sony have developed.

Lots of ambiguity until the March firmware update.

I'm don't think I would call this "brute force". The PSSR library has been changed. The games are making the same calls to the library they always did, but the algorithms used are now different. I would think the new algorithms would include better handling of engines like UE5 and tech like lumen since that has been the most problematic so far.

 

[Three]

I think you have to get past the idea that any of this is optimized for a specific game.

Depending on the milliseconds cost of different versions he's somewhat right in that games are optimised with the cost of that specific version in mind and tested with that version during QA for problems. I don't think there will be a lot of difference outside of the improved image quality though so we should be alright. Depending on how that millisecond cost goes we might get slightly different fps with a PSSR1 game 'forced' to PSSR2. We'll see soon enough.

 

[Ashamam]

We've seen on PC how the different upscalers can basically be injected in to games, hence why Sony can have a system wide toggle for PSSR2. Yeah they might change some parameters but the underlying upscaler is the same. I mean why do you think Int8 version of FSR4 exists if it isn't officially supported by any AMD GPU?

Nope I don't believe thats an accurate take based on all the public comments. Priorities around compute budgets and console needs mean they went with a distinct fork in implementation. The underlying upscaler is obviously similar in how its trained, but not the same. I'd argue that the reason the went with a fork instead of a port is they are performance profiling much harder and likely applying all sorts of deep optimisations in how and where they apply the upscale. So basically a lot more than just changing some parameters.

This kind of takes me back to the whole lets ignore Cerny when he talks about architecture thing when the PS5 released. People in the community seemed determined to simplify things down to the common elements and ignore the divergences, when its pretty obvious that Cerny/Sony spend a lot of time on those divergences because thats where they can make the most impact. Eg: The whole its just a PC thing. Sure if you sort of squint your eyes and blur the picture, yes its ubiquitous hardware, but it ignores the efforts Sony puts in to achieving their very specific and very bespoke goals.

If you believe otherwise you aren't listening to Cerny/AMD, or you think they are lying. Thats fine if people do, but at least come out and say so rather than trying to rewrite whats on public record.

 

[poppabk]

Nope I don't believe thats an accurate take based on all the public comments. Priorities around compute budgets and console needs mean they went with a distinct fork in implementation. The underlying upscaler is obviously similar in how its trained, but not the same. I'd argue that the reason the went with a fork instead of a port is they are performance profiling much harder and likely applying all sorts of deep optimisations in how and where they apply the upscale. So basically a lot more than just changing some parameters.

This kind of takes me back to the whole lets ignore Cerny when he talks about architecture thing when the PS5 released. People in the community seemed determined to simplify things down to the common elements and ignore the divergences, when its pretty obvious that Cerny/Sony spend a lot of time on those divergences because thats where they can make the most impact. Eg: The whole its just a PC thing. Sure if you sort of squint your eyes and blur the picture, yes its ubiquitous hardware, but it ignores the efforts Sony puts in to achieving their very specific and very bespoke goals.

If you believe otherwise you aren't listening to Cerny/AMD, or you think they are lying. Thats fine if people do, but at least come out and say so rather than trying to rewrite whats on public record.

They forked because they don't have the hardware on the PS5 pro - so there had to be a different version.
Here is Cerny saying that PSSR2 isn't anything at all like FSR4 "So the co-developed algorithm has already been released by AMD as part of FSR 4 on PC. And we're in the process of implementing it on PS5 and it will release next year on PS5 Pro. It's not a cut-down [version] of the algorithm. It's the full-fat version of the co-developed super resolution that we'll be releasing on PS5 Pro."

 

[adamsapple]

I would want to see what that screenshot looks like without PSSR at all as well.

Ask and ye shall receive.

Direct Series X capture.

More aliasing on the hair, otherwise very comparable, certainly not like comparing hair between Switch 2 and PS5 on FFVII Rebirth, at least.

 

[Inviusx]

I think you have to get past the idea that any of this is optimized for a specific game.

Well I suppose Im thinking about all of the different implementations of PSSR across various engines from Pros launch until now. All have used PSSR to upscale from different base resolutions and have different challenges in terms of RT implementations, RTGI etc etc. We have one great example in RE9 where Capcom have optimised the Pro version around a base resolution of 1080p, upscaling to 4k but not all games have had the same level of care with PSSR as I suspect Capcom have done with this title.

So yeah I do think PSSR2s results will still be extremely game specific.

 

[Loboxxx]

Ask and ye shall receive.

Direct Series X capture.

More aliasing on the hair, otherwise very comparable, certainly not like comparing hair between Switch 2 and PS5 on FFVII Rebirth, at least.

It's not just that PSSR offers better quality, it's that the performance improvement also enables ray tracing. The importance of PSSR is not only to show better AA, but also to offer more FPS at better quality.

 

[adamsapple]

It's not just that PSSR offers better quality, it's that the performance improvement also enables ray tracing. The importance of PSSR is not only to show better AA, but also to offer more FPS at better quality.

Sure, I know that's the whole point of ML like DLSS/PSSR etc.

But in this game's specific case, the RT mode has drops that PS5/SX don't have, so it's definitely incurring an overhead in the pipeline.

 

[Zathalus]

There is also the issue that he forced the L preset. When the game ships with the K preset.
But somehow, he still used the FSR 4.0.3, when the FSR 4.1 is out in the wild and brings big improvements to the performance mode, especially using vegetation.

The video was shot several days ago, FSR 4.1 wasn't out yet, nor did anyone even know about it. You'd likely get a 4.1 video from DF (Hardware Unboxed as well) soon.

 

[Tobimacoss]

So the Pro is finally not useless, congrats to the owners.

 

[poppabk]

Well I suppose Im thinking about all of the different implementations of PSSR across various engines from Pros launch until now. All have used PSSR to upscale from different base resolutions and have different challenges in terms of RT implementations, RTGI etc etc. We have one great example in RE9 where Capcom have optimised the Pro version around a base resolution of 1080p, upscaling to 4k but not all games have had the same level of care with PSSR as I suspect Capcom have done with this title.

So yeah I do think PSSR2s results will still be extremely game specific.

Yes in those terms - the games won't change at all - the upscaling will just be better.

 

[Zathalus]

So, playing RE9 on the Pro and PSSR2 is indeed great. The only issue is the RT implementation is dog shit. The RE denoiser is terrible and boiling artifacts are really distracting. No real fault of Sony here, AMDs Ray Regeneration can't come to consoles fast enough. But I'd assume that would be for the PS6.

 

[cafecat]

Game call PSSR upscaler API in PlayStation SDK

1. with system pssr enchancement, call the pssr 2.0 (or indeed the latest model)
2. without system pssr enchancement, call the version developer implemented in the game

Want to see, how many resource PSSR 2.0 use vs PSSR 1.0, and how some pssr1.0 games' performance change with system pssr enchancement, the latest 2 slient hill games, GOW frame (PSSR1.0 gain sharpness but tank about 10 fps vs nore TAAU)

 

[Negotiator101]

Once again, buying Pro was a probably the best purchase of 2025. The gift that keeps on giving.

One game? You're the gift that keeps on giving hilarious

 

[Panajev2001a]

Patching simply gives more developer control on the desired result. For example, let's say the whole upsampling stage takes 1.5 ms with 2.0 instead of 2 ms. What should happen with the remaining 0.5 ms freed up frame time? Higher resolution with DRS? Should the DRS window be expanded? Or conversely, what if 2.0 is slower? It's always good for devs to patch the game to the latest version with some basic QA. And the system level override is for those lazy fucks who just won't bother.

For GT7, for example, if PSSR 2.0 is faster, I would love for Polyphony to restore LODs to the non-PSSR 1.0, non-RT levels. That would be the ultimate version of the game.

Agreed, but I think there is more. Like the kind of inputs PSSR2 needs and how it makes some assumptions when it is reusing the PSSR1 inputs / setup that you could tweak to get an even better result IF you patch the game and make some small changes.

I would not be surprised if PSSR2 is integrated exactly how FSR4 is on PC and that alone might drive people to patch the games to target PSSR2 natively to reduce your tech debt.

 

[Panajev2001a]

Game call PSSR upscaler API in PlayStation SDK

1. with system pssr enchancement, call the pssr 2.0 (or indeed the latest model)
2. without system pssr enchancement, call the version developer implemented in the game

Want to see, how many resource PSSR 2.0 use vs PSSR 1.0, and how some pssr1.0 games' performance change with system pssr enchancement, the latest 2 slient hill games, GOW frame (PSSR1.0 gain sharpness but tank about 10 fps vs nore TAAU)

That I get, but I bet the PSSR2 API in the SDK has some changes, arguments and properties that have sane defaults and try to make assumptions to use the same setup as PSSR1, but could yield even better results if you target the new API yourself and make decisions that benefit your game (decisions that might not apply to every game in a positive manner and thus are not to be enforced by the system to every game).

 

[demigod]

• Able to apply PSSR 2 system-wide, no per-game patches required. Killer move
• PSSR 2 passes the "it's more than good enough" test, which DF claims PSSR1 did not
• Compared to FSR 4 and DLSS 4, the differences aren't actually that important. Reaches the "Good enough, I don't care" threshold
• PSSR 2 solves virtually all the PSSR 1 problems
• Reaches the point of not being able to tell the difference from native rendering at a normal sitting distance on a 65" display

PSSR 2 bringing the goods. It's pretty much FSR 4. Not quite DLSS 4.5 but close. Sounds like a home run. And RE9 is probably the best PS5 Pro upgrade yet?

PSSR 2 looks crisper than FSR 4 in that 4x zoom.

Damn this really must hurt Bojji

 

[Gp1]

Once again, buying Pro was a probably the best purchase of 2025. The gift that keeps on giving.

Not even close to the guy that bought 32gb of RAM, a 4 TB SSD and a next gen GPU mid year.

 

[PropellerEar]

One game? You're the gift that keeps on giving hilarious

Leagues better than the sack of shit you guys are getting fed.

 

[Karim]

So the Pro is finally not useless, congrats to the owners.

The level of stupidity and uselessness of post like this is something else. Are people like you never tired to trolling about the ps5 pro? PC higher end still exist, no one forcing to look at the ps5 pro hardware if you don't like it.

 

[winjer]

The video was shot several days ago, FSR 4.1 wasn't out yet, nor did anyone even know about it. You'd likely get a 4.1 video from DF (Hardware Unboxed as well) soon.

The first person I saw using FSR4.1 was "The Creator" on the guru3d forums, on Wednesday morning.
It doesn't take that long to shoot 3 or 4 scenes for comparison.

But that is not even the biggest problem. Is that he forced Preset L on the game, to pretend the difference to PSSR2 was bigger than what it really is.

 

[Karim]

No mentioned of improvement of PS4 enhancement features kind of means we might not see any.

To be fair the ps4 enhancement option it's sensibly improved during this year for my eyes perception but no one seem interested to cover it.

 

[RPS37]

So the Pro is finally not useless, congrats to the owners.

Thank you, I always knew this investment would pay off.

/s

 

[Fortuneelsewhere]

This is huge!!! Finally the Pro delivering on its promises… glad to see that Sony still have it in them…

I really think PS6 will be a fantastic console and this update on Pro PSSR only cement that feeling..

 

[winjer]

Talk about still not understanding how Amethyst works. You just don't get it. PSSR (1or2) have different priorities to FSR4. One is biased for fixed framerate variable resolution, the other is the opposite. One is targetted at large distant screens, the other close monitors. One is float the other is int. Not to mention a fixed hardware target vs multiple GPU'S.

The underlying techniques and models are shared but individually tuned. There is no merge, these are forks feeding back into a common codebase where it makes sense.

Your comment 'there is no point in separate upscalers' is utterly at odds with what is happening and what AMD and Cerny have decided is the best path.

Not exactly that. A lot of the code and training will be similar.
But the main difference is that FSR4 on PC will use FP8. While the PSSR2 version will use Int8. Int8 will produce lower slightly quality results, but at a distance on a big TV, it doesn't make that much difference.
Another thing to consider is that FSR4 used a 3.3 pixel grid. But looking at the results with the FSR4.1 dll, at the Performance Mode, it seems FSR4 might be now using a bigger grid. Maybe 5x5. Or 7x7.
For a console to render at around 1080p to 4K, this would be ideal, as it is closer to the 4K Performance Mode.

 

[Ashamam]

Not exactly that. A lot of the code and training will be similar.

Well yes? Thats the point of the collab (Amethyst). Its the differences though that are important in this context. I mean you pretty much highlighted a potential example with your pixel grid comment. Once you aren't on the same hardware you automatically change the limitations you are working to. So from a resources point of view they have challenges, but there is also a target result issue. They aren't shooting for the exactly the same result/priorities.

 

[Physiognomonics]

Nice catch on the last one. The whole frame looks blurrier (see letters on warning sign on right for reference), so it likely is dropping inference quality across the entire image to keep the framerate up (or may be it's lowering resolution when fps drops). The "not doing lines as well" is not the right conclusion. Different tradeoffs are being made based on performance target and normal viewing distance. It could also be just a difference in versions of FSR used, like pointed above.

This is probably down to different sharpening filter or grain filter settings (even DF thought some differences could be due to this as they didn't know which settings was on PS5 and there was no way to remove those). There shouldn't be any difference between FSR4.1 FP8 or INT8 if they use the same settings.

Also with the DLSS image it seems the sharpening is much stronger than both FSR4 images. We'll know more when the PSSR toggle will be there.

EDIT: PSSR2 looking sharper than FSR4 in cutscenes could be explained by using different (less) grain filter / CA settings there as they often do in cutscenes.

 

[winjer]

Well yes? Thats the point of the collab (Amethyst). Its the differences though that are important in this context. I mean you pretty much highlighted a potential example with your pixel grid comment. Once you aren't on the same hardware you automatically change the limitations you are working to. So from a resources point of view they have challenges, but there is also a target result issue. They aren't shooting for the exactly the same result/priorities.

I don't think you understand, but increasing the pixel grid size, if that had really happened, is something that benefits both AMD and Sony. So their thinking is perfectly aligned.
FSR4 4.0 weakest point has been the Performance Mode and 4.1 significantly improves it.

Just for comparison, DLSS 4.5 Preset L, is supposed to use a 9x9 grid. Though that is the reason why it's so heavy.

 

[Radical_3d]

This is weird, he should left default K preset in comparison.



Even setting nvidia app to "recommended" would use M preset in performance mode.

He should use default and show L as potential best case scenario.

Anyways, I couldn't resist and played last night. I'm going to be honest: I thought every RE game in PS5 looked like that but I don't tent to remember graphical shimmering as part of my memories. We'll probably have more comparisons in the future to see about that. The change is obvious when you compare to the 120Hz mode but then again I don't remember RE2 in PS5 looking that bad at 60. Worst case scenario is a quality mode at 60, whiteout the problems of FSR1-3, which is what the machine is all about. I'm playing in a 65", for reference.

That said, the Capcom denoiser destroys that mode. 80% of the time is not a bother but there are a significant 20% cases in the first couple of hours where moving the camera destroys what is near the border of the screen. Noticeable in volumetric lights and some white points in the dark that I assume they are reflections badly resolved. Curious how this is not shown in the videos about the fucking graphics. C'mon John!

THAT SAID, the game is a blast and I've lost 20 years of my life in the first (non tutorial) stage with Grace. I think I'd have preferred a full Grace game where all is a horror game, but a bit of RE4 isn't that bad.

Any PSSR1 fan thinking that is was "great" should look at this picture very carefully

Spoiler

 

[Sept Terrion Of Nothing]

I don't understand some people that they wouldn't like evolution of PSSR or FSR. I'm a PC guy from the DLSS team and got PS5 as well but the improvement of Deep Learning upscalers are beneficial for us gamers in the end.

Keep that coming from all Team Blue, Red, Green. We're eating well now and will eat well in the future.

 

[Nankatsu]

I suspect that the reason RDNA3 and RDNA2 haven't got INT8 FSR4 officially is because by helping AMD tuning up FSR4, Sony secured some kind of exclusive time gating deal on that regard, in order to lock how widespread the upscaler would be available.

Because let's be honest, having FSR4 running on the Pro first makes Sony the face here, and AMD becomes the heel.

It's clearly a win for Sony here.

 

[Karim]

This is huge!!! Finally the Pro delivering on its promises… glad to see that Sony still have it in them…

I really think PS6 will be a fantastic console and this update on Pro PSSR only cement that feeling..

Don't be worry at the first indie studio with shitty PSSR2 in their game ps5 pro will go back to the total failure class. I don't expect flawless pssr2 in all games, things surely will going better but the usual suspects and the eternal dissatisfied will come back in masses at the first occasion. There were already some really nice pssr1 uses but people only point out to the worst ones

 

[MisterXDTV]

I suspect that the reason RDNA3 and RDNA2 haven't got INT8 FSR4 officially is because by helping AMD tuning up FSR4, Sony secured some kind of exclusive time gating deal on that regard, in order to lock how widespread the upscaler would be available.

Because let's be honest, having FSR4 running on the Pro first makes Sony the face here, and AMD becomes the heel.

It's clearly a win for Sony here.

PS5 Pro has the AI Tops of the 9070, so RDNA4

RDNA3, let alone RDNA2, can't compare in any way

There's no big conspiracy here, it's just hardware specs

 

[Corporal.Hicks]

2.0 is still decent upscaler, it never sucked when it was around.

1.0 was ASS.



I want to see comparisons in more games. Even those with good implementations of PSSR1.

Agreed. I played Returnal yesterday, and even with DLSS 2, the image was incredibly sharp and clean during motion. Of course, the DLSS4 K preset looks even better, but DLSS2 still destroys FSR2/3 in this game. If the PS5Pro PSSR looked as good as the DLSS2, console gamers wouldn't complain.

 

[FUBARx89]

I'm looking forward to going back to some other games to see how PSSR 2 compares.

RE4 (Texture warbling effect because of how badly PSSR resolves sniper scopes and fizzle on vegetation)
Silent Hill 2
FF VII Rebirth (Tifas hair in her house at Nibelheim looks atrocious) but I am hoping for an in game implementation for PSSR2 along with higher shadow settings, fidelity mode NPC counts and a better LOD distance.
MH-W
Stellar Blade (Only a bad implementation at the end of the game)

Hopefully they announce quite soon when this will be coming with a presentation by Mark Cerny and his dreamy voice, rather than just dumping it at the end of march.

 

[Karim]

2.0 is still decent upscaler, it never sucked when it was around.

1.0 was ASS.



I want to see comparisons in more games. Even those with good implementations of PSSR1.

Yeah I seen on switch 2 You should come back to play some games with the old DLSS2 version If you said PSSR is shitty you should define an abominition the first DLSS2 try as minimum.

 

[squidilix]

Damn, so GT7 with RT will looks more better with the new update?
Same for GOYotei with RT? Even better for FF7Rebirth ?
TLOUp1/2 already on next level will look more better wtf..

Im curious to see how UE5 games running and the RT denoiser (SH2). Curious to also see if Alan Wake 2 have fixed some issue.

But that IQ for RE9 Requiem, with RT and 60fps, is already wild.

 

[Aurelius]

Managed to get permission from the wife by using women logic. "It's only going to get more expensive in the future, so I'm saving money by buying it now".