Support NeoGAF

[Bojji]

it is a bit strange that all other upsampling methods have very distinct naming conventions. this is true for XeSS, DLSS, FSR, and TSR. PSSR is the only one that is this similar to another one, which happens to be XeSS.

so it is curious. and it wouldn't be that far fetched, as Intel already plans to eventually make it entirely open source.
so it wouldn't be surprising if Sony worked with Intel to fork off of it. especially since XeSS is very close to DLSS in quality in many games

Xess has more presets and isn't one of them 900p on 4k output?

This already would not corelate with ~800p used in few games on pro.

 

[Fafalada]

We know last of us res is 1440p, just like PS5 version.

But you have no reference for 'no PSSR', nor what the GPU/CPU limits are on this mode (on either system).
Basically it tells you nothing about the cost of PSSR at all.

Sony's original information (2ms) would give you a cost of approximately 30% of frame at 120fps - but that % will change if framerate changes of course, and when/if run async, the cost will be proportionally lower (but that'll depend entirely on the title).
(eg. on PSVR, the GPU cost of reprojection was close to 1.5ms on a standard PS4, but in practice the runtime cost was lower as it was an async job).

 

[Bojji]

But you have no reference for 'no PSSR', nor what the GPU/CPU limits are on this mode (on either system).
Basically it tells you nothing about the cost of PSSR at all.

Sony's original information (2ms) would give you a cost of approximately 30% of frame at 120fps - but that % will change if framerate changes of course, and when/if run async, the cost will be proportionally lower (but that'll depend entirely on the title).
(eg. on PSVR, the GPU cost of reprojection was close to 1.5ms on a standard PS4, but in practice the runtime cost was lower as it was an async job).

Didn't devs left standard performance mode to choose? You can compare frame rate with pssr and with taa using that. I think slimy was referring to that in op.

 

[S0ULZB0URNE]

We don't really know what's going on under the hood

It's early so hopefully in time they come to grips with the added horsepower and learn more about PSSR.

 

[Fafalada]

Didn't devs left standard performance mode to choose? You can compare frame rate with pssr and with taa using that. I think slimy was referring to that in op.

But the video comparison isn't showing that - or at least it's claiming something completely different. And that's the numbers everyone's going off of in this thread.
Anyway - if we could confirm modes are identical and test that - it would still leave a question on CPU side, but indeed that would be potentially 'underestimating' the cost, not overestimating it at least.

 

[kevboard]

Oh yeah, I forgot they were using an Intel GPU.

...XeSS runs on anything that supports a certain shader model...

 

[Bojji]

But the video comparison isn't showing that - or at least it's claiming something completely different. And that's the numbers everyone's going off of in this thread.
Anyway - if we could confirm modes are identical and test that - it would still leave a question on CPU side, but indeed that would be potentially 'underestimating' the cost, not overestimating it at least.

This video is showing PS5 vs pro so yeah.

Game should have old performance mode left in so proper comparison can be made.

 

[ThatPolishGuy]

it is a bit strange that all other upsampling methods have very distinct naming conventions. this is true for XeSS, DLSS, FSR, and TSR. PSSR is the only one that is this similar to another one, which happens to be XeSS.

so it is curious. and it wouldn't be that far fetched, as Intel already plans to eventually make it entirely open source.
so it wouldn't be surprising if Sony worked with Intel to fork off of it. especially since XeSS is very close to DLSS in quality in many games

Yeah ths guy here mentions that PSSR uses the same terminology as XeSS

It was hidden in S2R files.

 

[kevboard]

Xess has more presets and isn't one of them 900p on 4k output?

This already would not corelate with ~800p used in few games on pro.

all upscalers can use arbitrary resolutions. the presets are just the default way to set it up. but games can also just have resolution scaling sliders, and some do on PC.

some PC games also support dynamic resolution + reconstruction through DLSS, XeSS or FSR

 

[Mr Moose]

...XeSS runs on anything that supports a certain shader model...

I thought it didn't run fully unless you're using Intel.
And why would Sony work with them on PSSR?

 

[kevboard]

I thought it didn't run fully unless you're using Intel.

it runs faster on Intel because Intel has XMX cores that run the XMX version of XeSS.

on any other GPU it uses the DP4a version that can run entirely on shaders but is therefore a bit more costly and was noticeably lower quality in the past. however the quality was improved a lot and now the DP4a version looks very close to the XMX version.

the base PS5 doesn't have the necessary RDNA2 hardware acceleration to run XeSS at fast enough speeds (it could run there too but really badly), so the jump to RDNA3, or whatever exactly the Pro uses, helps here as well, on top of the bigger GPU

 

[SKYF@ll]

It's funny how some people are in total denial about the fact that PS5 Pro is not as ‘Pro’ as expected, and is just a PS5 with a bit more raw power (which goes away when using PSSR) and AI rescaling for only $699 (disc drive and stand not included).

Games that use FSR2(3) simply replace the cost used with PSSR.
The extra power that's left over can be used for other tasks.
PSSR(FSR2) cost of 2ms at 120fps is still too expensive though.
I hope that the PSSR cost will go down in future updates.

 

[Gaiff]

Games that use FSR2(3) simply replace the cost used with PSSR.
The extra power that's left over can be used for other tasks.
PSSR(FSR2) cost of 2ms at 120fps is still too expensive though.
I hope that the PSSR cost will go down in future updates.

It's fine at 30 and 60fps. At 120fps where there is only 8.33ms per frame, it's too high. It could be why they use TAAU in F1 24 instead of PSSR in the 120Hz mode.

 

[kevboard]

It's fine at 30 and 60fps. At 120fps where there is only 8.33ms per frame, it's too high. It could be why they use TAAU in F1 24 instead of PSSR in the 120Hz mode.

F1 also uses TAAU in the 8K mode doesn't it?
so PSSR 4k to 8k might also be a bit overkill for the GPU currently in some titles.

 

[SKYF@ll]

For gamers who want a higher frame rate, it's good that they kept the TAA mode.
We can choose between high quality images (PSSR 90fps) or high frame rates (TAA 110fps).
Higher frame rates also reduce input lag.

 

[rofif]

For gamers who want a higher frame rate, it's good that they kept the TAA mode.
We can choose between high quality images (PSSR 90fps) or high frame rates (TAA 110fps).
Higher frame rates also reduce input lag.

There is redundancy in those modes like in many pro games.
leaving old modes there and having few pro modes is redundant.
Performance pro mode destroys the normal performance mode and there is no reason to even show it on the pro. Nobody going to feel 90 to 110fps in random spots. It's not even consistent
Same with stellar blade. They left all the old crap modes in there... and added 2 modes + hfr toggle. Why?
HFR toggle should be hidden and automatic based on vrr+120hz status from ps5 system settings. Old modes should GO AWAY and pro max mode is stupid. Looks and runs worse than PRO mode.

I don't like redundancy. Console gaming is all about lowering redundancy so I don't waste my attention looking at modes.

If anything, to make all the weirdos happy, there could be a system toggle in ps5 system settings "disable pro", so games recognize it as normal ps5.

 

[SlimySnake]

But you have no reference for 'no PSSR', nor what the GPU/CPU limits are on this mode (on either system).
Basically it tells you nothing about the cost of PSSR at all.

Sony's original information (2ms) would give you a cost of approximately 30% of frame at 120fps - but that % will change if framerate changes of course, and when/if run async, the cost will be proportionally lower (but that'll depend entirely on the title).
(eg. on PSVR, the GPU cost of reprojection was close to 1.5ms on a standard PS4, but in practice the runtime cost was lower as it was an async job).

Of course we do. did you not watch the video?

if the gpu can hit 120 fps on the base ps5 then its not a CPU limit. its a purely gpu limit.

here you can see the ps5 gpu outperform the pro gpu in terms of framerate. we know for a fact that the ps5 pro gpu is faster, has more vram bandwidth so if the framerate is lower despite the internal resolution being identical to the base ps5, we can assume that the difference is exclusively due to the pssr reconstruction costs.

sony's docs said a lot of things. we dont know what framerate they were looking at. they simply said 1080p. they likely picked the best case scenario like they did for the gpu which they said was 45% faster but we know its much lower than that.

this is only one data point. other games might have a smaller hit in line with the 2ms sony outlined. ive seen this on PC with dlss. varies from game to game. dlss is slightly less expensive than fsr but it still has a very high cost. definitely higher than 2ms. pssr is basically in line with dlss which is not a bad thing, but its still something we cant hand wave away just because we dont like it.

 

[SlimySnake]

There is redundancy in those modes like in many pro games.
leaving old modes there and having few pro modes is redundant.
Performance pro mode destroys the normal performance mode and there is no reason to even show it on the pro. Nobody going to feel 90 to 110fps in random spots. It's not even consistent
Same with stellar blade. They left all the old crap modes in there... and added 2 modes + hfr toggle. Why?
HFR toggle should be hidden and automatic based on vrr+120hz status from ps5 system settings. Old modes should GO AWAY and pro max mode is stupid. Looks and runs worse than PRO mode.

I don't like redundancy. Console gaming is all about lowering redundancy so I don't waste my attention looking at modes.

If anything, to make all the weirdos happy, there could be a system toggle in ps5 system settings "disable pro", so games recognize it as normal ps5.

hell no. gt7 runs at almost 100 fps at native 4k hitting 120 fps at times. the new fidelity mode would likely top out at 60 fps. fuck that. same goes for tlou2. the native 4k mode already ran at 60 fps 90% of the time dropping to 45 fps during some open world levels. on the pro, the extra horsepower will run gt7 at 120 fps locked and tlou2 at 60 fps locked at native 4k. no need for PSSR.

pro is about giving players choices. no dumb casual is buying this product. id argue that most gamers in general are smart and educated enough to know the difference between these toggles.

i do like the suggestion of a pro setting in the console that defaults in the pro mode when you boot up. they already have it for performance/fidelity modes in the base ps5 so adding a pro setting for pro consoles should be a no brainer.

 

[DirtyBast8rd]

Looking forward to this on pc. Il play it ultrawide 100+ fps and 4k like I did the first game.

Cry harder Pcmr. Oh wait, got confused.

 

[rofif]

hell no. gt7 runs at almost 100 fps at native 4k hitting 120 fps at times. the new fidelity mode would likely top out at 60 fps. fuck that. same goes for tlou2. the native 4k mode already ran at 60 fps 90% of the time dropping to 45 fps during some open world levels. on the pro, the extra horsepower will run gt7 at 120 fps locked and tlou2 at 60 fps locked at native 4k. no need for PSSR.

pro is about giving players choices. no dumb casual is buying this product. id argue that most gamers in general are smart and educated enough to know the difference between these toggles.

i do like the suggestion of a pro setting in the console that defaults in the pro mode when you boot up. they already have it for performance/fidelity modes in the base ps5 so adding a pro setting for pro consoles should be a no brainer.

Pssr makes all non pssr choices redundant tho. Native 4k will never look better than pssr if it’s anything like dlss. Unless the introduce psaa at native lol.

As for the choices, they don’t give us tools and info to make that decision. New modes are supposed to look better and run faster

edit: Like here


PS5 Pro PSSR performance is 95fps and PRO TAA mode is 112fps.
This is indistinguishable and should not be a choice on console. It's a console. Those who want this granular choice should just stick with pc.
Dev should make 1-2 optimal modes and that's it. Offer fps unlock for vrr screens at 120hz and that's it. In case of pro, 1 mode should be achievable.
Nobody themselves without DF or anal guy over here, is going to be able to realize that one mode is 97 fps and other is 112 fps.... And both look very similar. It makes no sense to have these redundant choices.
And to be honest, I had the same feelings towards pc settings for a long time. Devs feel like they must offer LOW-Ultra settings... and in many games and many of those settings, that's no fps difference of 2fps between low and ultra... with no effect on graphics or framerate. Or sometimes just making graphics worse but not helping performance. These settings should not be there. It's like having 15 more buttons in a car that don't do anything.
Starting up a pc game and deciding which setting in this particular game is the heavy weight was always a thing I did. It's a waste of everyone's time

 

[Fafalada]

here you can see the ps5 gpu outperform the pro gpu in terms of framerate. we know for a fact that the ps5 pro gpu is faster, has more vram bandwidth so if the framerate is lower despite the internal resolution being identical to the base ps5, we can assume that the difference is exclusively due to the pssr reconstruction costs.

I have no way of knowing how the two modes compare. At minimum - we'd need to see PS5 Pro run the PS5 mode to see where 'that' GPU lands (and if it landed on say - 111fps in the same screenshot, we'd know it's CPU limited). And that would still not tell us if the modes are actually equvialent outside of PSSR - that's just a 'trust me bro' statement right now.
So no - as far as test methodologies go - this one isn't it. Other than stating 'yes there's a cost' - which we knew already.

sony's docs said a lot of things. we dont know what framerate they were looking at.

Sony's doc explicitly stated time-cost of the upscaler from 1080p->4k.
Framerate has no relevance to any of this - I only referred to it because this thread was going off about % increases based on framerate.

this is only one data point. other games might have a smaller hit in line with the 2ms sony outlined.

Sony outlined the end-2-end compute cost - either because it's relatively frame-invariant (most screen space algorithms are) or because it's a good average.
The variance beyond that is less likely computational and more to do with things like async-execution etc(which I mentioned) but that's not something that can be predicted for different games, you'd need to measure it.
But for the purposes of this thread - I don't know what that measure is, there are no numbers for it - maybe you can run the tests on both Pro modes and we can at least have something to work from.

 

[vkbest]

PS5 Pro PSSR performance is 95fps and PRO TAA mode is 112fps.
This is indistinguishable and should not be a choice on console. It's a console. Those who want this granular choice should just stick with pc.
Dev should make 1-2 optimal modes and that's it. Offer fps unlock for vrr screens at 120hz and that's it. In case of pro, 1 mode should be achievable.

That is stupid. We have plenty of examples of developers doing bad decisions such ass FF XVI or elder ring prioritizing resolution over frame rate in performance mode. Or Wukong 45fps cap.

Alan wake 2 should run better than PS5 version in performance mode, but they decided to make higher settings and now the game is running worse than PS5 version despite they are using the same resolution.

 

[rofif]

That is stupid. We have plenty of examples of developers doing bad decisions such ass FF XVI or elder ring prioritizing resolution over frame rate in performance mode. Or Wukong 45fps cap.

Alan wake 2 should run better than PS5 version in performance mode, but they decided to make higher settings and now the game is running worse than PS5 version despite they are using the same resolution.

No way in hell I would want to go back to AW2 horrible fsr2 modes even if they ran 3x faster.

 

[deriks]

Maybe...but why waste their time when there's already FSR, DLSS, XeSS, and TAA? Adding a 5th solution won't be providing much.

But PSSR is doing supposedly better than those. Could be a gake changer

I think there's too many options to have like a minor percentage with gain to everyone. We need like three with better differences, and I can see PSSR being one of them

 

[twilo99]

DF spoke about thid in new direct weekly and they have much better game to judge this by - GOW Ragnarok patch whoch dropped. You can select taa or pssr with unlocked fps, so they were able to get idea of cost of pssr
tldr- around 2ms latency cost, pssr resolves much better image. There isnfps difference but a lot of paces you are like 70-100range and with vrr it doesnt really have big difference in feel

VRR is one of the most significant advances is gaming in the past 20 years.

 

[twilo99]

also interesting to note is that apparently in the code of the UE5 plugin for PSSR, PSSR uses nearly the exact same names and terminology for its variables and settings as those that previously were only used by XeSS.

Intriguing.

Do we have a source for that?

 

[winjer]

One thing to consider about these upscalers, is that they require LODs and MipMap bias to be adjusted, as to show correctly in the final upscaled image.
This means that when a game is upscaling a game, the cost is not just the upscaling, but also higher quality assets and textures.

 

[SlimySnake]

One thing to consider about these upscalers, is that they require LODs and MipMap bias to be adjusted, as to show correctly in the final upscaled image.
This means that when a game is upscaling a game, the cost is not just the upscaling, but also higher quality assets and textures.

this is true and i alluded to this in the OP.

. In some cases, it dips around 10 fps below the ps5 performance mode like screenshotted below but that could also be a vram bottleneck since the textures and other effects are still rendered at 4k.

All of this started in the pro leak threads because for some reason people didnt want to admit that 1440p native vs 1440p reconstructed was not going to be costly. people latched on to the 2ms figure from sony ignoring real world examples we can test at home today. All I was trying to point out was that the a big portion of the 45% gpu was going to go into reconstruction leaving few gpu resources to increase frames or settings in games that previously didnt have a reconstruction solution. that's it. And for that people jumped down my throat, called me names, and just made those threads a miserable fucking experience. Good to see basic fucking facts proven by actual game data.

 

[winjer]

this is true and i alluded to this in the OP.

All of this started in the pro leak threads because for some reason people didnt want to admit that 1440p native vs 1440p reconstructed was not going to be costly. people latched on to the 2ms figure from sony ignoring real world examples we can test at home today. All I was trying to point out was that the a big portion of the 45% gpu was going to go into reconstruction leaving few gpu resources to increase frames or settings in games that previously didnt have a reconstruction solution. that's it. And for that people jumped down my throat, called me names, and just made those threads a miserable fucking experience. Good to see basic fucking facts proven by actual game data.

I admit I was confused with what you meant by " textures and other effects are still rendered at 4k."
Now I get it.

 

[kevboard]

Intriguing.

Do we have a source for that?

someone posted a video earlier where it's shown I think.

 

[twilo99]

this is true and i alluded to this in the OP.



All of this started in the pro leak threads because for some reason people didnt want to admit that 1440p native vs 1440p reconstructed was not going to be costly. people latched on to the 2ms figure from sony ignoring real world examples we can test at home today. All I was trying to point out was that the a big portion of the 45% gpu was going to go into reconstruction leaving few gpu resources to increase frames or settings in games that previously didnt have a reconstruction solution. that's it. And for that people jumped down my throat, called me names, and just made those threads a miserable fucking experience. Good to see basic fucking facts proven by actual game data.

Maybe I am missing something again but wouldn’t turning pssr off free up that extra GPU power for “traditional” rasterization?

 

[winjer]

Maybe I am missing something again but wouldn’t turning pssr off free up that extra GPU power for “traditional” rasterization?

It depends. If the Pro is using WMMA, this means when shader units are being used to run the ML pass, they can't be used for other things. So turning off PSSR, would free shader units to do compute.

 

[vkbest]

No way in hell I would want to go back to AW2 horrible fsr2 modes even if they ran 3x faster.

Who said FSR? I'm talking about higher settings. Anyway, what is the point to play a performance mode is not even smooth because bad frame rate? I don’t think most people have a VRR TV

Also, GOW PSSR looks worse than TAA and the last run faster

 

[Codeblew]

It's fine at 30 and 60fps. At 120fps where there is only 8.33ms per frame, it's too high. It could be why they use TAAU in F1 24 instead of PSSR in the 120Hz mode.

Games should be working on the next frame while the PSSR + video out is doing its thing. It's not like everything stops waiting for that 2ms to end.

 

[Gaiff]

But PSSR is doing supposedly better than those. Could be a gake changer

I think there's too many options to have like a minor percentage with gain to everyone. We need like three with better differences, and I can see PSSR being one of them

PSSR is comparable to XeSs and DLSS, not necessarily better. According to John from DF who tested GOWR on both the Pro and PC, DLSS is still superior.

The point still stands that adding PSSR wouldn’t do much since DLSS is already as good and XeSS is often comparable, so it wouldn’t move the needle much.

 

[SlimySnake]

Maybe I am missing something again but wouldn’t turning pssr off free up that extra GPU power for “traditional” rasterization?

yes. if you have a ps5 pro, you can run the game in the regular performance mode and it should hit 120 fps seeing as how it sits around 80-90 fps on the base ps5.

 

[SlimySnake]

DF spoke about thid in new direct weekly and they have much better game to judge this by - GOW Ragnarok patch whoch dropped. You can select taa or pssr with unlocked fps, so they were able to get idea of cost of pssr
tldr- around 2ms latency cost, pssr resolves much better image. There isnfps difference but a lot of paces you are like 70-100range and with vrr it doesnt really have big difference in feel

i just watched this segment. they say its around 2ms compared to TAA upscaling which has its own cost and is not free either. TLOU2 is native 1440p and does not use any upscaling. your tv upscales the 1440p image to run at 4k.

they also said that the difference in frames could be 10 or 20 fps depending on the scene. And for some bizarre reason they used a cpu bound scene to determine the cost of pssr on the gpu which honestly makes no sense. regardless, 2ms is basically the best case scenario here just like the 45% figure they gave out in the leaks. i wish they had done some tests on the non-pro version to see just how well it runs on the pro to get a better feel for the ps5 pro gpu performance.

timestamped:

 

[twilo99]

someone posted a video earlier where it's shown I think.

I can't find it but it would make sense for them to partner up with someone on this, I just assumed that they would do this with AMD and not Intel

 

[kevboard]

I can't find it but it would make sense for them to partner up with someone on this, I just assumed that they would do this with AMD and not Intel

Intel already has a working and good looking ML reconstruction method that works on AMD hardware, while AMD doesn't.

which is hilarious and sad at the same time. but it would make sense to fork off of an already working one instead of waiting for AMD to get off their asses and finally catch up