Support NeoGAF

[Mr.Allan]

Honestly I'd like to see some pics from chapter 4. when you see tram moving fast in a distance. on fsr4 it looks super blurry. Not that it matters, just how pssr works there. probably the same as it's just watered version of fsr4.

 

[Salty Pickle]

So it's the same issue as RE9? when DoF caused massive aliasing? I never use it in any game but watched video and it looked insanely bad.

"insanely bad", really?

 

[Salty Pickle]

Honestly I'd like to see some pics from chapter 4. when you see tram moving fast in a distance. on fsr4 it looks super blurry. Not that it matters, just how pssr works there. probably the same as it's just watered version of fsr4.

In what way is pssr2 "watered down"? Just curious.

 

[Bojji]

Honestly I'd like to see some pics from chapter 4. when you see tram moving fast in a distance. on fsr4 it looks super blurry. Not that it matters, just how pssr works there. probably the same as it's just watered version of fsr4.

In what way is pssr2 "watered down"? Just curious.

FSR 4.1 and PSSR2 are more or less the exact same thing with almost identical results.

 

[Salty Pickle]

FSR 4.1 and PSSR2 are more or less the exact same thing with almost identical results.

Yeah that's what I thought.

 

[viveks86]

I get that but PSSR is running on INT8 originally written for FP8 I presume, therefore the same could be done with RR. However doubtful there's power left over for it in the Pro or if Sony thinks it's worth it since PS6 most likely will have it on FP8.

RR is too expensive to run on Pro even if they can port it to INT 8. Given the timing of it all, they might use it as a differentiator for next gen and not bother with it for the pro.

By the time next gen comes around, it may even be done with FP4, to recover some of that frame time.

 

[Mr.Allan]

In what way is pssr2 "watered down"? Just curious.

just look at pics.

it's from fsr 4 quality and there's a clean image, gpu on passive mode while doing 120fps, or 200 unlocked. I'm not saying it looks bad. just that there's something off with it in this game. not that it matters so don't lose sleep over it!

 

[Xeela_sunston]

If 800 is a scam then what the hell are Nvidia gpu:s? What are you comparing to?

Overpriced junk that lose on VRAM & Rastering compared to a 7900xtx/9070xt, Hence they need to brag about PT rendering. It funny they cry about FSR 4 making RTX on RDNA 3 actually painless, Yet they need DLSS to do Ultra with RT/4K textures because 12 ~ 16GB VRAM not enough at 1440p/4K.

 

[Salty Pickle]

just look at pics.

it's from fsr 4 quality and there's a clean image, gpu on passive mode while doing 120fps, or 200 unlocked. I'm not saying it looks bad. just that there's something off with it in this game. not that it matters so don't lose sleep over it!

So 864p internal to 4k I presume?

 

[Mr.Allan]

1440p I have 32 panels 70cm from my face. 4k is a waste for me. but yes. should be the same base res on quality. yet looks vast superior compared to sonystation toy.

 

[Hyper_hyper]

Posts like this is why age verification is important for forum registration.

Still not enough Phil Spencer's balls in your throat. Did Phil Spencer asked you for your age?

Then, I repeat, I asked (because it is a question) which game in the last 2 years has received a version or port on XSX that could be considered "BAD" compared to PS5... You answered only KCD2 (probably after having reviewed a huge list of games released in the last 1.5-2 years).

You haven't asked. You've mentioned you can't recall. That's not a question. I've just me mentioned KCD2 as a reminder to your brainwashed Xbox brain. Of course you can't recall which game performed better on PS5 because you've avoided that Digital Foundry threads.

LOL x2

The answer perfectly describes you...

And it all stems from a simple:

"The PS5 Pro and XSX versions are improved compared to the demo."

And then they start talking about others playing the victim and having an inferiority complex

Bye

LOL. You bragged how XSX version surpassed PS5 demo, c'mon. That's idiotic. Still whining how everyone attack Xbox, huh. Yep, you playing a victim card. You,Xslops, are playing victim card for more that a decade. Your brains are fried. Talking to me about complexity.

 

[Karim]

just look at pics.

it's from fsr 4 quality and there's a clean image, gpu on passive mode while doing 120fps, or 200 unlocked. I'm not saying it looks bad. just that there's something off with it in this game. not that it matters so don't lose sleep over it!

It's the DOF very low buffer upscaled in 4k. Just my guess. I remind MLAA cause similar effects on ps3 games in different edges.

 

[Karim]

LOL. That's called "asking a question to the air" and expecting an answer... and in fact, your reaction was an answer.
It's incredible how some of you need to overcomplicate things because you have trouble admitting things.

Then, I repeat, I asked (because it is a question) which game in the last 2 years has received a version or port on XSX that could be considered "BAD" compared to PS5... You answered only KCD2 (probably after having reviewed a huge list of games released in the last 1.5-2 years).

My question then was, if you consider the XSX version of KCD2 "bad" compared to the PS5 version... what adjective would you use to describe a prominent list of games released during that time where the XSX version had a clear advantage (up to 15fps, higher resolution, better graphics) over the PS5 version? Simple. Remember, we're in the sixth year of this generation, with the XSX possibly being the last version to receive attention and optimization time, and many of you were telling that by now the situation should be different, with the XSX even ceasing to receive games or suffering from poor optimization and performance compared to the PS5...
What's shocking is that those same people who told you this don't seem to find the situation relevant.


LOL x2

The answer perfectly describes you...

And it all stems from a simple:

"The PS5 Pro and XSX versions are improved compared to the demo."

And then they start talking about others playing the victim and having an inferiority complex

Bye

I thought exasperate any kinda of competition or try to evindence any kinda of superiority at any chance that was having an inferiority complex but sure whatever you say.

 

[Mr.Allan]

It's the DOF very low buffer upscaled in 4k. Just my guess. I remind MLAA cause similar effects on ps3 games in different edges.

yeah. found video from Daniel Owen
you can fast forward and it shows the difference. so DoF was the problem all together. It surely applies here as well.

 

[Karim]

Still doesn't make much sense, sorry. A neural network is a neural network, the rest is mostly standard - as can be seen on PC by the simplicity of DLL upgrades or plugin solutions like OptiScaler.


It matters for the h/w which it runs on, you can't run an FP8 based NN on INT8 h/w.

Can't care less if for you doesn't make sense and frankly this repeated attitude as know-it-all, it's childish and unnecessary, you are not a professor in an academic convention. Use the same neural network doesn't mean automatically FSR4 and PSSR2 are the same entity, it's really a simplistic generalisations which omit a lot of other work behind make on purpose just to continue to spread your bizzare convintion about sony who would want to steal the merit to AMD for unknown reasons.

 

[Bojji]

Can't care less if for you doesn't make sense and frankly this repeated attitude as know-it-all, it's childish and unnecessary, you are not a professor in an academic convention. Use the same neural network doesn't mean automatically FSR4 and PSSR2 are the same entity it's really a simplistic generalisations which omit a lot of other work behind.

We have Mark Cerny confirming that FSR4.1 and PSSR2 use the same model.

Upgraded PSSR Uses INT8 FSR 4 Implementation That AMD Denied Older RDNA 3 GPUs

Shortly after AMD released FSR 4, claiming that the tech was exclusive to the latest RDNA 4 GPUs, the company seemingly accidentally published the libraries that make up the backbone of the tech, revealing that there may have been a version of FSR 4 planned for RDNA 3 and RDNA 2 GPUs. While this...

www.techpowerup.com

With different training data, end results are almost the same.

 

[Karim]

We have Mark Cerny confirming that FSR4.1 and PSSR2 use the same model.

Upgraded PSSR Uses INT8 FSR 4 Implementation That AMD Denied Older RDNA 3 GPUs

Shortly after AMD released FSR 4, claiming that the tech was exclusive to the latest RDNA 4 GPUs, the company seemingly accidentally published the libraries that make up the backbone of the tech, revealing that there may have been a version of FSR 4 planned for RDNA 3 and RDNA 2 GPUs. While this...

www.techpowerup.com

With different training data, end results are almost the same.

I haven't denied that in my post.

 

[adamsapple]

We have Mark Cerny confirming that FSR4.1 and PSSR2 use the same model.

Upgraded PSSR Uses INT8 FSR 4 Implementation That AMD Denied Older RDNA 3 GPUs

Shortly after AMD released FSR 4, claiming that the tech was exclusive to the latest RDNA 4 GPUs, the company seemingly accidentally published the libraries that make up the backbone of the tech, revealing that there may have been a version of FSR 4 planned for RDNA 3 and RDNA 2 GPUs. While this...

www.techpowerup.com

With different training data, end results are almost the same.

Careful now Bojji .. that dude knows what TAA looks like, he has unparalleled tech knowledge

 

[Bojji]

I haven't deny that in my post.

But they are using the same model, differences are academic. Same will be true for PSSR3/FSR5 and whatever Xbox helix will use.

 

[Karim]

But they are using the same model, differences are academic. Same will be true for PSSR3/FSR5 and whatever Xbox helix will use.

I mean...probably you missed what was my point. I just continue to repeat to him say sony just steal the merit to AMD because PSSR2 is the same thing of FSR4, it's childish other than simplistic and unfair, because there are difference in other factors. It's practically that the deep argumentation sustained by him and I really don't get it this obstinacy to continue to spread this absurd conjecture.

 

[dgrdsv]

I get that but PSSR is running on INT8 originally written for FP8 I presume, therefore the same could be done with RR. However doubtful there's power left over for it in the Pro or if Sony thinks it's worth it since PS6 most likely will have it on FP8.

PSSR is using INT8 h/w to run the NN which was originally designed to run on FP8, this does have consequences and it does result in lower IQ.
RR is a much bigger NN which require more throughput and likely higher precision to run well. Looking at how AMD's own efforts in adding it to games ending up now I wouldn't expect the Pro to be able to run it at any ms usable in real time.

Use the same neural network doesn't mean automatically FSR4 and PSSR2 are the same entity

Yes, it does.

it's really a simplistic generalisations which omit a lot of other work behind make on purpose just to continue to spread your bizzare convintion about sony who would want to steal the merit to AMD for unknown reasons

This "lot of other work" was done by AMD who ported the NN to run on INT8 math.
If I had to guess Sony paid them for that (aka "the Amethyst Project" or w/e it's called) and one clause in this was to keep it exclusive to PS5Pro - which is why the leaked version of FSR4 INT8 never made it into release.
There is nothing "bizarre" in being fed up by Sony's/Cerny's blatant lies and pure distilled marketing being presented as some unbelievable tech innovations.

 

[MisterXDTV]

PSSR is using INT8 h/w to run the NN which was originally designed to run on FP8, this does have consequences and it does result in lower IQ.
RR is a much bigger NN which require more throughput and likely higher precision to run well. Looking at how AMD's own efforts in adding it to games ending up now I wouldn't expect the Pro to be able to run it at any ms usable in real time.


Yes, it does.


This "lot of other work" was done by AMD who ported the NN to run on INT8 math.
If I had to guess Sony paid them for that (aka "the Amethyst Project" or w/e it's called) and one clause in this was to keep it exclusive to PS5Pro - which is why the leaked version of FSR4 INT8 never made it into release.
There is nothing "bizarre" in being fed up by Sony's/Cerny's blatant lies and pure distilled marketing being presented as some unbelievable tech innovations.

 

[Imtjnotu]

PSSR is using INT8 h/w to run the NN which was originally designed to run on FP8, this does have consequences and it does result in lower IQ.
RR is a much bigger NN which require more throughput and likely higher precision to run well. Looking at how AMD's own efforts in adding it to games ending up now I wouldn't expect the Pro to be able to run it at any ms usable in real time.


Yes, it does.


This "lot of other work" was done by AMD who ported the NN to run on INT8 math.
If I had to guess Sony paid them for that (aka "the Amethyst Project" or w/e it's called) and one clause in this was to keep it exclusive to PS5Pro - which is why the leaked version of FSR4 INT8 never made it into release.
There is nothing "bizarre" in being fed up by Sony's/Cerny's blatant lies and pure distilled marketing being presented as some unbelievable tech innovations.

 

[rodrigolfp]

 

[Puscifer]

Why is the pro resolution reduced? How bizarre.

 

[Luipadre]

Why is the pro resolution reduced? How bizarre.

To pack in more RT features and PSSR2 doesnt need high ress to get good image quality on 4k sets so they can reduce resolution and get more stable framerate and higher quality visuals while maintaining a clean image

 

[Bojji]

To pack in more RT features and PSSR2 doesnt need high ress to get good image quality on 4k sets so they can reduce resolution and get more stable framerate and higher quality visuals while maintaining a clean image

I think RT is the same between PS5 and Pro version. But it looks worse on Pro because internal resolution of RT is lower.

 

[Prodyhig]

I feel nostalgic for when technological comparisons were made between consoles: PlayStation 3 vs. Xbox 360 / PlayStation 4 vs. Xbox One...

Now, the PlayStation 5 Pro is relegated to being compared with $3,000 hardware.

 

[Bojji]

I feel nostalgic for when technological comparisons were made between consoles: PlayStation 3 vs. Xbox 360 / PlayStation 4 vs. Xbox One...

Now, the PlayStation 5 Pro is relegated to being compared with $3,000 hardware.

WTF? You can enable PT on any nvidia hardware made after 2018.

Depending on your screen resolution - you can play it on 450$ GPU.

For 4k with 864p internal (so like Pro) you need ~5070ti level of power for fixed 60fps.

 

[Mr.Allan]

WTF? You can enable PT on any nvidia hardware made after 2018.

Depending on your screen resolution - you can play it on 450$ GPU.

For 4k with 864p internal (so like Pro) you need ~5070ti level of power for fixed 60fps.

That's for max settings. Also I don't remember any good panels having fixed 60hz cap so why are you bringing this up? it's not 2005 anymore.

 

[Bojji]

That's for max settings. Also I don't remember any good panels having fixed 60hz cap so why are you bringing this up? it's not 2005 anymore.

What are you even talking about? It's path tracing, I show what is required to run it, it's certainly not 3000$ hardware.

Who gives a shit about 60Hz when you have Gsync/VRR?

 

[Mr.Allan]

I mean it's Max settings + path tracing. on PC we have these things called options and you can just bring down unnecesary stuff to up framerate by 50%. People play all kind of games with PT on with 4060ti class gpu over 100fps, just not maxed everything.
Also clearly you give a shit as most of your screenshots show locked 60fps which is a slide show.

 

[Bojji]

I mean it's Max settings + path tracing. on PC we have these things called options and you can just bring down unnecesary stuff to up framerate by 50%. People play all kind of games with PT on with 4060ti class gpu over 100fps, just not maxed everything.
Also clearly you give a shit as most of your screenshots show locked 60fps which is a slide show.

Of course you can play with raster or raster+RT and have ~100FPS or more. But this comparison was about PT, that makes the game look significantly better, and you don't need 5090 to play it like that, that's the usual hyperbole from console owners.

To me 60fps is absolutely playable, but i play with a controller. Of course 100+FPS is nicer and more responsive, but I won't give a shit about it if it requires to cut important graphical features.

 

[Mr.Allan]

Of course you can play with raster or raster+RT and have ~100FPS or more. But this comparison was about PT, that makes the game look significantly better, and you don't need 5090 to play it like that, that's the usual hyperbole from console owners.

To me 60fps is absolutely playable, but i play with a controller. Of course 100+FPS is nicer and more responsive, but I won't give a shit about it if it requires to cut important graphical features.

Sorry if I wasn't clear. that picture says MAX SETTINGS + PT. you said you need 5070ti to play with PT 60fps which is not true. you can do low settings and pt and have more than 60fps on 4060ti. It's a bit nitpicking but these benchmarks are a rabbit hole if you don't understand what's what. Sorry again.

 

[sncvsrtoip]

For 4k with 864p internal (so like Pro) you need ~5070ti level of power for fixed 60fps.

thats probably what we will get with ps6

 

[Bojji]

Sorry if I wasn't clear. that picture says MAX SETTINGS + PT. you said you need 5070ti to play with PT 60fps which is not true. you can do low settings and pt and have more than 60fps on 4060ti. It's a bit nitpicking but these benchmarks are a rabbit hole if you don't understand what's what. Sorry again.

Low and max settings won't do much difference, PT is 90% responsible for the final framerate.

 

[Karim]

PSSR is using INT8 h/w to run the NN which was originally designed to run on FP8, this does have consequences and it does result in lower IQ.
RR is a much bigger NN which require more throughput and likely higher precision to run well. Looking at how AMD's own efforts in adding it to games ending up now I wouldn't expect the Pro to be able to run it at any ms usable in real time.


Yes, it does.


This "lot of other work" was done by AMD who ported the NN to run on INT8 math.
If I had to guess Sony paid them for that (aka "the Amethyst Project" or w/e it's called) and one clause in this was to keep it exclusive to PS5Pro - which is why the leaked version of FSR4 INT8 never made it into release.
There is nothing "bizarre" in being fed up by Sony's/Cerny's blatant lies and pure distilled marketing being presented as some unbelievable tech innovations.

Sony never talked of unbelievable tech innovation when presented the PSSR (not like Nvidia when showed DLSS5 cough cough ) and they just dare to say pssr and fsr4 are 2 different algorithm with the same neural network, so wow what hyperbolic marketing spin Anyway would be great to have any evidence outside your personal conjectures.
I missed the part where you said Amethyst project is basically a paid exclusive for FSR4 tech. You have reached a new high here. Fuck.

 

[Mr.Allan]

So MLID is right? 30fps will become 300fps on ps6? pls man.

 

[Mr.Phoenix]

I mean it's Max settings + path tracing. on PC we have these things called options and you can just bring down unnecesary stuff to up framerate by 50%. People play all kind of games with PT on with 4060ti class gpu over 100fps, just not maxed everything.
Also clearly you give a shit as most of your screenshots show locked 60fps which is a slide show.

Sorry... 60fps is a slideshow?

Ok, I'll just take this as one of those things that some people do or say that are just completely unrelatable.

 

[Mr.Allan]

Sorry... 60fps is a slideshow?

Ok, I'll just take this as one of those things that some people do or say that are just completely unrelatable.

only once you got used to 180fps. I always forget to add "to me". Sorry.

 

[Mr.Phoenix]

only once you're get used to 180fps. I always forget to add "to me". Sorry.

Fair enough, for me, once I get past 60fps, the difference is no longer obvious to me, barely noticeable if anything. But that's between 60 and 120... so I would say going from 30 to 60 is like night and day, going from 60 to 100/120 is hardly noticeable, if I may be honest, I think I placebo myself into thinking I see it. I have never played games at 160fps+ so I can't speak for anything higher.

 

[Mr.Allan]

Fair enough, for me, once I get past 60fps, the difference is no longer obvious to me, barely noticeable if anything. But that's between 60 and 120... so I would say going from 30 to 60 is like night and day, going from 60 to 100/120 is hardly noticeable, if I may be honest, I think I placebo myself into thinking I see it. I have never played games at 160fps+ so I can't speak for anything higher.

my lowest bar these days is 80fps. it makes huge difference. I played half of my life games in 60fps and it was fine. I knew that upgrading to 180hz panels would cost me 1000, in the end costed 2500. but there's really no coming back. I dunno why people are so fixated with 60fps on pc games while you cannot even buy normal 60hz panels these days. Well.

 

[Skifi28]

Fair enough, for me, once I get past 60fps, the difference is no longer obvious to me, barely noticeable if anything. But that's between 60 and 120... so I would say going from 30 to 60 is like night and day, going from 60 to 100/120 is hardly noticeable, if I may be honest, I think I placebo myself into thinking I see it. I have never played games at 160fps+ so I can't speak for anything higher.

Can't say I see much difference between 120 and 144, so I doubt 160/180 is gonna be a significant step up. While I can see/notice the difference from 60 to 120, it's not that big of a deal. Maybe in an FPS, anything third person and I don't care.

*Shrug*

 

[Mr.Phoenix]

my lowest bar these days is 80fps. it makes huge difference. I played half of my life games in 60fps and it was fine. I knew that upgrading to 180hz panels would cost me 1000, in the end costed 2500. but there's really no coming back. I dunno why people are so fixated with 60fps on pc games while you cannot even buy normal 60hz panels these days. Well.

I don't play my PC games on a monitor. I have an HTPC (in a Silverstone GD11 case), connected to my living room TV. So the best I would be getting from my TV is 120fps anyway.

 

[Mr.Allan]

I don't play my PC games on a monitor. I have an HTPC (in a Silverstone GD11 case), connected to my living room TV. So the best I would be getting from my TV is 120fps anyway.

Sorry for Offtopic. but I never said monitors. only panels. Which for me is tvs and monitors. EOT as I derailed topic too much.

 

[FalconPunch]

WTF? You can enable PT on any nvidia hardware made after 2018.

Depending on your screen resolution - you can play it on 450$ GPU.

For 4k with 864p internal (so like Pro) you need ~5070ti level of power for fixed 60fps.

The 5060ti 16gb doesn't even give you 60fps at 1080p using DLSS quality. However while you can enable pt on any card, you shouldn't.

Using any DLSS other than DLAA at 1080p is already a fail for PT as far as I'm concerned. The minimum base performance for viable path tracing is to be able to path trace at 1080 >60fps native.

That gives DLSS good data for excellent image quality.

 

[Bojji]

The 5060ti 16gb doesn't even give you 60fps at 1080p using DLSS quality. However while you can enable pt on any card, you shouldn't.

Using any DLSS other than DLAA at 1080p is already a fail for PT as far as I'm concerned. The minimum base performance for viable path tracing is to be able to path trace at 1080 >60fps native.

That gives DLSS good data for excellent image quality.

That graph is bugged, you can see 16GB version of 5060ti losing to 8GB version and regular 5060.

What's wrong with quality DLSS at 1080p? It is with Ray Reconstruction as well.

 

[saintjules]

 

[SKYF@ll]

PSSR is using INT8 h/w to run the NN which was originally designed to run on FP8, this does have consequences and it does result in lower IQ.
RR is a much bigger NN which require more throughput and likely higher precision to run well. Looking at how AMD's own efforts in adding it to games ending up now I wouldn't expect the Pro to be able to run it at any ms usable in real time.


Yes, it does.


This "lot of other work" was done by AMD who ported the NN to run on INT8 math.
If I had to guess Sony paid them for that (aka "the Amethyst Project" or w/e it's called) and one clause in this was to keep it exclusive to PS5Pro - which is why the leaked version of FSR4 INT8 never made it into release.
There is nothing "bizarre" in being fed up by Sony's/Cerny's blatant lies and pure distilled marketing being presented as some unbelievable tech innovations.

PSSR2 primarily uses INT8 inference, but unlike a PC-side INT8 implementation of something like FSR4.0, it does not inherently result in lower image quality compared to FSR4.1. In fact, INT8 inference is a well-suited and efficient choice for real-time game upscaling workloads.

For stages that require higher numerical precision, FP8 can be advantageous, but on PS5 Pro this role is likely handled with FP16 instead. The underlying algorithmic structure is essentially the same; only the precision format used in certain layers differs.

 

[adamsapple]

The frequent dips out of VRR shows why they didn't do fixed 1440p but a DRS solution would have worked fine, the game has enough headroom for 1440p in quieter moments.