Support NeoGAF

[ShadyAcshuns]

You mean DLSS? It is proprietary to Nvidia

I realize PSSR is also proprietary, but hasn't there been issues with it compared to that other one? Isn't it FSR1?

 

[Topher]

I realize PSSR is also proprietary, but hasn't there been issues with it compared to that other one? Isn't it FSR1?

PSSR has been hit and miss depending on the developer. I would imagine PSSR is closely related to FSR, but we don't really know. Either way, PSSR 2 will hopefully be a big step up.

 

[Danjin44]

I'm no tech expert but does this "PSSR 2.0" coming as system update or new version PS5 pro with new PSSR?

 

[Three]

I'm no tech expert but does this "PSSR 2.0" coming as system update or new version PS5 pro with new PSSR?

System update obviously. There isn't a PS5 Pro 2. That would be bigger news.

 

[RSMan]

What are the chances that the PSSR update will also improve PS4 games? Because the current PS4 enhancement feature on the Pro feels pretty useless.

 

[Gaiff]

What are the chances that the PSSR update will also improve PS4 games? Because the current PS4 enhancement feature on the Pro feels pretty useless.

None? As far as I'm aware, no PS4 game uses PSSR, unless there is a PS5 version, at which point it's a PS5 game.

 

[Gitaroo]

What are the chances that the PSSR update will also improve PS4 games? Because the current PS4 enhancement feature on the Pro feels pretty useless.

from the translated x qoute " Upscales low-res classic games to high resolution", probably if the original post was even real. We have enter the era where monitor has built in NPU, LG 39GX950B that can do ML base up scaling now to any source below 4K. No reason why PS5 pro cannot do it beside Sony lackluster support which is usually the case. MS also has AutoSR around for a year for low power SoC like snapdragon X which would be a perfect for upscaling any older games.

 

[skit_data]

If it's a notable difference I might end up getting a Pro after all, especially if we get more definitive proof Sony might end up delaying the release of PS6 due to the market conditions.
But PSSR 2.0 will have to be way more consistent in terms of quality and implemetation than the first iteration because as of now I'm rather happy I didn't get a Pro which is wierd to say the least.

 

[Jesb]

What was the point even of PSSR if other related tech is available.

 

[Duchess]

What was the point even of PSSR if other related tech is available.

As I understand it, AMD was still working on the specs when the PS5 Pro was releasing, and so Sony decided to have a crack at it themselves.

Now that AMD has finished their work, Sony's going to use their implementation.

Something like that.

 

[Traxtech]

What are the chances that the PSSR update will also improve PS4 games? Because the current PS4 enhancement feature on the Pro feels pretty useless.

It's useless in some games, but I think screen size plays into it alot. On my 42 oled it's meh, on my 77 oled the difference in some games is very apparent

 

[Physiognomonics]

As I understand it, AMD was still working on the specs when the PS5 Pro was releasing, and so Sony decided to have a crack at it themselves.

Now that AMD has finished their work, Sony's going to use their implementation.

Something like that.

Cerny probably picked INT8 because it's the most efficient way to do inference but algo need to be customized fo it. AMD solution is more general purpose.

Tesla also massively favor INT8 for their inference chips for the same reasons.

 

[Toots]

Is it possible the PSSR 2.0 update rolls out between now and march ?
We don't have any info about it, wouldn't sony communicate more on it if that was the case ?

 

[Radical_3d]

I swear: every time this gets bumped with no actual info it sparks a murderous humor in me.

 

[Forth]

Hopefully they use Marathon as a promotional for this new PSSR.

 

[Panajev2001a]

As I understand it, AMD was still working on the specs when the PS5 Pro was releasing, and so Sony decided to have a crack at it themselves.

Now that AMD has finished their work, Sony's going to use their implementation.

Something like that.

Sony and AMD are working on the solution together as partners, Sony had their hands in making FSR4 as good as it is.

 

[Fake]

I'm no tech expert but does this "PSSR 2.0" coming as system update or new version PS5 pro with new PSSR?

System update, plus patches.

 

[Bojji]

None? As far as I'm aware, no PS4 game uses PSSR, unless there is a PS5 version, at which point it's a PS5 game.

This is probably about that sharpening filter update already used in PS4 games (that didn't got PS4 Pro support).

Hopefully they use Marathon as a promotional for this new PSSR.

It should have clean IQ regardless, they are aiming for 5k internal resolution.

Sony and AMD are working on the solution together as partners, Sony had their hands in making FSR4 as good as it is.

And they sacrificed PSSR in the process hahaha.

 

[Panajev2001a]

And they sacrificed PSSR in the process hahaha.

no, that is callide agile iteration (releasing a 1.0 version) despite fanboys preferring static cameras over gameplay .

 

[MisterXDTV]

What was the point even of PSSR if other related tech is available.

Blame AMD, FSR 3.x was poor quality and not ML so Cerny was forced to make ML upscaling himself when PS5 Pro was developed

Now Cerny is helping them with FSR 4

As simple as that

 

[Bojji]

no, that is callide agile iteration (releasing a 1.0 version) despite fanboys preferring static cameras over gameplay .

Same gameplay that shows you motion blur, radial blur, chromatic abberation and drops dynamic resolution the moment you move the camera? Same movement that you have to see through LCD motion blur?

If PSSR and FSR4 development was connected then Sony did a really shitty job with it. As far as we know, very late in PSSR development AMD used some Sony help with ML training for FSR4 but it seems both models are separated. Now, with PSSR2 it's obvious that it will be based on FSR4.

 

[Panajev2001a]

Same gameplay that shows you motion blur, radial blur, chromatic abberation and drops dynamic resolution the moment you move the camera?

Pick it up with all the reviews that say that in motion PSSR cleaned up IQ better than the native PS5 version .

Nobody says FSR4 and PSSR2 will not be better but PSSR1 is demonised in funny and not very very unbiased ways .

 

[Bojji]

Pick it up with all the reviews that say that in motion PSSR cleaned up IQ better than the native PS5 version .

Nobody says FSR4 and PSSR2 will not be better but PSSR1 is demonised in funny and not very very unbiased ways .

Just look at results in many games. It's not demonized because people are mean Xbox fanboys...

 

[Panajev2001a]

Just look at results in many games. It's not demonized because people are mean Xbox fanboys...

I do, some of the same people think Switch 2 has a good LCD screen, pretend not to see ghosting and smearing and tons of other artefacts in UE5 games and tech demos, and overly focus on fizzing details when the camera is still vs when the camera is moving (see PSSR in Sony's games for example)…

Not saying they are Xbox fanboys (some are the usual "everyone owns a 5090" PCMRs ), not all, they just like to troll Sony stuff and while PSSR 1.0 is not perfect, by far, exaggerating does not really help .

 

[Felessan]

If PSSR and FSR4 development was connected then Sony did a really shitty job with it. As far as we know, very late in PSSR development AMD used some Sony help with ML training for FSR4 but it seems both models are separated. Now, with PSSR2 it's obvious that it will be based on FSR4.

FSR4 is the next iteration of project Amethyst, so no, Sony didn't do a shitty job
The timing of PSSR was just kinda bad as it was released just before pure CNN models became obsolete, and Pro relies on INT8fot ML made it hard to immediately use transformers (those on base level might have very big or very low value of weights, outside of INT8 range).
It really was a bad luck for Sony how timing of Pro and timing of next generation of ML models aligned.

PSSR2 even though might be similar to FSR4 in model architecture, will most probably follow PSSR tech path and not FSR4 - fixed memory utilization with fixed execution time. The goal of Cerny for PSSR is for ML to be efficient in resources utilization and not eat too much into limited resources of console.

 

[Bojji]

I do, some of the same people think Switch 2 has a good LCD screen, pretend not to see ghosting and smearing and tons of other artefacts in UE5 games and tech demos, and overly focus on fizzing details when the camera is still vs when the camera is moving (see PSSR in Sony's games for example)…

Not saying they are Xbox fanboys (some are usual everyone owns a 5090 PCMRs ), not all, they just like to troll Sony stuff and while PSSR 1.0 is not perfect, by far, exaggerating does bit really help .

PSSR works ok in many games, especially Sony games. It's better than what FSR3 or TSR can achieve but it's not in the same realm of quality as DLSS3 or FSR4.

But with third some third party games this wasn't the case, and after many problems with PSSR - a lot of developers decided just to not bother with it at all.

FSR4 is the next iteration of project Amethyst, so no, Sony didn't do a shitty job
The timing of PSSR was just kinda bad as it was released just before pure CNN models became obsolete, and Pro relies on INT8fot ML made it hard to immediately use transformers (those on base level might have very big or very low value of weights, outside of INT8 range).
It really was a bad luck for Sony how timing of Pro and timing of next generation of ML models aligned.

PSSR2 even though might be similar to FSR4 in model architecture, will most probably follow PSSR tech path and not FSR4 - fixed memory utilization with fixed execution time. The goal of Cerny for PSSR is for ML to be efficient in resources utilization and not eat too much into limited resources of console.

Cerny talked about end result being on par with FSR4 in quality. In reality we don't know ANYTHING more about PSSR2, it's all just speculation.

 

[simpatico]

A firm release date was given, but PSSR used that data point to extrapolate 6 others, and that's where we get Jan - March. The AI checked it, and it checks out. No worries.

 

[SABRE220]

Same gameplay that shows you motion blur, radial blur, chromatic abberation and drops dynamic resolution the moment you move the camera? Same movement that you have to see through LCD motion blur?

If PSSR and FSR4 development was connected then Sony did a really shitty job with it. As far as we know, very late in PSSR development AMD used some Sony help with ML training for FSR4 but it seems both models are separated. Now, with PSSR2 it's obvious that it will be based on FSR4.

At first I was also baffled by the PSSR debacle but it seems like it was joined development. FSR4 wasn't ready for the ps5 pro launch and even at launch relies on FP8 for delivering its target quality. Sony basically used the work they were doing in unison until then and adapted that into PSSR1 porting it to int8. PSSR2 is a separate iteration that builds on PSSR1 and not FSR4 directly. From the leaked information its meant to be both cheaper in terms of performance and delivering at least equivalent image quality but aiming for better. Its clearly designed for the console environment and thus is more specialized and has to be more efficient due to the fixed specs. They worked on the framework together and now have developed their independent solutions and while it needs to be taken with a pinch of salt butseems Sony has outpaced Amd with PSSR2 and future development atleast the next gen version.

That being said I will double down on the fact that Sony effed up with PSSR1 it was a terrible launch and support for upgrades on the ps5 pro. It failed to deliver effectively on most of its main selling points especially RT beyond basic reflections. A premium console launch needs to match its premise it needs to be clean, impressive and seamless not a stuttering, delayed and bug riddled scenario.

 

[Gitaroo]

DLSS is praised to be the god tier ML upscaler, the recent 4.5 has big image regression very similar to PSSR1 artifacts especially for UE5 games. Not to say that PSSR is amazing but for a first attempt, not bad. I still think FG should be a priority a long with IQ improvement upscaling from lower internal res.

 

[Ashamam]

At first I was also baffled by the PSSR debacle but it seems like it was joined development. FSR4 wasn't ready for the ps5 pro launch and even at launch relies on FP8 for delivering its target quality. Sony basically used the work they were doing in unison until then and adapted that into PSSR1 porting it to int8. PSSR2 is a separate iteration that builds on PSSR1 and not FSR4 directly. From the leaked information its meant to be both cheaper in terms of performance and delivering at least equivalent image quality but aiming for better. Its clearly designed for the console environment and thus is more specialized and has to be more efficient due to the fixed specs. They worked on the framework together and now have developed their independent solutions and while it needs to be taken with a pinch of salt butseems Sony has outpaced Amd with PSSR2 and future development atleast the next gen version.

That being said I will double down on the fact that Sony effed up with PSSR1 it was a terrible launch and support for upgrades on the ps5 pro. It failed to deliver effectively on most of its main selling points especially RT beyond basic reflections. A premium console launch needs to match its premise it needs to be clean, impressive and seamless not a stuttering, delayed and bug riddled scenario.

Conceptually its clear that in terms of code paths Amethyst is the core, PSSR and FSR are the forks. Some aspects will no doubt get upstreamed back to the core, some won't as they are hardware specific or have context specific to console/PC. Its not like the collaboration just stopped with the advent of FSR 4..... or will cease with PSSR2.

Some people are unable to process the arrangement and are determined to uncouple the two, presumably so they can criticise the one they don't like?

 

[SABRE220]

Conceptually its clear that in terms of code paths Amethyst is the core, PSSR and FSR are the forks. Some aspects will no doubt get upstreamed back to the core, some won't as they are hardware specific or have context specific to console/PC. Its not like the collaboration just stopped with the advent of FSR 4..... or will cease with PSSR2.

Some people are unable to process the arrangement and are determined to uncouple the two, presumably so they can criticise the one they don't like?

I agree with the first part of your premise. They developed a core framework and forked from their for their own solutions depending on their needs. That being said, I have heard nothing about Sony and AMD working together on PSSR2 and future developments together, nor of Sony working with AMD to contibute towards FRS5.

The collaboration will be there for making better solutions generally since Sony has to rely exclusively on AMD for all its hardware and its both their benefits to design better products and more efficient and pioneering software but from here on out in terms of image reconstruction technologies and evolving versions of iterative solutions theres no concrete basis to say they will continue to keep themselves heavily involved with their continuously evolving forked solutions or if Sony will even directly share the particular efficiencies of PSSR2 and beyond with AMD for fsr5 etc.

 

[Bojji]

DLSS is praised to be the god tier ML upscaler, the recent 4.5 has big image regression very similar to PSSR1 artifacts especially for UE5 games. Not to say that PSSR is amazing but for a first attempt, not bad. I still think FG should be a priority a long with IQ improvement upscaling from lower internal res.

It has some regresses, especially in RT games. But it's completely optional and has to be forced by users, unlike PSSR that can't even be turned off in many game.

So giving players options is usually the best solution for stuff like this, like in hell is us. PSSR in this game is slightly sharper than TSR but it has horrible artifacts across entire image.

 

[mckmas8808]

I agree with the first part of your premise. They developed a core framework and forked from their for their own solutions depending on their needs. That being said, I have heard nothing about Sony and AMD working together on PSSR2 and future developments together, nor of Sony working with AMD to contibute towards FRS5.

The collaboration will be there for making better solutions generally since Sony has to rely exclusively on AMD for all its hardware and its both their benefits to design better products and more efficient and pioneering software but from here on out in terms of image reconstruction technologies and evolving versions of iterative solutions theres no concrete basis to say they will continue to keep themselves heavily involved with their continuously evolving forked solutions or if Sony will even directly share the particular efficiencies of PSSR2 and beyond with AMD for fsr5 etc.

While correct, it'd be weird for AMD to run from Sony's help considering Amethyst has already excelerated their ML work to catch up close to Nvidia with DLSS 4. And for Sony, it's also in their best interest to stick to building with AMD because they'll continue putting money into the R&D in order to maintain pace with Nvidia on a yearly basis.

Sony and AMD need each other to be honest. Because Nvidia is the King of ML in the games space right now.

 

[SABRE220]

While correct, it'd be weird for AMD to run from Sony's help considering Amethyst has already excelerated their ML work to catch up close to Nvidia with DLSS 4. And for Sony, it's also in their best interest to stick to building with AMD because they'll continue putting money into the R&D in order to maintain pace with Nvidia on a yearly basis.

Sony and AMD need each other to be honest. Because Nvidia is the King of ML in the games space right now.

No one would be running from anyone's help though. The collaboration remains, and they will continue developing new technologies and work towards general breakthroughs. The only thing that would be separate is that Sony will keep their iterative tech that builds upon the works of their collaboration as they keep advancing separately. If Sony improves upon the work of FSR/PSSR1, making it specialized, cheaper, more efficient, and sharper, they are not obliged to share that back to incorporate into AMD's open solution tech since they are entitled to enjoy the competitive advantage to their console platform, whereas AMD's solutions will benefit the competitors as well.

Basically, imagine if they achieve a breakthrough in developing a new lighting pipeline for AMD hardware, making vertain parts of the pathtracing workloads run on AMD hardware better. They can name it whatever, and both incorporate that into their work, but Sony might specialize that further for their console, making it more efficient and robust pound for pound, and keep the somewhat spicier sauce for their own hardware. That being said, the general concept and learnings can then be incorporated towards r&d to make new ground into other techniques they can develop together.

 

[mckmas8808]

No one would be running from anyone's help though. The collaboration remains, and they will continue developing new technologies and work towards general breakthroughs. The only thing that would be separate is that Sony will keep their iterative tech that builds upon the works of their collaboration as they keep advancing separately. If Sony improves upon the work of FSR/PSSR1, making it specialized, cheaper, more efficient, and sharper, they are not obliged to share that back to incorporate into AMD's open solution tech since they are entitled to enjoy the competitive advantage to their console platform, whereas AMD's solutions will benefit the competitors as well.

Basically, imagine if they achieve a breakthrough in developing a new lighting pipeline for AMD hardware, making vertain parts of the pathtracing workloads run on AMD hardware better. They can name it whatever, and both incorporate that into their work, but Sony might specialize that further for their console, making it more efficient and robust pound for pound, and keep the somewhat spicier sauce for their own hardware. That being said, the general concept and learnings can then be incorporated towards r&d to make new ground into other techniques they can develop together.

The bolded is exactly what I expect actually.

 

[Ashamam]

That being said, I have heard nothing about Sony and AMD working together on PSSR2 and future developments together, nor of Sony working with AMD to contibute towards FRS5.

Thats the point of the forks, they aren't working on each others implementations. But they do both contribute back to the core project. I should add that my opinion is only based on whats been publicly stated. Maybe its completely different, but not if you take the statements of Cerny and AMD at face value.

 

[Three]

I do, some of the same people think Switch 2 has a good LCD screen, pretend not to see ghosting and smearing and tons of other artefacts in UE5 games and tech demos, and overly focus on fizzing details when the camera is still vs when the camera is moving (see PSSR in Sony's games for example)…

Not saying they are Xbox fanboys (some are usual everyone owns a 5090 PCMRs ), not all, they just like to troll Sony stuff and while PSSR 1.0 is not perfect, by far, exaggerating does bit really help .

The laser focus on some minute fizzle when standing perfectly still in a game will never cease to amaze me. I remember R&C where people first found that fizzle to complain about meanwhile totally ignoring that the other upscaler had completely removed moving light effects from the game.

 

[SABRE220]

Thats the point of the forks, they aren't working on each others implementations. But they do both contribute back to the core project. I should add that my opinion is only based on whats been publicly stated. Maybe its completely different, but not if you take the statements of Cerny and AMD at face value.

Agreed, the general concept and learnings from their work on their specialized solutions can be incorporated towards r&d to make new ground into other techniques they can develop together, but again within reason. That is at least what I would hope for, but corporate statements aren't always indicative of the real extent of collaboration that would happen in real time.

It might be more restrictive than the statement implies, with certain parameters firmly set to serve as limits for the information shared, especially since Sony was burned very badly by IBM in the past, when they were too polite and trusted IBM with their "partnership" based on their good relationship and partnership on the CELL. Meanwhile, IBM basically screwed them over in the most scumbag way possible and sold most of their R&D work over to Microsoft for their Xenon cpu. IBM employees were literally hiding work and files from their Sony colleagues in the opposite cubicle meanwhile having meetings with Microsoft and Sony in the same building hiding their deal with Microsoft for years, providing the knowledge and tech from their partnership with Sony. This debacle was a major reason for Crazy Ken getting demoted.

 

[Fafalada]

Meanwhile, IBM basically screwed them over in the most scumbag way possible and sold most of their R&D work over to Microsoft for their Xenon cpu.

I'm not sure I'd go that far, given that the overlap was gcpu cores in Cell which for a fair duration of the project weren't even PPC based.
Yes in retrospect picking PPC arch for that clearly cost them(then again MS didn't get a great deal there either) but none of the core IP that made up Cell made it across, and that was really Ken's baby.

This debacle was a major reason for Crazy Ken getting demoted.

I'd always thought it was building a machine that lost 300+ at 600 price point, nearly killing the brand in process.
There were many things that went wrong on ps3 hw, not least of them being the complete screw up with GPU, but Cell at least lived up to its design specs in the end.

 

[Panajev2001a]

PSSR works ok in many games, especially Sony games. It's better than what FSR3 or TSR can achieve but it's not in the same realm of quality as DLSS3 or FSR4.

Nobody ever stated it was, Although compared to DLSS3 in games that use both (you did not say 4.5) PSSR tends to fight ghosting artefacts in motion better (this was evident in screenshots of the videos DF did even though DF did not always call it out).

 

[Panajev2001a]

the recent 4.5 has big image regression very similar to PSSR1 artifacts especially for UE5 games.

This is the release where they traded some performance too in order to improve IQ in motion and fight off against ghosting artefacts… who'd is what PSSR1 prioritised too so it is a bit fun if so hehe (as well as have variable Frame Gen which is a big deal for those that want to use Frame Gen).

 

[Felessan]

Agreed, the general concept and learnings from their work on their specialized solutions can be incorporated towards r&d to make new ground into other techniques they can develop together, but again within reason. That is at least what I would hope for, but corporate statements aren't always indicative of the real extent of collaboration that would happen in real time.

It might be more restrictive than the statement implies, with certain parameters firmly set to serve as limits for the information shared, especially since Sony was burned very badly by IBM in the past, when they were too polite and trusted IBM with their "partnership" based on their good relationship and partnership on the CELL. Meanwhile, IBM basically screwed them over in the most scumbag way possible and sold most of their R&D work over to Microsoft for their Xenon cpu. IBM employees were literally hiding work and files from their Sony colleagues in the opposite cubicle meanwhile having meetings with Microsoft and Sony in the same building hiding their deal with Microsoft for years, providing the knowledge and tech from their partnership with Sony. This debacle was a major reason for Crazy Ken getting demoted.

A lot of some crazy assumptions here
And it seems a lot of people doesn't really understand the core of AI/ML research
There is a models, those certainly researched together along with it's possible hardware requirements
And there is implementations of model those done separately by Sony and AMD as their hardware is different and even approach of how ML handled by hardware is different - AMD went with separate AI cores and Sony utilizing existing GPU power, making it trickier to use but power is more universal and can be shifted across workloads

 

[RSMan]

Me everytime this thread gets bumped

 

[Zathalus]

AMD went with separate AI cores and Sony utilizing existing GPU power, making it trickier to use but power is more universal and can be shifted across workloads

AMD does not use separate AI cores. The SIMD compute units include matrix–acceleration instructions baked directly into the shader ALUs. Neither the PS5 Pro nor RDNA4 use dedicated Tensor cores or a separate NPU, the approach taken by both are very similar, although the work done on PS5 Pro is heavily customized as RDNA4 was not fully ready in time for the PS5 Pro.

 

[Panajev2001a]

it was a terrible launch and support for upgrades on the ps5 pro. It failed to deliver effectively on most of its main selling points especially RT beyond basic reflections. A premium console launch needs to match its premise it needs to be clean, impressive and seamless not a stuttering, delayed and bug riddled scenario.

That is debatable. Nobody liked the increase in price compared to the way they launched the PS4 Pro especially, but there were plenty of titles where the PS5 Pro upgrade / edition was a noticeable improvement.
No matter the improvements, a Pro console's main goal and expectation is to allow for improved graphics and/or performance without requiring many changes if almost any at all from the devs and those choices lead to the path of holding some HW improvements back too (where to take advantage of them would require complexity normally reserved for a generation upgrade).

RT mode at 60 FPS for Ghost of Yotei, higher quality RT mode at 40/60 FPS for Spider-man 2, noticeable resolution and shadows improvement for Astro-Bot, noticeable in-game RT (and likewise for PSVR2) integration in GT7, etc… there are quite a few examples of PS5 Pro being a more than decent step change in action. With 2-3x improved RT cores (from base RDNA2) we are still a bit behind from 40x0 level RT, so no, PT upgrades in titles were not expected.

Looking at the sales, customers seems to have jumped aboard just like for PS4 Pro and if Sony does have a good batch of consoles in the warehouses they could take advantage of the rising PC costs to increase their PS5 Pro user base which would in turn help devs to allot more time to PS5 Pro improvements… so win win…

 

[Zathalus]

That is debatable. Nobody liked the increase in price compared to the way they launched the PS4 Pro especially, but there were plenty of titles where the PS5 Pro upgrade / edition was a noticeable improvement.
No matter the improvements, a Pro console's main goal and expectation is to allow for improved graphics and/or performance without requiring many changes if almost any at all from the devs and those choices lead to the path of holding some HW improvements back too (where to take advantage of them would require complexity normally reserved for a generation upgrade).

RT mode at 60 FPS for Ghost of Yotei, higher quality RT mode at 40/60 FPS for Spider-man 2, noticeable resolution and shadows improvement for Astro-Bot, noticeable in-game RT (and likewise for PSVR2) integration in GT7, etc… there are quite a few examples of PS5 Pro being a more than decent step change in action. With 2-3x improved RT cores (from base RDNA2) we are still a bit behind from 40x0 level RT, so no, PT upgrades in titles were not expected.

Looking at the sales, customers seems to have jumped aboard just like for PS4 Pro and if Sony does have a good batch of consoles in the warehouses they could take advantage of the rising PC costs to increase their PS5 Pro user base which would in turn help devs to allot more time to PS5 Pro improvements… so win win…

The PS5 Pro reception would have been much better received if PSSR had the same quality as FSR4. The hardware is perfectly fine, good really, just the upscaler needed some more time to cook.

 

[PaintTinJr]

Agreed, the general concept and learnings from their work on their specialized solutions can be incorporated towards r&d to make new ground into other techniques they can develop together, but again within reason. That is at least what I would hope for, but corporate statements aren't always indicative of the real extent of collaboration that would happen in real time.

It might be more restrictive than the statement implies, with certain parameters firmly set to serve as limits for the information shared, especially since Sony was burned very badly by IBM in the past, when they were too polite and trusted IBM with their "partnership" based on their good relationship and partnership on the CELL. Meanwhile, IBM basically screwed them over in the most scumbag way possible and sold most of their R&D work over to Microsoft for their Xenon cpu. IBM employees were literally hiding work and files from their Sony colleagues in the opposite cubicle meanwhile having meetings with Microsoft and Sony in the same building hiding their deal with Microsoft for years, providing the knowledge and tech from their partnership with Sony. This debacle was a major reason for Crazy Ken getting demoted.

I wouldn't draw the same parallel between the collaboration of PlayStation with AMD, and PlayStation with IBM because in hindsight it was IBM that came out far worse in the long run from them failing to protect the Cell BE processor success - undermining it with their Xbox Xenon deal.

POWER as a high-end populist chipset option has fallen into obscurity since they undermined the Cell, even letting Microsoft torch POWER's reputation by a post-release overclock to RROD levels to try and chase performance parity. All while IBM could have been reaping the future gen rewards of PS4, PS5 and eventually PS6 sales all with POWER APUs, and had a chipset that could have evolved being rapidly modernized and been in contention for the low power, graphics and emerging AI market too had the STI alliance continued. If anything it ranks as one of IBM's poorest business decisions alongside the original right's over MS DOS that IBM OS/2 could never overcome. By contrast AMD and Sony both understand their own positions in the collaboration and how openly working together helps them both leapfrog their own shortcomings while still doing business with others in a way that doesn't undermine their collaboration.

AMD like Nvidia(and Intel) have a desire for power efficient TOPs throughput for large scale with a heavier software runtime for scalability, whereas PlayStation wants power efficient TOPs throughput with versatility and a lightweight low latency software runtime so the depth of the collaboration seems partitioned by each company's solution objective and protected by patents, rather than any risks of disclosure IMO.

 

[PaintTinJr]

The PS5 Pro reception would have been much better received if PSSR had the same quality as FSR4. The hardware is perfectly fine, good really, just the upscaler needed some more time to cook.

That is so insincere. DF always stack the deck to undermine PlayStation tech every gen, because it allows them to farm maximum audience engagement.

So regardless of how good a modified 2019 silicon APU at 300TOPs using just ~1ms of ~9, ~16 or ~33ms in PSSR would have been, they would have always placed it against 2024 silicon with 1500 TOPs with a latency hiding 5ms run-time of umbrella software with a custom model chosen on a worst offending game (AW2 with its original AW1 XB1 exclusive roots), and then still not call out parity flaws in the PC version to suit their narrative and muddy waters further with RT noise being fixed up by other solutions like RR on PC and pass it off as a fair comparison.

As an example you only need look as Demon's Souls PSSR mode and outside of greater RT features or on a 5090 with MFG to run it @240 ffps( fake fps), ask yourself how a different solution was going to meaningfully better the "quality"?

Even from a point that the best examples of PSSR are on games exclusive to PS5/Pro and can't be bettered by DLSS/FSR4/XeSS shows that there is excellent examples of it not needing more time to cook, but just needing framed differently by those doing coverage.

 

[Zathalus]

That is so insincere. DF always stack the deck to undermine PlayStation tech every gen, because it allows them to farm maximum audience engagement.

DF praises Sony far, far more than any other developer. They were the ones praising GoY and DS2 graphics last year, when people dismissed GoY outright on this very forum. They are the ones that get exclusive interviews with Mark Cerny, who, in his own words, is a fan of them. Not only that they praise PSSR in a lot of titles. Pointing out the bad as well isn't a crime. Also, I'm not sure why you brought DF up? I certainly didn't mention them, nor are they are hardly the only one who have pointed out issues with PSSR noise in some titles.

they would have always placed it against 2024 silicon with 1500 TOPs with a latency hiding 5ms run-time of umbrella software with a custom model chosen

"latency hiding 5ms run-time"? I'm not even sure what that means. 1500 TOPs? I can only assume you mean the 5090? Which DF doesn't usually use in PS5 Pro comparisons. The first batch of comparisons they did in 2024 was before the 5090 released, and the AW2 comparison you seem to have a problem with was done on a 4070, which has less TOPs than a PS5 Pro.

on a worst offending game (AW2 with its original AW1 XB1 exclusive roots), and then still not call out parity flaws in the PC version to suit their narrative and muddy waters further with RT noise being fixed up by other solutions like RR on PC and pass it off as a fair comparison.

Some odd things here:

1. AW2 is by far not the worst offending game. Off the top of my head, Avatar, Star Wars Outlaws, Silent Hill 2, Silent Hill F, Rise of the Ronin, Metal Gear Solid 3 are all worse. Star Wars Outlaws and Avatar got patched and look decent now, although Outlaws could use some work.
2. AW1 XB1 exclusive roots? AW1 was a 360 game and on a completely separate engine. AW2 is based on the Northlight engine, which is one of only four that uses mesh/primitive shaders on current gen. The engine had roots on XB1, but it works really well on PS5, albeit it very demanding on both PS5/XSX. Control uses the same engine and it is categorically better on the Pro vs the XSX.
3. Any comparisons to PSSR has always been just with the regular DLSS versions and near as possible matched settings. Never with RR, unless it is a separate video showing off the benefits of RR, but PSSR never enters that discussion.

As an example you only need look as Demon's Souls PSSR mode and outside of greater RT features or on a 5090 with MFG to run it @240 ffps( fake fps), ask yourself how a different solution was going to do meaningfully better the "quality"?

Demon Souls is a good example of PSSR. Its also not on PC so I'm not sure what a 5090 or MFG has got to do with anything.

Even from a point that the best examples of PSSR are on games exclusive to PS5/Pro and can't be bettered by DLSS/FSR4/XeSS shows that there is excellent examples of it not needing more time to cook, but just needing framed differently by those doing coverage.

Even Sony is replacing it with a newer version based off FSR4. They clearly understand it is lacking in certain respects vs other upscalers. That is not to say PSSR can't look good, it absolutely can, but it can be better. And it will be better. Unless you think Sony and Mark Cerny are liers?

 

[omegasc]

The PS5 Pro reception would have been much better received if PSSR had the same quality as FSR4. The hardware is perfectly fine, good really, just the upscaler needed some more time to cook.

After the main issues became more evident, they should have acknowledged it and say they were working on those. It's ok to select a few best cases for release, but then the bad cases came and they still acted like everything was ok for too long.
They should showcase improvements made instead of forbidding devs to talk about PSSR versions. I don't understand their reasons, unless work on PSSR was already just maintenance until Amethyst/PSSR 2.0 was announced.