Support NeoGAF

[Bojji]

What the fuck are you talking about? dont make me pull up the receipts again about your bullshit:


Again this wasn't even the only time. I called out your flip flopping before:


Queue the new excuses.

That was about PSSR1, it it's quite fucking obvious that it had nothing to do with FSR4 looking at quality of output... They started developing PSSR1 way before AMD thought about FSR4.

And then: "Now, with PSSR2 it's obvious that it will be based on FSR4."

 

[Three]

That was about PSSR1, it it's quite fucking obvious that it had nothing to do with FSR4 looking at quality of output...

And then: "Now, with PSSR2 it's obvious that it will be based on FSR4."

Which again was identical training data but was an earlier iteration before FSR4 and FSR4.1/PSSR2. People were telling you this and you were denying it:

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.

no, that is callide agile iteration (releasing a 1.0 version)

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.

Besides you fucking liked a post saying "their joint effort went left and FSR4 went right" claiming the PS5 Pro couldn't handle FSR4. When talking about PSSR2. Don't backpedal now.

 

[winjer]

It's not about using the full amount, it's about frametime budget, the upscale will take longer to run on the WMMA units than it does on PS5 Pro, and then the cost is too high to implement it.

You have no idea of the upscale time for FSR4 on the Pro and neither on RDNA3 WMMA.

 

[Bojji]

Which again was identical training data but was an earlier iteration before FSR4 and FSR4.1/PSSR2. People were telling you this and you were denying it:





Besides you fucking liked a post saying "their joint effort went left and FSR4 went right" claiming the PS5 Pro couldn't handle FSR4. When talking about PSSR2. Don't backpedal now.

You even track what post I liked? Fuck me, that's some dedication!

News Rumor Hype  Post in thread '[MLiD] PS6 & PSSR2 Dev Update'

Dec 7, 2025

The Pro getting PSSR2 as well as FSR4 is breaking news isn't it? I thought Cerny described the upgrade as essentially FSR4 but with the PSSR nomenclature.

IMT8 version of FSR4 (in development build) already leaked and runs on GPUs with far weaker ML performance than PS5 Pro:



Even if they don't optimize it further it should run on PS5 Pro just fine, but IF PSSR2 has similar quality but runs faster no one will use FSR4.

• Bojji

• 

March 2025

Analysis  Post in thread '[Digital Foundry] Mark Cerny: FSR 4 'Re-Implemented' For PS5 Pro - Next-Gen PSSR Coming 2026'

Mar 10, 2025

Basically a port to different (custom) architecture of PS5 Pro. Will it have the same quality? We will see.

This also improves possibilities of FSR4 port to RDNA3 (or not, it doesn't support INT8).

• Bojji

Training data may be indentical for PSSR1 and FSR4 but one of them was vastly superior!

 

[Three]

You even track what post I liked? Fuck me, that's some dedication!

Nope but clearly a week apart I had already called you out on that and still remember. you keep doing this song and dance.

News Rumor Hype  Post in thread '[MLiD] PS6 & PSSR2 Dev Update'

Dec 7, 2025

The Pro getting PSSR2 as well as FSR4 is breaking news isn't it? I thought Cerny described the upgrade as essentially FSR4 but with the PSSR nomenclature.

IMT8 version of FSR4 (in development build) already leaked and runs on GPUs with far weaker ML performance than PS5 Pro:



Even if they don't optimize it further it should run on PS5 Pro just fine, but IF PSSR2 has similar quality but runs faster no one will use FSR4.

• Bojji

• 

March 2025

Analysis  Post in thread '[Digital Foundry] Mark Cerny: FSR 4 'Re-Implemented' For PS5 Pro - Next-Gen PSSR Coming 2026'

Mar 10, 2025

Basically a port to different (custom) architecture of PS5 Pro. Will it have the same quality? We will see.

This also improves possibilities of FSR4 port to RDNA3 (or not, it doesn't support INT8).

• Bojji

Training data may be indentical for PSSR1 and FSR4 but one of them was vastly superior!

Nobody was denying that the later iteration FSR4 was better than the earlier iteration PSSR1 and now PSSR2 is better than that. BTW you posting that quote shows the opposite. You clearly still thought they were separate by replying to this:

"The Pro getting PSSR2 as well as FSR4 is breaking news isn't it? I thought Cerny described the upgrade as essentially FSR4 but with the PSSR nomenclature."

And not just saying yes PSSR2 is just FSR4.1 specialised for PS5 Pro hardware. You still thought they would be using it in place of it.

 

[Bojji]

Nope but clearly a week apart I had already called you out on that and still remember you keep doing this song and dance.

Nobody was denying that the later iteration FSR4 was better than the earlier iteration PSSR1 and now PSSR2 is better than that. BTW you posting that quote shows the opposite. You clearly still thought they were separate by replying to this:

"The Pro getting PSSR2 as well as FSR4 is breaking news isn't it? I thought Cerny described the upgrade as essentially FSR4 but with the PSSR nomenclature."

And not just saying yes PSSR2 is just FSR4.1 specialised for PS5 Pro hardware. You still thought they would be using it in place of it.

We all were speculating last year, Cerny was saying that he want to reach FSR4 quality, not that console will be using FSR4. But it was quite obvious where this shit is going, and I suspect we have reached this point with that combined FSR4.1/PSSR2 release:

At this point in time FSR4 destroys PSSR1. Cerny wants to reach FSR4 quality with PSSR2 in 2016 (for current Pro). I doubt future version of PSSR will look better than FSR.

In the future I think both technologies will merge together, there is no point of two separate Ai upscalers when both companies are collaborating (Amethyst).

I mean, you and your friends were always super hardcore about PSSR2 being something completely different, specialized for console. When it basically is the same fucking thing with a different name (probably small differences here and there as well). It uses the same SR algorithm as FSR 4.1 as confirmed by Cerny himself.

 

[Three]

I mean, you and your friends were always super hardcore about PSSR2 being something completely different, specialized for console. When it basically is the same fucking thing with a different name (probably small differences here and there as well). It uses the same SR algorithm as FSR 4.1 as confirmed by Cerny himself.

No I wasn't, i merely mentioned the algorithm being different because it was adapted to PS5 Pro hardware. Not that they weren't partners in developing PSSR/FSR. It was you always trying to suggest they weren't, how they went separate ways and how Cerny was failing.

 

[Bojji]

No I wasn't i merely mentioned the algorithm being different because ot was adaptedto PS5 Pro hardware. Not that they weren't partners in developing PSSR/FSR. It was you always trying to suggest they weren't, how they went separate ways and how Cerny was failing.

Cerny failed with PSSR1 because AMD ouclassed him 3+ months later. He learned his lessons and now we have great ML upscalers for both platforms.

You are reading my posts the wrong way but whatever, I don't think there is any point of discussing this further. It's clear that from now on FSR/PSSR are interconnected and whatever branding Xbox Helix will use it's also connected to this.

 

[Karim]

Cerny failed with PSSR1 because AMD ouclassed him 3+ months later. He learned his lessons and now we have great ML upscalers for both platforms.

You are reading my posts the wrong way but whatever, I don't think there is any point of discussing this further. It's clear that from now on FSR/PSSR are interconnected and whatever branding Xbox Helix will use it's also connected to this.

Why continue to spit out this nonsense? PSSR at the beginning was supposed to be in a temporary state to evolve with the AI learning. Flaws were taken in the count. It's not like sony was caught off the guard. It happened what hypothetically planned from the beginning. Asking to a fully functional AI upscaler at the first iteration when need to be experienced and it was barely a concept, it's a paradox.

 

[Three]

Cerny failed with PSSR1 because AMD ouclassed him 3+ months later. He learned his lessons and now we have great ML upscalers for both platforms.

You are reading my posts the wrong way but whatever, I don't think there is any point of discussing this further. It's clear that from now on FSR/PSSR are interconnected and whatever branding Xbox Helix will use it's also connected to this.

You still don't get it. They didn't "outclass" eachother. They were partners. The Pro was delayed a year but PSSR succeded FSR3. Sony/AMD worked on FSR4 together. AMD released cards with FP8 after PS5 Pro launched, and they released the next iteration FSR4 later.
Sony/AMD continued working together on FSR4.1 adapting FSR4.1/PSSR2 to Int8 again later. The gaps in these weren't even big. AMD left their older cards in the dust. It's not a "from now on" thing they were always partners especially on model training.

 

[Felessan]

If pssr (3 at that point) is behind fsr5, they can just use fsr5 on PS6. Doesn't really matter.

If pssr3 will be behind fsr5 it would mean that PS6 hardware doesn't have blocks to run latter
Model R&D is joint efforts of both, so if something is worse than other it means that this something just can't run better version, that was the case with pssr1 and FSR4.0

I mean, you and your friends were always super hardcore about PSSR2 being something completely different, specialized for console.

The common opinion was that ~models~ developed together and ~implementation~ is tailored for specific hardware

When it basically is the same fucking thing with a different name (probably small differences here and there as well). It uses the same SR algorithm as FSR 4.1 as confirmed by Cerny himself.

It a kinda obvious that whatever best model project Amethyst came by will be used by both. Doesn't really means that they are the same "algorithm" as hardware is different and so are targets

 

[Bojji]

Why continue to spit out this nonsense? PSSR at the beginning was supposed to be in a temporary state to evolve with the AI learning. Flaws were taken in the count. It's not like sony was caught off the guard. It happened what hypothetically planned from the beginning. Asking to a fully functional AI upscaler at the first iteration when need to be experienced and it was barely a concept, it's a paradox.

FSR4 was fully functional ML upscaler - AMD first try...

You still don't get it. They didn't "outclass" eachother. They were partners. The Pro was delayed a year but PSSR succeded FSR3. Sony/AMD worked on FSR4 together. AMD released cards with FP8 after PS5 Pro launched, and they released the next iteration FSR4 later.
Sony/AMD continued working together on FSR4.1 adapting FSR4.1/PSSR2 to Int8 again later. The gaps in these weren't even big. AMD left their older cards in the dust. It's not a "from now on" thing they were always partners especially on model training.

Gap between PSSR1 and PSSR2/FSR4.1 is big, especially in games with low resolutions and problems with denoisers fucking up PSSR1 (and constant film grain in pretty much all games).

 

[Three]

Gap between PSSR1 and PSSR2/FSR4.1 is big, especially in games with low resolutions and problems with denoisers fucking up PSSR1 (and constant film grain in pretty much all games).

I'm not talking about the technical gap. I'm talking about the time in iterations of the upscaler PS/AMD are jointly developing. You mentioned "3 months", I'm talking about that.

 

[J Required]

Cerny failed with PSSR1 because AMD ouclassed him 3+ months later. He learned his lessons and now we have great ML upscalers for both platforms.

You are reading my posts the wrong way but whatever, I don't think there is any point of discussing this further. It's clear that from now on FSR/PSSR are interconnected and whatever branding Xbox Helix will use it's also connected to this.

Failed? Guy helped AMD catch up to Nvidia in no time, apparently Nvidia needs a guy like him after DLSS5 blunder. LMAO

 

[winjer]

Failed? Guy helped AMD catch up to Nvidia in no time, apparently Nvidia needs a guy like him after DLSS5 blunder. LMAO

AMD and Sony developed PSSR1 and FSR4 separated.
Amethyst only came after both released. FSR4.1 is the first result of the partnership.

 

[Bojji]

I'm not talking about the technical gap. I'm talking about the time in iterations of the upscaler PS/AMD are jointly developing. You mentioned "3 months", I'm talking about that.

FSR4 launched 3 months after PSSR and gap between them was massive. If they were developing them together (like some of you claimed), something went wrong for PSSR...

 

[Three]

AMD and Sony developed PSSR1 and FSR4 separated.
Amethyst only came after both released. FSR4.1 is the first result of the partnership.

no it didn't. Amethyst was announced (publicly) in December 2024. PS5 Pro came out November 2024. FSR4 was not released before Amethyst.

FSR4 launched 3 months after PSSR and gap between them was massive. If they were developing them together (like some of you claimed), something went wrong for PSSR...

Your timeline is messed up.

 

[Mr.Phoenix]

Doesn't really matter if its int/fp. PS6 will do both. Especially considering the more mature dual issue framework. There are things that an int model is great for, like upscaling, and there are things the fp model is good for, eg. RT and generative AI. This is even built into the AMD spec, called mixed precision something.... basically allowing the GPU to simultaneously run models on different precision simultaneously.

 

[Karim]

FSR4 was fully functional ML upscaler - AMD first try...



Gap between PSSR1 and PSSR2/FSR4.1 is big, especially in games with low resolutions and problems with denoisers fucking up PSSR1 (and constant film grain in pretty much all games).

Yeah let's pretend FSR1, 2 and 3 experience not give any lesson to AMD...

 

[Bojji]

no it didn't. Amethyst was announced (publicly) in December 2024. PS5 Pro came out November 2024. FSR4 was not released before Amethyst.

Your timeline is messed up.

You think they developed FSR4 in 3 months?

Mentioned in Sep 2024:

AMD FSR 4 To Go The Full "AI" Route, Follows NVIDIA DLSS & Intel XeSS To Deliver Better Visual Fidelity

AMD's next-gen FidelityFX Super Resolution technology known as FSR 4 will be leveraging AI in a grander scale to improve visual fidelity.

wccftech.com

Yeah let's pretend FSR1, 2 and 3 experience not give any lesson to AMD...

FSR1 and FSR2/3 have nothing to do with ML upscaling.

 

[Three]

You think they developed FSR4 in 3 months?

What are you talking about? Try to make some sense.

 

[winjer]

no it didn't. Amethyst was announced (publicly) in December 2024. PS5 Pro came out November 2024. FSR4 was not released before Amethyst.

Your timeline is messed up.

You forget that these things take time to develop.
PSSR and FSR4 were developed in parallel by both companies, independently.
FSR4 was the first result from AMD. And PSSR was the first result from Sony.
And FSR4.1 + PSSR2 is the first result from Amethyst.
If Sony and AMD had been making FSR4 as long as you claim, then PSSR2 would have released a year ago, at the same time as FSR4.0

 

[Karim]

You think they developed FSR4 in 3 months?



FSR1 and FSR2/3 have nothing to do with ML upscaling.

You are telling FSR4 is not based to TAA algorithm? You think the past upscalers didn't trained AMD to improve the algorithm logics?

 

[Bojji]

You are telling FSR4 is not based to TAA algorithm?

It is. But it's not similar to how FSR1 and FSR2/3 work.

Sony games were using TAA for over a decade and yet, they failed to make PSSR1 great at first try (like AMD with FSR4).

 

[Felessan]

I am pretty sure that FSR4 was also joint efforts of Cerny and AMD so it's not really correct to say that one outclassed another
It's just transformer evolution steps made it impossible to use it on Pro for 1st version and this was tied to specifics of transformer

This relate to transformer internal issues and until some workaround was researched, model was not usable on Int8. It was not a Cerny problem, it was model research problem that affect everyone on the market. And given that fp8 are less efficient calculations compared to Int8 - affected all upscalers performance too.
And until general research haven't find solution everyone suffered, but breakthrough came and model progressed to be usable even with only int8

AMD and Sony developed PSSR1 and FSR4 separated.
Amethyst only came after both released. FSR4.1 is the first result of the partnership.

That's ain't true
Both PSSR and FSR4 was already under Amethyst umbrella

 

[Karim]

It is. But it's not similar to how FSR1 and FSR2/3 work.

Sony games were using TAA for over a decade and yet, they failed to make PSSR1 great at first try (like AMD with FSR4).

Have you idea how algorithm logics improve or evolve? You don't completely scratch from zero a new one.

 

[winjer]

That's ain't true
Both PSSR and FSR4 was already under Amethyst umbrella

This would imply that Sony was using PSSR1 for a year, while FSR4 code was available. That doesn't make any sense.
The reality is that developing an ML model takes a lot of time. And the result of the Amethyst project only now is showing it's first result.
Before that Sony and AMD were developing separate models.

 

[Bojji]

Have you idea how algorithm improve or evolve? You don't completely scratch from the nothing a new one.

You do if you want to make something completely different. FSR1 and 2 are similar in names only.

 

[Karim]

You do if you want to make something completely different. FSR1 and 2 are similar in names only.

You are completely wrong. You preserve what work in the new one. Otherwise why persist to use TAA as basis?

 

[Bojji]

You are completely wrong. You preserve what work in the new one.

Even if it doesn't help them in any way?

 

[TheRedRiders]

Doesn't really matter if its int/fp. PS6 will do both. Especially considering the more mature dual issue framework. There are things that an int model is great for, like upscaling, and there are things the fp model is good for, eg. RT and generative AI. This is even built into the AMD spec, called mixed precision something.... basically allowing the GPU to simultaneously run models on different precision simultaneously.

Question is will dual issue compute offer a meaningful difference in such workloads considering it's bandwidth heavy, and the PS6 bandwidth isn't as impressive as its other specs, I think less than 50% improvement over the Pro?

 

[Karim]

Even if it doesn't help them in any way?

OK. No offence really but seems obvious to me you don't understand how this stuff really work.

 

[winjer]

You are completely wrong. You preserve what work in the new one. Otherwise why persist to use TAA as basis?

The only thing that comes from PSSR1 are the inputs. The AI model is clearly a new one, made from scratch. There is no PSSR1 code in PSSR2, besides the inputs.

 

[Bojji]

OK. No offence really but obviously you don't understand how this stuff work.

Tell me how spatial upscaler (FSR1) helped them in any way when they created image reconstruction (FSR2)?

Or how FSR2/3 helped then they were building Ai model? You could say that Sony could have used Insomniac reconstruction tech to develop PSSR1 - it would be much faster!

 

[Three]

You forget that these things take time to develop.
PSSR and FSR4 were developed in parallel by both companies, independently.
FSR4 was the first result from AMD. And PSSR was the first result from Sony.
And FSR4.1 + PSSR2 is the first result from Amethyst.
If Sony and AMD had been making FSR4 as long as you claim, then PSSR2 would have released a year ago, at the same time as FSR4.0

Has nothing to do with it. You clearly said

AMD and Sony developed PSSR1 and FSR4 separated.
Amethyst only came after both released. FSR4.1 is the first result of the partnership.

Blatantly incorrect. These things do take time to develop which was why PSSR was launched earlier. They continued working on the next iteration which was FSR4 but the algorithm was fp8 while the delayed PS5 Pro was int8. That took time and effort to implement and AMD did not even bother doing so for their older PC GPUs. Doesn't mean PSSR and FSR4 were separate models before though.

Right then and there when FSR4 released "3 months" after:

https://overclock3d.net/news/software/amd-confirms-that-sony-playstation-assisted-in-fsr-4-development/

 

[Mr.Phoenix]

You guys are looking at the whole Sony and AMD thing wrong.

Sony and AMD have always been working together, even before they put it in some sort of official moniker. They are so in bed with each other, that even the design patents Sony develops pretty much go into AMD GPUs.

What happened with PSSRv1 was simply a time issue. By the time Sony was finalizing the PS5pro and sending out dev kits, FSR4 did not exist yet. PSSR is kinda like an early version of what became FSR4, and Sony had to do that so it had some form of machine learning running on the Pro. Its all down to release roadmaps, that of a console is far more protracted than that of a GPU. Case in point, sony would have to have PSSR in their SDK 6-8 months before their hardware even releases vs AMD, which can release a GPU and upload the drivers, allowing a version of FSR4 that was just finilized two days before the GPU's release. Consoles do not have that luxury.

 

[winjer]

Has nothing to do with it. You clearly said

These things do take time to develop which was why PSSR was launched earlier. They continued working on the next iteration which was FSR4 but the algorithm was fp8 while the delayed PS5 Pro was int8. That took time and effor to implement and AMD did not even bother doing so for their older PC GPUs. Doesn't mean PSSR and FSR4 were separate models before though.

Right then and there when FSR4 released "3 months" after:

https://overclock3d.net/news/software/amd-confirms-that-sony-playstation-assisted-in-fsr-4-development/

Converting a model from Int8 to FP8 and vice versa is not the thing that takes more time. It's the training.
Saying that Sony was using PSSR1 for a year, while having made FSR4, makes no sense at all.

 

[Felessan]

This would imply that Sony was using PSSR1 for a year, while FSR4 code was available. That doesn't make any sense.

Keep playing dumb?
FSR4.0 not usable on Pro - it doesn't support fp8
And until ~transformer~ itself can't be implemented on Int8, and there were some tricks with it as some parts of transformers very sensitive to outliers leading to almost mandatory FP usage, Pro was out of luck

And it was not like Sony didn't wanted or didn't spent efforts - it's just problem is complex and hardly can be solved fast by efforts of one company. I posted white paper in another thread regarding this problem - it was from 2022 (very new for this type of things). Nvidia, who probably have the most power in upscaling tech, developed int8 transformer just at the end of 2025, not that far away from Sony/AMD

But yeah, ignore details and specifics and stick to "simple and obvious (and wrong" answer

The reality is that developing an ML model takes a lot of time. And the result of the Amethyst project only now is showing it's first result.
Before that Sony and AMD were developing separate models.

Those are your fantasies
We have some statements indicating that models were developed jointly and FSR4 based on pssr1
And it's understandable as model development is iterative process where the more smart guys you have the better

You just assume that just because Sony didn't use FSR4 outright means that they were behind, completely ignoring fp8 shenanigans those are very important to understand reasoning and flow of events

 

[winjer]

Keep playing dumb?
FSR4.0 not usable on Pro - it doesn't support fp8
And until ~transformer~ itself can't be implemented on Int8, and there were some tricks with it as some parts of transformers very sensitive to outliers leading to almost mandatory FP usage, Pro was out of luck

And it was not like Sony didn't wanted or didn't send efforts - it's just problem is complex and hardly can be solved fast by efforts of one company. I posted white paper in another thread regarding this problem - it was from 2022 (very new for this type of things). Nvidia, who probably have the most power in upscaling tech, developed int8 transformer just at the end of 2025, not that far away from Sony/AMD

But yeah, ignore details and specifics and stick to "simple and obvious (and wrong" answer

Those are your fantasies
We have some statements indicating that models were developed jointly and FSR4 based on pssr1
And it's understandable as model development is iterative process where the more smart guys you have the better

You just assume that just because Sony didn't use FSR4 outright means that they were behind, completely ignoring fp8 shenanigans those are very important to understand reasoning and flow of events

Of course FSR4 couldn't run on the Pro. It was made by AMD alone, for the units in RDNA4.
Amethyst made sure that it FSR4.1 model could run both on FP8 and Int8, on RDNA4 and the Pro.
You have no statements that show that FSR4.0 was developed in conjunction with AMD and Sony.
You only have statements that from December onwards AMD and Sony would collaborate on a new model.

 

[Karim]

Tell me how spatial upscaler (FSR1) helped them in any way when they created image reconstruction (FSR2)?

Or how FSR2/3 helped then they were building Ai model? You could say that Sony could have used Insomniac reconstruction tech to develop PSSR1 - it would be much faster!

Are you really serious? You really asking a company who has worked on TAA upscaler for years how can be smarter to design a new one based on AI compared to the other with zero experience?

 

[Three]

Converting a model from Int8 to FP8 and vice versa is not the thing that takes more time. It's the training.
Saying that Sony was using PSSR1 for a year, while having made FSR4, makes no sense at all.

Come on. What you're saying makes no sense and you know it. If the training for FSR4 was ready why didn't they just use it then on PS5 Pro if "Converting a model from Int8 to FP8 and vice versa is not the thing that takes more time"? Because the int8 algorithm wasn't ready. The training model was worked on together but they had to make it work on PS5 Pro hardware. Something AMD hasn't even done for their PC GPUs yet if it were so easy.

 

[Bojji]

Are you really serious? You really asking a company who has worked on TAA upscaler for years how can be smarter to design a new one based on AI compared to the other with zero experience?

Zero experience? Sony games use TAA since ~2016. Insomniac developed on of the first image reconstructions (before FSR2).

 

[Felessan]

Converting a model from Int8 to FP8 and vice versa is not the thing that takes more time. It's the training.

It actually is
Default transformer just can't be run on Int8, it loses a lot of precision due to attention blocks (the thing that define transformer over CNN) relies on outliers those just can't be fit in int

Training actually depends on processing power and I doubt Sony/AMD have lots of problem with it

 

[Karim]

Zero experience? Sony games use TAA since ~2016. Insomniac developed on of the first image reconstructions (before FSR2).

You start to make a lot of confusion. Software house and sony engineers aren't exactly the same crew.

 

[Bojji]

You start to make a lot of confusion. Software house and sony engineers aren't exactly the same crew.

Same is true for AMD. Years ago they had amazing hardware but shit driver team.

Sony used to share tech between developers in PS3 days using ICE team. If they wanted to develop image reconstruction they could have asked Insomniac to help with it. They were doing reconstructions even on PS3 using MSAA...

 

[Mr.Phoenix]

Question is will dual issue compute offer a meaningful difference in such workloads considering it's bandwidth heavy, and the PS6 bandwidth isn't as impressive as its other specs, I think less than 50% improvement over the Pro?

I don't see any reason why it shouldn't work as advertised. Being that this time around, both AMD and sony are doing a lot of things to get a lot of this stuff right. Eg.. its no longer picky about what kinda instructions it can see to qualify for a VOPD activity. The register rules arent picky anymore. And the compression thing they talked about in that keynote targets all data types for compression, which means less impact on bandwidth.

And another thing, everyone seems to mix up about bandwidth. Bandwidth determines how fast you can spit out frames. Consoles like the PS6.. has a priority to do 60fps and likely at no higher than a natively rendered 1440p resolution. You don't need a LOT of bandwidth to do that. Hell or even to do 120fps. And then there is FG too.

 

[Karim]

Same is true for AMD. Years ago they had amazing hardware but shit driver team.

Sony used to share tech between developers in PS3 days using ICE team. If they wanted to develop image reconstruction they could have asked Insomniac to help with it. They were doing reconstructions even on PS3 using MSAA...

Jesus Christ how you can be so obtuse seriously the argumentations you use are completely fool. How you can be serious I can't believe it. You trolling me.

 

[Felessan]

Of course FSR4 couldn't run on the Pro. It was made by AMD alone, for the units in RDNA4.
Amethyst made sure that it FSR4.1 model could run both on FP8 and Int8, on RDNA4 and the Pro.
You have no statements that show that FSR4.0 was developed in conjunction with AMD and Sony.
You only have statements that from December onwards AMD and Sony would collaborate on a new model.

You are very uninformed
Cerny himself told that FSR4 is a joint collaboration at release of fsr4

Mark Cerny: FSR 4 for PS5 Pro is the "next evolution of PSSR"

The RDNA 4 upscaler is the first result of AMD/Sony's Project Amethyst collaboration

www.digitalfoundry.net

You just know nothing about this stuff and pushing stupid agenda

 

[Bojji]

Jesus Christ how you can be so obtuse seriously the argumentations you use are completely fool. How you can be serious I can't believe it. You trolling me.

Because in your mind AMD developing FSR2 helped them in any way with FSR4. This isn't true at all, those things are separate just like Insomniac reconstruction is to PSSR.

 

[winjer]

You are very uninformed
Cerny himself told that FSR4 is a joint collaboration at release of fsr4

Mark Cerny: FSR 4 for PS5 Pro is the "next evolution of PSSR"

The RDNA 4 upscaler is the first result of AMD/Sony's Project Amethyst collaboration

www.digitalfoundry.net

You just know nothing about this stuff and pushing stupid agenda

And where is the PSSR equivalent of FSR4.0? There isn't because PSSR and FSR4 were designed separately.
FSR4.1 is the first result of Amethyst.
There is no version of FSR4 that looks as bad as PSSR1. And there is no version of PSSR1 that looks as good as FSR4.0
But there is now a version of PSSR2 that looks as good as FSR4.1.