Support NeoGAF

[Physiognomonics]

The 45% is most likely the rendering pipeline

It's likely an average between bandwith improvement (28%) and Fflops improvement (single issue, 65%). Like 14 + 4 Cus on PS4 it's not really meaningful, on forums hours and hours of discussions had being done on 14 + 4 but it was all worthless in the end.

Here patched PS5 Pro game won't see this (raster rendering) improvement as they'll either uncap fidelity modes and we'll get twice the framerate or we'll get more RT effects at the same framerate. I think we'll see this in patched games that won't add either of those modes.

 

[Loxus]

In the Road to PS5 refers to RDNA2 CU as much LARGER than GCN, it's the same thing: @35:09 to 35:28



I don't get why you put 40 in there... The count doesn't make sense as 4 are disabled

Why should they count disabled CUs???

Every leak suggest it's a 60 active CUs , 54 seems out of the window now.

Also Navi 48 has 64 CUs

He said a 67% larger GPU, not 67% more CUs though.
The PS5 GPU has 40CUs.

RDNA3 CUs are actually smaller than RDNA2 CUs.
Both on 7nm, a RDNA3 CU is 3mm² vs RNDA2 4.33mm²

 

[MisterXDTV]

He said a 67% larger GPU, not 67% more CUs though.
The PS5 GPU has 40CUs.

RDNA3 CUs are actually smaller than RDNA2 CUs.
Both on 7nm, a RDNA3 CU is 3mm² vs RNDA2 4.33mm²

He said "Larger" in transistor count not die size.....

If you take a 36 CUs GPU and add 24 more CUs. You should get a GPU that is 67% "LARGER"

But hey, we will see about that...

Still it doesn't make sense to me to count disabled CUs that are basically useless....

 

[Loxus]

He said "Larger" in transistor count not die size.....

If you take a 36 CUs GPU and add 24 more CUs. You should get a GPU that is 67% "LARGER"

But hey, we will see about that...

Still it doesn't make sense to me to count disabled CUs that are basically useless....

I still don't understand why you assume he's referring to only CU count.

That number could include a number of things such as even Infinity Cache.

This is what a GPU consists of.
All of these contribute to the size of the GPU, not just CUs.
6700XT

 

[MisterXDTV]

I still don't understand why you assume he's referring to only CU count.

I mean it's CLEARLY written in the leaks by Henderson

EXCLUSIVE - More PlayStation 5 Pro Specs Detailed

Following this week's leaks on the PlayStation 5 Pro GPU specs and performance targets, Insider Gaming has learned more PS5 Pro specs.

insider-gaming.com

30 WGPs running specialised BVH8 traversal shaders vs 18 WGPs running BVH4 tranversal shaders on the standard PlayStation 5.

18 WGPs = 36 CUs
30 WGPs = 60 CUs

36->60 = +67%

Seems pretty straightforward to me

 

[Loxus]

As for transistors.
RDNA2 36 CUs = 212 million × 36 =7.740 billion
RDNA3 36 CUs = 331 million × 36 = 11.916 billion.
RDNA3 60 CUs = 331 million × 60 = 19.860 billion.

7.740 billion + 67% = 12.925 billion.

Going by transistors is out of the question.

 

[nosoynadie22]

I hope to see Spiderman2 with path-tracing, I trust in Cerny's magic

 

[Loxus]

I mean it's CLEARLY written in the leaks by Henderson

EXCLUSIVE - More PlayStation 5 Pro Specs Detailed

Following this week's leaks on the PlayStation 5 Pro GPU specs and performance targets, Insider Gaming has learned more PS5 Pro specs.

insider-gaming.com

30 WGPs running specialised BVH8 traversal shaders vs 18 WGPs running BVH4 tranversal shaders on the standard PlayStation 5.

18 WGPs = 36 CUs
30 WGPs = 60 CUs

36->60 = +67%

Seems pretty straightforward to me

Where does he mention larger here?

 

[MisterXDTV]

Where does he mention larger here?

He doesn't but this absolutely proves that the disabled CUs are NOT counted in these leaks

If that was the case, he would have said 20 WGPs instead of 18 for standard PS5 as:

20 WGPs = 40 CUs

So he's talking about ACTIVE CUs

 

[Loxus]

He doesn't but this absolutely proves that the disabled CUs are NOT counted in these leaks

If that was the case, he would have said 20 WGPs instead of 18 for standard PS5 as:

20 WGPs = 40 CUs

So he's talking about ACTIVE CUs

To be honest, he doesn't know if it's active or total. He's just reporting on what he heard.

He even admitted he's not tec savvy.

 

[MisterXDTV]

To be honest, he doesn't know if it's active or total. He's just reporting on what he heard.

He even admitted he's not tec savvy.

He is copying and pasting documents he has access to....

He's not analyzing anything

Even my mom could do it

 

[Loxus]

He is copying and pasting documents he has access to....

He's not analyzing anything

Even my mom could do it

Documents should have clock speed and TFLOPs. Those are the most important specs to leak.

Where are they?

 

[MisterXDTV]

Documents should have clock speed and TFLOPs. Those are the most important specs to leak.

Where are they?

TFLOPS are there:

33.5 TF Dual Issue

Clock speed can be calculated but it's not known yet

60 CUs -> 33.5 TF = 2.18 Ghz clock

 

[Loxus]

TFLOPS are there:

33.5 TF Dual Issue

Clock speed can be calculated but it's not known yet

60 CUs -> 33.5 TF = 2.18 Ghz clock

You can't see one set of info and be blind to the others.

 

[MisterXDTV]

You can't see one set of info and be blind to the others.

The difference is that Kepler doesn't have any document

Henderson has proven himself for 2 years....

See PS5 Slim(mer) and PS Portal

But hey, you are free to be trust who you want

 

[HeisenbergFX4]

The difference is that Kepler doesn't have any document

Henderson has proven himself for 2 years....

See PS5 Slim(mer) and PS Portal

But hey, you are free to be trust who you want

Tom also nailed the earbuds and he called the PS Pro controller long before anyone else

 

[MisterXDTV]

Tom also nailed the earbuds and he called the PS Pro controller long before anyone else

Absolutely and he clearly said that he WAS SHARED documents from developers. He's not posting his thoughts:

"Insider Gaming, who was also shared documentation from the developer portal under the condition that it’s not shared publically or privately can also confirm that Devkits have been available to first-party studios since September 2023, third-party since January 2024, and from Spring 2024 Testkits will also be available which will be identical to the final product.

PS5 Pro Specs Leak are Real, Releasing Holiday 2024

The leaked PS5 Pro specs leaked earlier today are real and the PlayStation 5 is still tentatively targeting a 2025 holiday release.

insider-gaming.com

 

[HeisenbergFX4]

Absolutely and he clearly said that he WAS SHARED documents from developers. He's not posting his thoughts:

"Insider Gaming, who was also shared documentation from the developer portal under the condition that it’s not shared publically or privately can also confirm that Devkits have been available to first-party studios since September 2023, third-party since January 2024, and from Spring 2024 Testkits will also be available which will be identical to the final product.

PS5 Pro Specs Leak are Real, Releasing Holiday 2024

The leaked PS5 Pro specs leaked earlier today are real and the PlayStation 5 is still tentatively targeting a 2025 holiday release.

insider-gaming.com

My 2 cents when it comes to future PS hardware there is absolutely no one better than Tom

 

[onQ123]

It's likely an average between bandwith improvement (28%) and Fflops improvement (single issue, 65%). Like 14 + 4 Cus on PS4 it's not really meaningful, on forums hours and hours of discussions had being done on 14 + 4 but it was all worthless in the end.

Here patched PS5 Pro game won't see this (raster rendering) improvement as they'll either uncap fidelity modes and we'll get twice the framerate or we'll get more RT effects at the same framerate. I think we'll see this in patched games that won't add either of those modes.

They wouldn't do it that way because rendering techniques hit their bottlenecks differently so they wouldn't use the average of different parts.

It's either they moved to 3 SE or they upgraded the rendering pipeline in each SE



Basically a move to 3 SE with the lower clock rate posted give you around 45% more fixed function rendering.

Same thing with the ROPs if they go from 64 ROPs to 96 while using the lower clockrate it will be able to push 45% more pixels ( when not using compute to render the pixels )

You can't see one set of info and be blind to the others.

In a system that already used variable clock rates I don't think it will break compatibility especially when there is more of everything in the GPU.

 

[Loxus]

The difference is that Kepler doesn't have any document

Henderson has proven himself for 2 years....

See PS5 Slim(mer) and PS Portal

But hey, you are free to be trust who you want

In that case, the PS5 Pro doesn't have 2 SE, since that came from Kepler only.

Tom Henderson never mentioned anything about Shader Engines or if that 30WGP is active or total.

Since PS4, 1CU in GCN or 1 WGP in RDNA was disabled in every SE.

PS4 has 2 SE with 1CU per SE disabled for yields.
PS4 has 4 SE with 1CU per SE disabled for yields.

With RDNA, 2CU are grouped together to create a work group processor. So two CUs get disabled when disabling a WGP for yields.

We all know how BC works on PS4 Pro and PS5.
CUs and clocks have to match.

• PS4
  18 CUs
  800 MHz
• PS4 Pro
  36 CUs
  911 MHz

It's done by turning off SE not CUs.
For PS5 Pro BC to work, common sense would tell you it's 3 SE.
PS4: 1SE / 18CU
PS4 Pro/5: 2 SE / 36CU

This leaves us with the third SE having the same number of CU, 18.
3SE × 18CU = 54CU total.

We know Sony disables 1WGP with RDNA architecture, meaning the full chip has 60CUs.

It's common sense when applying logic to the leaks.

 

[MisterXDTV]

It's common sense when applying logic to the leaks.

We will know sooner or later....

But now we know for sure that Navi 48 has 64 CUs/32 WGPs

Assuming PS5 Pro is based on that, and I see no reason to assume otherwise....

Your logic suggests that Sony disabled 10 CUs/5 WGPs for yields....

I find that highly unlikely

 

[Loxus]

We will know sooner or later....

But now we know for sure that Navi 48 has 64 CUs/32 WGPs

Assuming PS5 Pro is based on that, and I see no reason to assume otherwise....

Your logic suggests that Sony disabled 10 CUs/5 WGPs for yields....

I find that highly unlikely

When did I say Sony disables 10 CUs for yields?
I said it's 1WGP disabled per SE for yields.

Navi48 has 4SE.

Shader Engines are always symmetrical.
Example, 6800XT has 4SE/80CUs total, with 8 disabled.

So 1WGP has to be disabled in all 4 SE.
32WGP - 4WGP =28WGP/56CU (Active)
or
30WGP + 4WGP = 34WGP/68CU (Total)

 

[Loxus]

Another example of symmetrical SE, is the 6800.

It's base on Navi21 and only has 60CUs.
AMD actually disables an entire SE to achieve this.

 

[MisterXDTV]

At the end of the day it all comes down to this:

You think the specs leaked in the documents ARE counting useless disabled CUs, that no developer will ever care about....

I simply don't...

 

[Panajev2001a]

At the end of the day it all comes down to this:

You think the specs leaked in the documents ARE counting useless disabled CUs, that no developer will ever care about....

I simply don't...

Dev docs would not count disabled CUs, but then again to protect sources numbers may have been changed a bit? Is that possible?

 

[MisterXDTV]

Dev docs would not count disabled CUs, but then again to protect sources numbers may have been changed a bit? Is that possible?

I have no idea: but what's the point of leaking the specs, if you CHANGE the specs making them wrong???

I mean, it sounds absurd to me

 

[shamoomoo]

But those same peoples are also highly dismissive of PC in general. So pot, meet Kettle.

How? PC is an open platform with generally some advantages over console gaming when games are programmed right. But there are few games that pushes what the hardware is cable of to the max.

 

[Loxus]

Dev docs would not count disabled CUs, but then again to protect sources numbers may have been changed a bit? Is that possible?

Exactly this.

Also, documents would show the total amount of WGP/CU.

Example.
XBSX documents showed or talked about the full amount of WGP, which is 28.

 

[Panajev2001a]

I have no idea: but what's the point of leaking the specs, if you CHANGE the specs making them wrong???

I mean, it sounds absurd to me

Changing does not mean making HUGE differences… but .

 

[Physiognomonics]

At this point I am totally confused by those alleged specs and like Neo I am expecting to wake up at any moment.

 

[Gaiff]

But those same peoples are also highly dismissive of PC in general. So pot, meet Kettle.

Always funny when you see some people claim they don’t see the difference between 4K at Ultra and DLAA on PC vs 720p reconstructed to 1440p at Medium settings on consoles but then find the very same posters in DF comparison threads fighting over 1296p vs 1440p or a 2fps difference.

 

[playsaves3]

Of course it will be up to the dev and their time/resources etc to go back and patch. That said, if this approach by Sony is what it appears to be, it should be very little effort to do so. At the very least, they could just use PSSR to take the existing quality mode and deliver it at 60fps as is, whether it's running natively at 2160p, 1440p, or something in between. The result with minimal effort should be the same quality mode running much faster. With some more effort, they could do other things to clean up the IQ (i.e. push DRS targets higher, push native internal res higher, add higher quality graphics settings) and add RT effects if desired.

I expect that we'll see some Sony 1st Party games do the extra work to deliver on the full promise of the Pro. Titles like Returnal, GT7, Spiderman 2, and Ratchet & Clank will be the poster children for the Pro (at least initially) with added and/or improved RT effects running at 60fps with fidelity mode visuals. Recent releases such as Rise of the Ronan, Stellar Blade, and Helldivers 2 should also be improved quite a bit. As for third parties, it'll depend on the amount of effort to update but Sony knows this and I suspect it will be pretty painless to update.

Ratchet and clank id expect to run fidelity mode at 60 (and if you have a vrr screen allow you to uncap up to 120) while also adding the rt shadows and ambient occlusion from the pc version should easily be doable. They can optionally allow to have a toggle to turn on pssr in fidelity mode while doing this but hopefully that’s optional if you like the look

 

[playsaves3]

Console players will be OK with 30 fps for a cutting edge feature like path tracing. Just as PC players got off of their "60fps or bust" high horse to experience it.

Some older games could do it at 60fps maybe they can offer a performance mode with pt at like 720p

 

[playsaves3]

I don't know what you are smoking here, this is the difference between 7900XTX and 7700XT (closest GPU to Pro in raster power):

PS5 - 10.29 TFLOPS
XSX - 12.15 TFLOPS

vs.

7900XTX - 61.39 TFLOPS
7700XT - 35.17 TFLOPS (more than Pro)

Isn't far off...

The closer gpu is the 7800xt it’s better in some aspects and worse in others

 

[playsaves3]

Always funny when you see some people claim they don’t see the difference between 4K at Ultra and DLAA on PC vs 720p reconstructed to 1440p at Medium settings on consoles but then find the very same posters in DF comparison threads fighting over 1296p vs 1440p or a 2fps difference.

For me I notice differences more when it’s rt off compared to on which is where pc shines

 

[playsaves3]

Cyberpunk with PT should be possible. Especially if they use 1 ray instead of 2 rays.

Now if the 2x RT performance increase is what most devs can get out of it then no, you are not getting PT on the PS5. You can look at 2x more powerful RDNA2 cards trying to run PT and its virtually impossible. Or you are running at resolutions so low that it looks like shit because of all the RT noise artifacts.

Now if we are looking at 4x then it puts the PS5 Pro at around 3080 levels of performance, and ive been able to run it at 30 fps on 4k dlss performance with some drops to 25 fps. If they are able to lower the number of rays from 2 to 1 like that PC mod does then it should be doable.

But you have to take the higher end of Sony's own estimates to get PT and we cant say just how likely that is going to be. If Sony had simply said 4x instead of 2-4x then i would be more upset at Richard.

Path tracing on AW2 is very expensive and i wasnt able to run it over 20 fps. most of the time it was under 20 fps at 4k dlss performance.

I thought the 3080 is 3-3.5x the ps5 in rt not 4x wouldn’t 4x put it between 4070 and 4070ti?

 

[playsaves3]

I think the vanilla game has 2x bounces and 2x rays. As you mentioned before, 1x bounce and 2x rays could be feasible on the PS5 Pro. It's much more performant than the default 2x2 while not looking much worse. Also, I assume if Sony wanted it, they would aim for 30fps, not 60, so something like 1440p with an internal res of 1080p (or 960p?) with PSSR, and 1x bounce and 2x rays sounds plausible with a 30fps lock. Visually, it would look way better than the RT mode on the regular PS5 minus a softer IQ but if PSSR is decent, the IQ should still be passable.

The thing is, I'm not sure how much interest Sony will have in implementing something of the sort.

Whatever the case, I think it's a fair question and Rich just laughing it off is just bizarre. At that point, best come clean and just admit you were completely wrong about the performance of the machines instead of being befuddled whenever the Series X doesn't perform like you think it should. Didn't Cerny mention that it was much more difficult to keep a higher number of CUs busy, thus explaining in part why he went with 36? I vaguely remember him touching upon this in Road to PS5. Not that he'd need to tell us that anyway because it was obvious beforehand. DF are really looking like a bunch of tools there,

I think the gpu is actually strong enough to do path tracing at the current ps5 settings at native 1440p the reason we won’t have path tracing is cause of the rumored cpu you can see in benches cpu usage rises expoentially with it.

 

[Bojji]

The closer gpu is the 7800xt it’s better in some aspects and worse in others

It's 7800XT but downclocked and without inifinity cache. Plus of course better RT performance and Ai hardware.

I thought the 3080 is 3-3.5x the ps5 in rt not 4x wouldn’t 4x put it between 4070 and 4070ti?

I posted hybrid RT benchmarks before where AMD fares much better but in pure PT?

There is no 6700 on this chart but of course it would have lower FPS than 6700XT, assuming 5FPS it would put 4x better GPU in 20FPS section. Of course Pro will have 2-4x increase so norm would most likely be 2x or 3x, 4x will be rarer.

So... 3070 level.

 

[DirtyBast8rd]

If John is fired, they could still hire this shill

Thought I smelled something stink at work. Now I know what it was

 

[Little Mac]

Based on all the info we have, can someone predict Helldivers 2 performance on the pro?

I think the current PS5 is 1080p60 in performance mode?

 

[TrebleShot]

Based on all the info we have, can someone predict Helldivers 2 performance on the pro?

I think the current PS5 is 1080p60 in performance mode?

So probably 1440p 60 with upscaling to 4k maybe 90fps

 

[Bojji]

So probably 1440p 60 with upscaling to 4k maybe 90fps

Straight 1440p (or upscaled to 4K with PSSR) and around 60FPS, this game is for some reason super demanding for AMD GPUs

 

[Gaiff]

It's 7800XT but downclocked and without inifinity cache. Plus of course better RT performance and Ai hardware.



I posted hybrid RT benchmarks before where AMD fares much better but in pure PT?

There is no 6700 on this chart but of course it would have lower FPS than 6700XT, assuming 5FPS it would put 4x better GPU in 20FPS section. Of course Pro will have 2-4x increase so norm would most likely be 2x or 3x, 4x will be rarer.

So... 3070 level.

Cyberpunk seems broken on AMD cards.

The regular PS5 beats the 6700 by 42% with just RT shadows.

 

[Bojji]

Cyberpunk seems broken on AMD cards.

The regular PS5 beats the 6700 by 42% with just RT shadows.

It also beats Nvidia cards here. I think this can be memory bandwidth limited scenario. But who knows, maybe we see how differently PS5 api works with rt compared to dxr (I wonder how series x performs here with dxr?).

 

[Gaiff]

It also beats Nvidia cards here. I think this can be memory bandwidth limited scenario. But who knows, maybe we see how differently PS5 api works with rt compared to dxr (I wonder how series x performs here with dxr?).

Yeah, but that's like a 3fps difference and it can actually drop to 2-1fps. They're basically a match. The 6700 though just crumbles.

 

[FireFly]

It also beats Nvidia cards here. I think this can be memory bandwidth limited scenario. But who knows, maybe we see how differently PS5 api works with rt compared to dxr (I wonder how series x performs here with dxr?).

AMD cards do much better in the latest Cyberpunk 2.1 benchmarks.

Cyberpunk 2077 v 2.1 тест GPU/CPU | Action / FPS / TPS | TEST GPU

Мы выполнили тестирование Cyberpunk 2077 v 2. 1 на наивысших параметрах графики, на видеокартах из серий GEFORCE RTX и RADEON R

gamegpu.com

I know PL is more demanding, but I don't think it can account for the difference.

 

[Bojji]

Yeah, but that's like a 3fps difference and it can actually drop to 2-1fps. They're basically a match. The 6700 though just crumbles.

Yeah I have no idea why it's like that, this test was unlike any other in that DF comparison. Maybe a bug in AMD driver?

AMD cards do much better in the latest Cyberpunk 2.1 benchmarks.

Cyberpunk 2077 v 2.1 тест GPU/CPU | Action / FPS / TPS | TEST GPU

Мы выполнили тестирование Cyberpunk 2077 v 2. 1 на наивысших параметрах графики, на видеокартах из серий GEFORCE RTX и RADEON R

gamegpu.com

I know PL is more demanding, but I don't think it can account for the difference.

Different places to test probably but still here 6700x is almost 2x slower than 3070 and in TPU test it was almost 4x. Probably both game code patches and AMD drivers improved performance (a lot).

 

[winjer]

Cyberpunk seems broken on AMD cards.

The regular PS5 beats the 6700 by 42% with just RT shadows.

Of course it's broken on AMD.
It's an nvidia sponsored game. The most sponsored in recent times.

 

[Bojji]

Of course it's broken on AMD.
It's an nvidia sponsored game. The most sponsored in recent times.

Yeah compared to other games here even standard RT kills AMD hardware. But for path tracing benchmarks we only have this and AW2, no AMD sponsored game with PT exist (yet) to see AMD optimized version of it.

When I had 6800 most games with RT performed surprisingly well (when you read comments all the time that AMD RT is complete dogshit) but Cyberpunk wasn't one of them, lol.

But counter argument could be made that CP RT is much more heavy than most games and that's why it performs like that, without NV sponsored conspiracy theory.

Truth is probably somewhere in between...

 

[winjer]

Yeah compared to other games here even standard RT kills AMD hardware. But for path tracing benchmarks we only have this and AW2, no AMD sponsored game with PT exist (yet) to see AMD optimized version of it.

When I had 6800 most games with RT performed surprisingly well (when you read comments all the time that AMD RT is complete dogshit) but Cyberpunk wasn't one of them, lol.

But counter argument could be made that CP RT is much more heavy than most games and that's why it performs like that, without NV sponsored conspiracy theory.

Truth is probably somewhere in between...

Path Tracing is just Ray tracing with the Monte Carlo simulation.
It does send a lot more rays. And that is why performance drops, on all systems.