Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
What do you mean, because DXR is an API not an implementation.
So I don't know why that would by default mean default AMD RT solution.
I mean acceleration structure that's compatible with DXR. DXR is an API, yes, but it does have it's requirements and limitations and if you want to accelerate parts of the process, it needs to work with DXR.
Well yeah one thing that hasn't really been talked about is if AMD's RT hardware scales with the size of the chip like Nvidia's does.... then it's possible XSX ends up with more "RT cores" and therefore better "RT performance"
Assuming of course that AMD's RT acceleration is based on the patent we know of, it scales with the amount of TMUs (and assuming CUs become bottleneck, with the amount of CUs)
 
Well yeah one thing that hasn't really been talked about is if AMD's RT hardware scales with the size of the chip like Nvidia's does.... then it's possible XSX ends up with more "RT cores" and therefore better "RT performance"
Yeah, good catch. But ofc clock and BW would also be factor of performance additioanlly to CU count, and Sony could choose to spend 2 RT units per TMU maybe? And we don't know if both have similar RT solution at all.
 
Assuming of course that AMD's RT acceleration is based on the patent we know of, it scales with the amount of TMUs (and assuming CUs become bottleneck, with the amount of CUs)

Well if the leaks are to be believed then XSX will have 224 TMU's and PS5 144 TMU's....
 
Is there any possibility that the 36 CUs don't represent the entire chip and is a Pro+/Enhanced BC mode for patched PS4 games; and when additional CUs are activated it may run at a lower clock (but still an overall higher TF)?

For eg.

18CU @ 800MHz = PS4 Base BC
36CU @ 911MHz = PS4 Pro BC
36CU @ 2000MHz = PS4 Pro+ / Enhanced BC (for Patched Games)
~52CU @ ~1800MHz = PS5 Native (not yet shown in testing?)
 
Is there any possibility that the 36 CUs don't represent the entire chip and is a Pro+/Enhanced BC mode for patched PS4 games; and when additional CUs are activated it may run at a lower clock?

For eg.

18CU @ 800MHz = PS4 Base BC
36CU @ 911MHz = PS4 Pro BC
36CU @ 2000MHz = PS4 Pro+ / Enhanced BC (for Patched Games)
>40CU @ <2000MHz = PS5 Native (not yet shown in testing?)

The problem with that theory is that the GitHub leak had a note "full chip" next to the 36CU 2Ghz tests AFAIK.
 
Regression test file for Oberon A0 is called "native". In iGPU test tab there are 3 tests, GEN0, GEN1 and GEN2.

GEN0 - PS4
GEN1 - PS4Pro
GEN2 - Ariel spec config

Now, for GEN2 whats interesting is that in same file, different tab there are macro settings which indicate GEN2 is using Ariel Spec iGPU and that sclk speed of 2.0GHz is defined as "upperlimit 100%".

Also, "full chip result" comment by WGP18 might be only valid for Ariel, and not Oberon, we dont know it for sure although it does point to 40CU chip with 36 active.

It seems like there is a script that is pulling a test template from Project and then fields are populated with actual results, thats why in BC1 there are test template from Arden while results are referring to BC1 - PS4 Pro

owbFLzp_d.jpg


ns11hTK.png
 
Last edited:
By this point I think it's just utilitarian. The style being consistent across three generations makes the font as much part of the brand recognition. The banality of it probably works in PS's favour, because it doesn't really need any flashy 'hey look at me'-ness; everyone knows PlayStation.

It also doesn't look dated yet. Maybe at some point it'll look gauche, like those 'modern' looking 80s fonts now look so horribly clumsy and retro. But for 2020, PS5's logo doesn't strike me as a negative. I guess more noise could have been generated with the big reveal by going with some new logo but, as I say, at this point in the brand, fifth one in, it's like Coca-Cola - the logo doesn't need to change because the message isn't changing. It's old faithfully. It's the same box you always had.
Personally I preferred the PS2 logo and the original Playstation Font
 
Is there any possibility that the 36 CUs don't represent the entire chip and is a Pro+/Enhanced BC mode for patched PS4 games; and when additional CUs are activated it may run at a lower clock (but still an overall higher TF)?

For eg.

18CU @ 800MHz = PS4 Base BC
36CU @ 911MHz = PS4 Pro BC
36CU @ 2000MHz = PS4 Pro+ / Enhanced BC (for Patched Games)
~52CU @ ~1800MHz = PS5 Native (not yet shown in testing?)

The 2.0GHz mode just being for boosting PS4 Pro games makes no sense...
 
Finally something most of us agree here...The PS5 logo is boring...and sucks.
The 2.0GHz mode just being for boosting PS4 Pro games makes no sense...
Agree.
Neither is for a full chip ps5. Heat, performance, cost, etc.
The only thing that makes sense for the real ps5 is above 50 CUs and a much lower clock.
Sure. But it makes perfectly sense for a devkit. You don't care about heat or cost for a devkit. And Ariel / Oberon at 36CUs make perfectly sense for complete BC regressions tests. They just needed 36CUs for that.

As for PS5, well, Cerny actually said to DF that the APU (or design ?) was smart. So we can clearly state that Cerny is definitely not modest.
 
If they changed the logo, you can be sure the internet would explode in the other direction "why did they change it for this shit? There was nothin wrong with the other one! Sony sucks I'm canceling my preorder if they don't bring back the old logo!"

...or they don't show the logo at all, "why didn't they show the logo? How can they unveil it without a logo? Is there something wrong with the logo?"
 
Finally something most of us agree here...The PS5 logo is boring...and sucks.

Agree.

Sure. But it makes perfectly sense for a devkit. You don't care about heat or cost for a devkit. And Ariel / Oberon at 36CUs make perfectly sense for complete BC regressions tests. They just needed 36CUs for that.

As for PS5, well, Cerny actually said to DF that the APU (or design ?) was smart. So we can clearly state that Cerny is definitely not modest.
Not like he would said "we just made a stupid APU design" anyway(not saying it's a stupid design).

Neither is for a full chip ps5. Heat, performance, cost, etc.
The only thing that makes sense for the real ps5 is above 50 CUs and a much lower clock.
I feel like some of you guys expecting a lot from PS5.
 
The serif SONY logo is the same since 1957.

Not like he would said "we just made a stupid APU design" anyway(not saying it's a stupid design).


I feel like some of you guys expecting a lot from PS5.
The only way 50+ CU happens with ps5 is with two skus. And that would be too close in performance, doesn't seem to be worth the cost of making two different chips.
 
Neither is for a full chip ps5. Heat, performance, cost, etc.
The only thing that makes sense for the real ps5 is above 50 CUs and a much lower clock.

It's not the same thing though. You are not pushing a chip that much just for BC...but pushing a 36CU chip for the new generation is at least somewhat believable. I'm not even sure that would work for BC given the instability it would introduce since it's over double the frequency of what the games originally ran on. Cerny even said tinkering with different frequencies introduced problems for BC on Pro.
 
There is a rumor about the Zen engineers helping on clock and efficiency for RDNA2, which would increase clock if the consoles are using RDNA2. It would be on 7nm+, which also help clock higher.
The CEO stated that some members from the Zen team had moved to RTG, and there was discussion about introducing some of the methodologies to the graphics tech.
No specific product or generation was specified as being the target, and the most recently disclosed optimized silicon is the Vega-based Renoir GPU.


Which part of RDNA1 has hardware acceleration for raytracing again?
That's right. There's absolutely no reason to think either console would be RDNA1, especially since based on the patents the RT acceleration is part of TMUs and I really doubt they'd mix'n'match TMUs of one generation with CUs of another
(only chance I can really see for RDNA1 in a console is Sony if by some miracle they're not using AMD RT)
That's much of what is made available for semi-custom clients. Cerny described a process where they were offered a menu of IP choices from current and future products, which could be put into their product.
Sony modified how the GPU L2 functioned with compute and used an expanded ACE front end for the PS4, and modified the SIMD hardware and ROP export path to include 2xFP16 and the ID buffer, respectively.

Microsoft's GPU included custom inclusions for legacy texture formats into the TMU hardware, as well as modifications for allowing for activating AF on legacy applications that didn't have it on Scorpio.

edit: about PS4 Pro, Sony never said it's "Polaris based", they said it uses "many new features from the Polaris architecture as well as several even beyond it".
There's few different possibilities, but either the 2xFP16 was trivial change or the gfx complex is Vega rather than Polaris. Think of AMD GPUs as Lego bricks, they have bunch of different bricks which are interchangeable and can work together (and to my understanding CUs aren't separate block, but rather part of bigger GFX block which would tie said block to specific architecture. Probably best recent example of this is the Intel "Vega", it's graphics block is Polaris, it's HBCC memory controller was first introduced in Vega and who knows which versions of vce, vcd, display controller and so on it includes
There's been some indications that the Pro's GPU is Sea Islands.
If you believe the github tests are valid information, their backwards compatibility tests specifically exclude certain modern instruction types from the PS4 and Pro modes that are available natively.
In other instances, either Sony specifically described how it brought in only certain next-gen features from Vega, or games that used RPM did not mention using any other useful architectural changes introduced after Sea Islands. If cross-lane operations and more effective scalar memory operations could have improved uniform calculations, I doubt an optimization team like the one used for Doom's Pro version would have ignored them.

Sony has been willing to customize its GPU beyond trivial changes, so I would say inserting RPM back into a prior generation ISA is non-trivial but not disruptive to the whole GPU.
 
Neither is for a full chip ps5. Heat, performance, cost, etc.
The only thing that makes sense for the real ps5 is above 50 CUs and a much lower clock.

stadia CU count is reported to be 56, I’ve always had this feeling that the ps5 would be 54, 2 being inactive. but l don’t know sh@t.
 
stadia CU count is reported to be 56, I’ve always had this feeling that the ps5 would be 54, 2 being inactive. but l don’t know sh@t.
They can't do this because of how the CUs are now grouped by 2 into one WGP. They need to deactivate the whole WGP. One on each sides of the butterfly design so 4CUs in total.
 
Status
Not open for further replies.
Back
Top