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

[iroboto]

https://www.resetera.com/threads/ps5-and-xbox-series-speculation-ot12-aw-hell-transistors-running-this-again.173318/page-30

A lot of Eras say OBR/Ariel is Navi 10 (or Navi 10 lite). Is't it real?

Uhh. Yea just speak with Absolute on this one. I didn’t actually follow along to the detail that he did. I only followed along when DF covered it; and there was no mention of Any console being any architecture except the talk of Navi.

 

[Esrever]

that's not true.
I think people are hell bent on proving that GitHub leaks are real.
that doesn't mean that PS5 can't be more than 9.2TF.

doesn't mean ps5 isn't less than 9.2 either, why are people focused on it being more?

 

[Scott_Arm]

Oh the horrors!

Honestly, the whole argument over the specs is arguing the numbers for the sake of the numbers and not for the product on the screen ... Well, the other reason is just pure entertainment from people blowing up.

 

[RDGoodla]

Its not that. Its that people think Ariel = Oberon. Its not true, never was. I talked about this for months, hence I am called cult leader. Ariel was known since December 2018. Ariel was known as GFX1000, it was Navi 10 derivative, RDNA1 chip.

View attachment 3637

Oberon is later chip, its 2nd APU, and entire Github repo for Oberon was made up of Native/BC1/BC2 tests which are done on Ariel iGPU testlist. Therefore, since Ariel was Navi 10 Lite, you couldn't have ray tracing and variable rate shading running Ariel's testlist.

People are trying to discredit a leak without understanding basics about it, its frustrating.

https://www.eurogamer.net/articles/...laystation-5-xbox-series-x-spec-leak-analysed
Gonzalo is also Navi10 lite/

 

[iroboto]

doesn't mean ps5 isn't less than 9.2 either, why are people focused on it being more?

I guess there hasn’t been much information on it being less than 9.2 TF. Making it less probable. There have been lower early specs maybe that is leaking around; but I wouldn’t hold those to end up as being accurate in the final release

 

[N2O]

https://www.eurogamer.net/articles/...laystation-5-xbox-series-x-spec-leak-analysed
Gonzalo is also Navi10 lite/

*DF think it's Navi10 lite.

Back when everyone think both console will be RDNA, you know.

 

[AbsoluteBeginner]

https://www.eurogamer.net/articles/...laystation-5-xbox-series-x-spec-leak-analysed
Gonzalo is also Navi10 lite/

And Gonzalo contains iGPU named Ariel, PCI ID 13E9.

AMD 2G16002CE8JA2_32/10/10_13E9



So, as I said

Ariel - 13E9/13F8 (Gonzalo platform). Navi 10 Lite (RDNA1)
Oberon - 13F9 (Flute platform). RDNA2

 

[RDGoodla]

*DF think it's Navi10 lite.

Back when everyone think both console will be RDNA, you know.

You can see Absolutebeginner posts.

Ariel and Gonzalo are RDNA1 chips and OBR is RDNA2 chip.

Basically we can say there are at least 2 chips made by SONY.

 

[temesgen]

My personal opinion is the hw theoretical performance won't matter if they scale by resolution. 9TF is 75% of the compute power of 12TF.

x * y = 3840 * 2160 * 0.75
16/9 * y * y = 6,220,800
y2 = 6,220,800/1.78
y = 1869

x = 6,220,800 / y
x = 3328

So 3840 x 2160 for 12TF console vs 3328 x 1869 for 9TF console. How will anyone ever live with the compromise? I'm sure people will 100% be able to tell the difference.

Cerney said 8TF was needed for 4k. I think the right analogy is a vending machine. Anything less than a buck (8tf) won't get you a candy bar (4k) but $1.50 might get you a hostess cupcake (4k at 60fps).

We don't have any way of knowing if 12TF is enough to get there (cupcake could cost more than 12tf) but we do know both platforms will get you something out of the machine (4k).

We could be looking at one kid with a buck who gets a candy bar out (4k at 30fps) and another kid with a $1.50, who gets the same candy bar and some change (4k at 50fps) or a hostess cupcake (4k at 60fps). Noway to know yet.

 

[anexanhume]

Cerney said 8TF was needed for 4k. I think the right analogy is a vending machine. Anything less than a buck (8tf) won't get you a candy bar (4k) but $1.50 might get you a hostess cupcake (4k at 60fps).

We don't have any way of knowing if 12TF is enough to get there (cupcake could cost more than 12tf) but we do know both platforms will get you something out of the machine (4k).

We could be looking at one kid with a buck who gets a candy bar out (4k at 30fps) and another kid with a $1.50, who gets the same candy bar and some change (4k at 50fps) or a hostess cupcake (4k at 60fps). Noway to know yet.

That comment was largely believed to be in relation to the time, since PS4 was ~2TF and 4K is 4x the pixels, hence 4x the TF to maintain visuals.

 

[Scott_Arm]

Cerney said 8TF was needed for 4k. I think the right analogy is a vending machine. Anything less than a buck (8tf) won't get you a candy bar (4k) but $1.50 might get you a hostess cupcake (4k at 60fps).

We don't have any way of knowing if 12TF is enough to get there (cupcake could cost more than 12tf) but we do know both platforms will get you something out of the machine (4k).

We could be looking at one kid with a buck who gets a candy bar out (4k at 30fps) and another kid with a $1.50, who gets the same candy bar and some change (4k at 50fps) or a hostess cupcake (4k at 60fps). Noway to know yet.

I don't understand this analogy. I'm showing that if you just scale by resolution to make up the difference the compromises will be very small. There is no fixed number of TF to reach 4k. You can make a game that looks like Superhot run at 4k very easily. Developers choose where the flops go.

 

[McHuj]

People actually thinking this invalidates the Gospel. Oh you poor souls, Oberon Native test is regression test with Ariel iGPU spec (GX10 - ie. Navi 10 derivative), it wasn't suppose to contain RT/VRS.

2.0GHz was only thing not making complete sense, yet, it does now...

Also, not only will we see 2.0+ GHz chips, IPC has been improved by 15% as well. I think we are probably looking at 9.2TF RDNA2 beating 2080.

To quote another insider, "bingo!"

I think today's AMD conference only strengthens the idea that PS5 is around 9.2 TF. It will probably come in sub 200W and at $399 as well.

I think Abert Panello was guessed right while he was at MS when he expected a PS5 around 8TF and at $399. MS wouldn't win a head to head $399 console battle with equally spec consoles, but they can make up significant ground with a $299/$499 low-tier/high-tier approach.

And if MS only "wins" US/UK markets with this approach, it doesn't matter because that guarantees no third party exclusives.

 

[AbsoluteBeginner]

To quote another insider, "bingo!"

I think today's AMD conference only strengthens the idea that PS5 is around 9.2 TF. It will probably come in sub 200W and at $399 as well.

I think Abert Panello was guessed right while he was at MS when he expected a PS5 around 8TF and at $399. MS wouldn't win a head to head $399 console battle with equally spec consoles, but they can make up significant ground with a $299/$499 low-tier/high-tier approach.

And if MS only "wins" US/UK markets with this approach, it doesn't matter because that guarantees no third party exclusives.

It also fits with another insider claiming new consoles are RTX2080 level. If there is 15% increase in IPC, then guess where 36CUs at 2.0GHz takes you?

 

[anexanhume]

It also fits with another insider claiming new consoles are RTX2080 level. If there is 15% increase in IPC, then guess where 36CUs at 2.0GHz takes you?

33% more than 9TF is XSX, which would make it 2080 Ti level. Jason said both. You can’t mix and match the part of rumors you like and ignore the rest.

 

[AbsoluteBeginner]

33% more than 9TF is XSX, which would make it 2080 Ti level. Jason said both. You can’t mix and match the part of rumors you like and ignore the rest.

I cannot, its more tongue in cheek. He also said not to hold him for his word and that he doesnt know exactly, therefore you wont see me quouting something like that in serious discussion.

 

[RDGoodla]

To quote another insider, "bingo!"

I think today's AMD conference only strengthens the idea that PS5 is around 9.2 TF. It will probably come in sub 200W and at $399 as well.

However bloomberg said PS5 uses an "unusually expensive" cooling solution.

If PS5 is RDNA2 9.2TF I don't think it needs super good cooling since RDNA2 is 1.5x performance per watt of RDNA1.

Considering the expensive cooling solution, it strengthens the idea that PS5 final silicon changes very very much from Gonzala/Ariel:
SONY can change to RDNA2 why they maintain the same CU number?
It seems Odium's 11.6TF (which implies 44CUs@2.06GHz) is achievable with RDNA2 thus Sony needs an expensive cooling.

 

[BRiT]

Jason said both. You can’t mix and match the part of rumors you like and ignore the rest.

But everything changes. So sayeth the insiders. So you can pick and choose.

 

[iroboto]

It seems Odium's 11.6TF (which implies 44CUs@2.06GHz) is achievable with RDNA2 thus Sony needs an expensive cooling.

Odium also said they were RDNA 1.9 and not 2.0

So...

it's inconclusive; it's really a matter of which leakers people want to believe. There is just so little data on PS5. We have a snapshot of something with Github. But outside of that, not much to work with.

 

[AbsoluteBeginner]

However bloomberg said PS5 uses an "unusually expensive" cooling solution.

If PS5 is RDNA2 9.2TF I don't think it needs super good cooling since RDNA2 is 1.5x performance per watt of RDNA1.

Considering the expensive cooling solution, it strengthens the idea that PS5 final silicon changes very very much from Gonzala/Ariel:
SONY can change to RDNA2 why they maintain the same CU number?
It seems Odium's 11.6TF (which implies 44CUs@2.06GHz) is achievable with RDNA2 thus Sony needs an expensive cooling.

I still think 36CU at 2.0GHz and 16GB of GDDR6 RAM (16-18Gbps) will require better cooling then Pro or PS4 had.

I dont know how much watts do these chips require, but it must be quite a bit (when 5700XT comes with 8GB 14Gbps).

Will still probably hit ~180-190W I assume.

 

[N2O]

However bloomberg said PS5 uses an "unusually expensive" cooling solution.

If PS5 is RDNA2 9.2TF I don't think it needs super good cooling since RDNA2 is 1.5x performance per watt of RDNA1.

Considering the expensive cooling solution, it strengthens the idea that PS5 final silicon changes very very much from Gonzala/Ariel:
SONY can change to RDNA2 why they maintain the same CU number?
It seems Odium's 11.6TF (which implies 44CUs@2.06GHz) is achievable with RDNA2 thus Sony needs an expensive cooling.

If PS5 isn't very large, a good cooling solution without jet engine sound could very well be more expensive than usual.
And yeah RDNA2 performance is better, but come on doesn't mean Sony can suddenly add more CU in short time...