• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

PS5 Die Shot has been revealed

Bingo! Co processors
20200329143109.jpg


Seems like a core part of the PS5s design is to bring in a ton of data from the I/O and to do it often. All these custom pieces of hardware is to ensure the performance doesn't tank every time the I/O is accessed.
 
That seems on point to me. It certainly fits with what Cerny was saying about so much of the i/o pipelining optimization just "being there", and not requiring programmatic intervention.

From an engineering standpoint it makes sense: Basically why spend money on more cores when you can achieve the same result by effectively offloading a lot of heavy data pre-processing onto an intermediary layer between mass-storage and CPU/GPU?

I think alot of people focus on how fast the I/O is but I think it's important to look at how the I/O affects performance.
 

MonarchJT

Banned
Stop crying man if you can't comprehend fact from fiction..........I don't need to wake up because I never took things as fact when I knew they were speculation. How childish are you?
Hahahaha your telling me they are writing bad things about RTG on B3D? Is the world ending oh no..............................................
keep moving the goalpost i don't care about that youtuber. we can close it here.
Is you that care for him so you can fill forum with some tales from HIS ass like rdna3 ...IC ..and all the others fud
 

ethomaz

Banned
No, xsx is rdna 2 there's nothing contradicting this.If evidence are provided then we will talk.
Evidence about what exactly because what I posted is true.

If you mean Xbox 25% CU perf/clock increase from Xbox One X to Series X is the Hotchips presentation.

If you mean the over double digit CU perf/clock permanece increase for RDNA 2 over AMD RDNA it is the RDNA 2 presentation.

If you mean the 25% CU perf/clock increase for RDNA over GCN it is the AMD RDNA presentation.
 
Last edited:

ethomaz

Banned
Last edited:

cormack12

Gold Member
I think, and I hope it's not too controversial, that the coloured pictures are much better but they also remind me of gasoline puddles.

I think the new information is like a green potato chip. You've been loving eating your potato chips then you find a green one. You're not quite sure if you should eat it. But then you do, and everything is just fine.
 

azertydu91

Hard to Kill
MS confirmed 25% in their HotChips presentation.
But it was ambiguous if it was opposed to rdna 1 CU or GCN CU because if it was opposed to rdna 1 you are wrong (which I think is happening)but if it was opposed to GCn you are right.
Once again don't do like others are quick to do here and take one simple part out of context, but if you have a slide with explicitely stated 25% over gcn then yes I would admit that you are right.
 

M1chl

Currently Gif and Meme Champion

Clear

CliffyB's Cock Holster
I think alot of people focus on how fast the I/O is but I think it's important to look at how the I/O affects performance.

Yeah, I mean potentially it could be hugely beneficial. I mean I guess the standard way of thinking is that performance is like a drinking contest where whoever can take the biggest, fastest gulps (of data) wins. This is like a funnel! Just open the throat and down it goes.
 
Yeah, I mean potentially it could be hugely beneficial. I mean I guess the standard way of thinking is that performance is like a drinking contest where whoever can take the biggest, fastest gulps (of data) wins. This is like a funnel! Just open the throat and down it goes.

Definitely seems to be one of the things that Sony and Epic are betting on. Streaming in huge amounts of data seems to be what they are focused on. That could hurt performance whenever data is streamed in if you don't have the right hardware.

I guess we have to wait and see what these big data steaming games look like to see if it's worth it.
 

DaGwaphics

Member
Evidence about what exactly because what I posted is true.

If you mean Xbox 25% CU perf/clock increase from Xbox One X to Series X is the Hotchips presentation.

If you mean the over double digit CU perf/clock permanece increase for RDNA 2 over AMD RDNA it is the RDNA 2 presentation.

If you mean the 25% CU perf/clock increase for RDNA over GCN it is the AMD RDNA presentation.

AMD ever say what happens to the perf/clock increase if the Infinity Cache is removed from the equation?
 

FritzJ92

Member
I would expect a console releasing one year after and at $100 more to be better otherwise it would be stupid unless your Nintendo. The One X is a fantastic console.
Yeah bro... i agree. It just found it weird to say something less capable is better because it’s cheaper. That’s just a weird take. You can say a hardware is great without having to down talk another.
 
Last edited:
MS confirmed 25% in their HotChips presentation.
They boasted 25% performance increase in CU perf/clock from Xbox One X (GCN) to Series X.


I’m not even sure where you are getting 50% lol

Som people are living in the fantasy, would have been nice if Sony did also the hotchip stuff.... And now Cerny's conference was not that deep like hotchips...

But it was ambiguous if it was opposed to rdna 1 CU or GCN CU because if it was opposed to rdna 1 you are wrong (which I think is happening)but if it was opposed to GCn you are right.
Once again don't do like others are quick to do here and take one simple part out of context, but if you have a slide with explicitely stated 25% over gcn then yes I would admit that you are right.

Here is the official full HotChips conference for XSX. Of course, there isn't slide for every word they've said. It's impossible to make slides that much. LOL

But anyway :

Architecturally these CUs have 25% better performance per clock on average graphics workloads relative to the GCN generation

UpNlOnE.jpg


Timestamped at 16:50

 
Last edited:

azertydu91

Hard to Kill
Here is the official full HotChips conference for XSX. Of course, there isn't slide for every word they've said. It's impossible to make slides that much. LOL

But anyway :



UpNlOnE.jpg


Timestamped at 16:50


Alright that's interesting but is in total contradiction with amd's RDNA 2 presentation about RDNA 2 CU's, could it simply be a mistake from the man speaking ?
Because I think otherwise their are only presenting the same gains from GCN than RDNA 1 CU's which means that it is indeed RDNA 1 Cu's (that I don't believe) r maybe that had to make concessions to fit that much CU on the die (I'm not sure about it either).
Honestly I believe this is a simple mistake made by the person speaking.
You can look for RDNA1/2 ipc gains slide from AMD.

Edit: Or it could be a scenario where we are not talking exactly about the same thing it could simply be a difference between theoretiical and practical gains.
 
Last edited:

Riky

$MSFT
No, xsx is rdna 2 there's nothing contradicting this.If evidence are provided then we will talk.

You can't reason with stupidity.

"powered by AMD’s latest “Zen 2” and RDNA 2 architectures. Xbox Series X|S are the only next-generation consoles with full hardware support for all the RDNA 2 capabilities AMD showcased today."


It's even there in black and white in a statement at the actual AMD RDNA2 unveiling.
 

azertydu91

Hard to Kill
MS confirmed 25% in their HotChips presentation.
They boasted 25% performance increase in CU perf/clock from Xbox One X (GCN) to Series X.


I’m not even sure where you are getting 50% lol
Sorry I missed your edit the 50% gains is from the AMD RDNA2 slides it stated :
GCN --->+25% RDNa1 ---> +25% Rdna2.Which explains exactly why sony CU's are a 50% boost from GCN cu's.
 

M1chl

Currently Gif and Meme Champion
Sorry I missed your edit the 50% gains is from the AMD RDNA2 slides it stated :
GCN --->+25% RDNa1 ---> +25% Rdna2.Which explains exactly why sony CU's are a 50% boost from GCN cu's.
50% would be really insane jump....and I don't see it honestly, GPU is much higher clocked so that alone brings huge advantage over GCN, but on same clock, I don't know.
 

ethomaz

Banned
Sorry I missed your edit the 50% gains is from the AMD RDNA2 slides it stated :
GCN --->+25% RDNa1 ---> +25% Rdna2.Which explains exactly why sony CU's are a 50% boost from GCN cu's.
On AMD presentation they talk about “double digit” CU performance/clock increase from RDNA to RDNA2.

I understand “double digital” as bigger than 10%.
So if you take 25% + 11% (it is 11% over the 25%) that means 38% minimum GCN to RDNA 2.

AMD were not clear how much “double digit” it was.
 
Last edited:

azertydu91

Hard to Kill
50% would be really insane jump....and I don't see it honestly, GPU is much higher clocked so that alone brings huge advantage over GCN, but on same clock, I don't know.
ethomaz ethomaz Sory I missed your post I couldn't quote you in my response.

taken from an AMD slide
rdna-ameliorations.jpg
 
Last edited:
Alright that's interesting but is in total contradiction with amd's RDNA 2 presentation about RDNA 2 CU's, could it simply be a mistake from the man speaking ?
Because I think otherwise their are only presenting the same gains from GCN than RDNA 1 CU's which means that it is indeed RDNA 1 Cu's (that I don't believe) r maybe that had to make concessions to fit that much CU on the die (I'm not sure about it either).
Honestly I believe this is a simple mistake made by the person speaking.
You can look for RDNA1/2 ipc gains slide from AMD.

Edit: Or it could be a scenario where we are not talking exactly about the same thing it could simply be a difference between theoretiical and practical gains.

Well, i'm sure Xbox engineers wouldn't make that mistake at Hot Chips for deep dive technical presentation of their new shiny XSX
 

ToTTenTranz

Banned
It can't be one of the active CU's.
You ever heard of a GPU with 35 or 55 active CU's?
It's most likely, 36 Active - 3 Disable - 1 repurpose as Tempest Engine.

Why would Cerny waste CU's like that?
Cerny is smart enough not to waste CU's disabling, but rather repurpose them.
I meant to say that Tempest is most probably a completely separate block inside the non-identified area, and not within the GPU area.

Like I said, that fact that the caches were stripped down means it may look very different to the CUs you see in the GPU portion.
 

ethomaz

Banned
taken from an AMD slide
rdna-ameliorations.jpg
That slide includes the higher clock on the left side... and the right side are the account of each part for that increase.

Only 50% from these 50% come from IPC gains... that means 25%.

That is confirmed in the next slide:


images
 
Last edited:

azertydu91

Hard to Kill
Well, i'm sure Xbox engineers wouldn't make that mistake at Hot Chips for deep dive technical presentation of their new shiny XSX
That's a paradox right there, either they are too competent to make mistakes then they wouldn't have made the mistake to use rdna 1 cu's or they are incompetent and either misspoke or use bad CU's.
But in you sillogical propostion is a bit too reductive forgetting that anybody however competent they are can make on simple mistake or mispoke.Beside that engineer mispeaking falls in line way more with what amd said about RDNa2.
 

azertydu91

Hard to Kill
That slide includes the higher clock on the left side... and the right side are the account of each part for that increase.

Only 50% from these 50% come from IPC gains... that means 25%.

That is confirmed in the next slide:


images
for rdna 1 yeah and they boasted the exact same gains for RDNA2.25 + 25 = 50%.
 
Top Bottom