Could be more RSX info...

Tahir2 said:
Yes I agree the 400% figure is garbage but:

If you want to do a proper comparison with HL2 and XBOX vs PC then get a PIII 733, 64MB of RAM and a GF4 Ti 4600 is you like. I think you will find it wont work properly.

But this is a GPU comparison, not a whole system comparison. We all know that a PC needs far more RAM and CPU power to equal a console, but on the GPU side, things are fairly even. i.e. the NV2a is capable of in real terms about the same things it would be capble of if it were in a PC. Its has a slight advantage for being in a console but its not enough to take it beyond Ti4200 observed performance IMO.
 
Brimstone said:
I don't understand why Toshiba/Sony IP would end up on the GPU?

If they were going to do something custom, I think it would be more reasonable to guess on removing vertex shaders and sticking more pixel shaders. Since geometry shaders fit in between vertex and pixel shading, a GPU of just geometry shaders and pixels shaders is another possibility.

The most extreme situation I can see is

CELL = Vertex Shader
RSX = Geometry Shader + Pixel Shader

XDR2 128 bit memory pool [multithreading provides 5x the performance over GDDR-3(Plenty of bandwidth for HDR+A.A.)]


Thats a possibility but Nvidia have already stated that RSX will include pixel and vertex shaders in the GPU. But of course a dev can still make Cell send vertex calculations to the GPU.

Of note, the RSX will also include these so called Multiway programmable FP shaders pipelines independent of the vertex/pixel shaders. how do you think they will fit in the GPU design?
 
pjbliverpool said:
But this is a GPU comparison, not a whole system comparison. We all know that a PC needs far more RAM and CPU power to equal a console, but on the GPU side, things are fairly even. i.e. the NV2a is capable of in real terms about the same things it would be capble of if it were in a PC. Its has a slight advantage for being in a console but its not enough to take it beyond Ti4200 observed performance IMO.

I think Xen is arguing both ways for the GPU and the system comparison - that is what my posts refer to, but Xen (hi Xen) seems he to be unsure a little as to what his exact parameters for discussion are.
 
Xen said:
Of note, the RSX will also include these so called Multiway programmable FP shaders pipelines independent of the vertex/pixel shaders. how do you think they will fit in the GPU design?
Sorry, what makes you think they weren't talking about Pixel Shaders?
 
Joe DeFuria said:
Doesn't every GPU aside from Xenos have independent pixel/vertex shaders?

I dont think the term independent means the pixel and vertex shaders are independent (which i'm sure they are even in RSX)

Its saying the conventional vertex/pixel shaders in todays pc and in the RSX will be independent from these so called Multi-way parallel FP shader pipelines.

Its like this

Vertex shaders pipes
Pixels shaders pipes
Multi-way parallel FP shader pipes

All independent from each other

How they all work in the GPU god only knows (well Nvidia and sony)
 
Tahir2 said:
Anyway.. it is all PR rubbish and certainly some people are going to be disappointed with the un"presidented" levels of performance.

Not if more games look like Medal of Honor: Airborne.;)
 
Last edited by a moderator:
4 GPU's

4 Cell's is different than 8 cores...

The RSX may have four cores (in one GPU). So all this talk about 4 GPU's is kinda senseless.
 
Xen said:
I dont think the term independent means the pixel and vertex shaders are independent (which i'm sure they are even in RSX)

Its saying the conventional vertex/pixel shaders in todays pc and in the RSX will be independent from these so called Multi-way parallel FP shader pipelines.

Well I think that sums up the merit of your argument quite nicely. Amusing to say the least.
 
Xen said:
Thats quite a contradictary post. First you say Derek Perez didnt say RSX had the equivalent performance to a Quad SLI. Then you say PS3 will have similar graphical quality to a top end pc, which right now is a pc equiped with Quad SLI.

And your right it is absurd to think PS3 would contain 4 GPUs and no sane person here are making such wild claims. I'm saying like you said yourself the performance will be similar.

Actually much better when put it into context of a closed system with huge bandwidth benefits like the PS3.

Lets break it down short bus style. Graphics performance is pretty simple.

So scenario 1, equivelent performance as 4 GTX512 cards in SLI:
Die Size ~800 sq mm. Die cost, $400+. PS3 Cost ~1K

So scenario 2, equivelent performance to 1 GTX512 card:
Die Size 200-240 sq mm. Die cost $50-100. PS3 cost $500-600.

Pick one. I'm betting if its scenario 1 you'd make a bloody fortune shorting Sony.

Aaron Spink
speaking for myself inc.
 
Dave Baumann said:
Oh dear lord...

Your very welcome to have your own opinion on this matter. But its seems less of a discussion here and more of just labeling anyone with an opposing opinion crazy.

Its like

"hey guys there could be something not discussed about RSX"

"Its a G70+connection to cell if its more it would be in a PC"

"But what about these multi-way parallel FP shader pipes"

"Its a G70+connection to cell if its more it would be in a PC"

"They're independent from pixel/vertex pipes they said each can do 136 shader ops a cycle in parallel"

"Its a G70+connection to cell if its more it would be in a PC"

"oh forget it"


I'm only trying to point you guys to the slides from E3 and the accompanying speech about RSX that though admittingly vague does atleast backup what i'm trying to say. Its obviously all open to interpretation at this point as there's no hard facts. But if you think i'm wrong then look at the same E3 slides and listen to Jen-Hsun Huang's speech again and tell me what you think.

You'll find the slides to me confirm these Parellel FP shader pipelines are independent from the conventional vertex/pixel pipes on RSX (its there in black&white)

You'll find Jen-Hsun Huang's speech confirms their theorectical power "each one of these FP shader pipelines can process 136 shader ops simultaneously"

And to me the very fact he's discussing these processors/pipelines in isolation to me reinforces they're independent from the pixel/vertex shaders.

And the fact he says "SIMULTANEOUSLY" confirms to me there's more than ONE

I mean for gods sake, he explains there will be a "farm" of these things in the RSX his words not mine.

So in summary for those wanting to discuss/point and laugh...

RSX will include Multi-way parallel programmable FP shader pipelines
RSX will have independent pixel/vertex pipelines
(yes independent from the pipes above)
RSX will have a "farm" of these programmable FP shader pipelines
(meaning more than 1...)
RSX will be able to calculate 136 shader ops per cycle in parallel from each of these programmable FP shader pipes simultaneously
(again must mean more than 1...)

If you dont believe me fine i know i'm sticking my neck out here, but the information is freely there for anyone to find ive just got the infomation there is and this is my interpretation.
 
Xen, Regardless of whether it says the VS/PS are independent does it mean to say that these are different from the "Multiway parallel FP shader pipelines". What is, for instance, a group of pixel shaders if they aren't inhernatly "Multiway parallel FP shader pipelines"? I could equally describe 6 quads of shader pipelines as a "farm" with each quad being a "Multiway parallel FP shader pipeline".
 
Xen said:
RSX will include Multi-way parallel programmable FP shader pipelines
RSX will have independent pixel/vertex pipelines
(yes independent from the pipes above)
You've taken this quote in the wrong context. The context is not "independent from the pipes text above" but indepentant from each other (i.e. they are not unified).
 
Back
Top