Talk in London about Xenos

[bbot]

Xenos is equivalent to a 32-pipe gpu. It's WGF 1.0, not WGF 2.0 (can't create vertices).

I think I'll go back to PC gaming. No PPU, no WGF2.0.

http://www.bit-tech.net/news/2005/05/25/ati_xbox_360_london/

 

[Johnny Awesome]

 

[Carl B]

Well of course it doesn't have WGF 2.0 - the standard hasn't even been finalized yet, how could it include it?

 

[benjamin]

Is that like Warewolf in London, just with GPUs?

 

[X-AleX]

I'm not sure of this, I've already read somewhere else that the chip would have supported WGF 2.0.

 

[dukmahsik]

ATI revealed this afternoon that every single Xbox 360 game will be required to run at a minimum of 720p resolution with 4x Anti-Aliasing.

Speaking at a press conference in London, ATI technology specialist Rene Froeleke told journalists that Microsoft had specified that games must run at the 1280x720 resolution at 4xAA with no slowdown. Every single game will be supported at this graphical specification, which we think is great news for gamers playing on big-screen TVs.

When questionned as to why the minimum wasn't 1080p, Rene responded that most of Europe hadn't even reached 720, let alone 1080.

Because of the embedded DRAM architecture of Xbox 360, the AA was described as 'Absolutely free'.

The technology fulfills all the criteria of Windows Graphics Foundation 1.0, aka DirectX 9.0c, which includes Shader Model 3.0. However, it lacks features which are required for WGF 2.0, most notably the ability to create vertices, rather than just shade and manipulate them.

Whilst the architecture of the 360 is not comparable to a PC design, Rene said that the raw performance would be roughly equivalent to a 32-pipeline PC graphics card.

The shared memory technology was 'Definitely not comparable to Hypermemory', ATI's shared memory technology for the PC.

The chip doesn't have an official ATI name, since ATI aren't marketing it. Whilst the codename is R500, we were told that the name meant 'As much as the code on my suitcase' in terms of relevance to existing architecture.

 

[marconelly!]

Whilst the architecture of the 360 is not comparable to a PC design, Rene said that the raw performance would be roughly equivalent to a 32-pipeline PC graphics card.

Does such a card even exist? Doesn't 6800U, and upcoming G70 and/or R520 have only 24 pipelines (or do vertex pipes count in also?)

 

[Dave Baumann]

Actually, verts can be created and deleted on the graphics, although its a multipass op.

 

[AlNom]

Does such a card even exist? Doesn't 6800U, and upcoming G70 and/or R520 have only 24 pipelines (or do vertex pipes count in also?)

Could be referencing SLI performance.

 

[Carl B]

I think he's (Rene) just going on the prior speculated breakdown (obviously internal to ATI this time) that tries to compare it to a conventional card by translating it's unified architecture into 'pipe' equivelents using the percentage of assumed time/cycles doing shader vs vertex ops.

 

[benjamin]

Does such a card even exist? Doesn't 6800U, and upcoming G70 and/or R520 have only 24 pipelines (or do vertex pipes count in also?)

Could be referencing SLI performance.

The 6800U has 4 fragment processors that each operate on 4 fragments at a time, leading to the common reference of 16 fragment/pixel pipelines.

The new G70 is slated to have 24 I believe. I'm not sure about the R520....

 

[phat]

Actually, verts can be created and deleted on the graphics, although its a multipass op.

So the answer really is no?

 

[Acert93]

Actually, verts can be created and deleted on the graphics, although its a multipass op.

Why must the booming voice from the heavens come down and taunt us so!




Thanks for the info Dave, that was one of my Xenos questions! This must be one of the features ATI had in mind when they mentioned the chipset features being "SM 3.0+". Your article is going to be great Dave, thanks for the update

 

[DeanoC]

Actually, verts can be created and deleted on the graphics, although its a multipass op.

So the answer really is no?

No, it can create vertices but its not as flexible as WGF2.0 was originally talking about. Its basically a fixed function tesselator (however R500's architecture means its actually semi-programable, I'll leave Dave's article to explain...)

 

[marconelly!]

The 6800U has 4 fragment processors that each operate on 4 fragments at a time, leading to the common reference of 16 fragment/pixel pipelines.

The new G70 is slated to have 24 I believe. I'm not sure about the R520....

Ah, true. So he might be referring to SLI-ed 6800s... It's surprising to me how much Ati people seem to be downplaying the raw power of this chip when talking about Xbox 360, and playing up it's cost effectiveness or power saving or elegant design or whatever. I wonder why they keep doing that, when the chip seems very capable. Do they actually expect RSX to outgun this by a far, because to me it doesn't seem that will happen if it has 24 pipes and R500 is equivalent to 32...

 

[tEd]

Actually, verts can be created and deleted on the graphics, although its a multipass op.

So the answer really is no?

No, it can create vertices but its not as flexible as WGF2.0 was originally talking about. Its basically a fixed function tesselator (however R500's architecture means its actually semi-programable, I'll leave Dave's article to explain...)

A semi-flexible fixed function tesselator. Now no matter how good it actually is it sounds absolutely amazing

 

[ERP]

Actually, verts can be created and deleted on the graphics, although its a multipass op.

So the answer really is no?

No, it can create vertices but its not as flexible as WGF2.0 was originally talking about. Its basically a fixed function tesselator (however R500's architecture means its actually semi-programable, I'll leave Dave's article to explain...)

I was just thinking about this and I think you might be able to do completly generic tessalation in the shader. Although I'm not sure how well you exploit the parallelism at that point.

 

[Titanio]

The 6800U has 4 fragment processors that each operate on 4 fragments at a time, leading to the common reference of 16 fragment/pixel pipelines.

The new G70 is slated to have 24 I believe. I'm not sure about the R520....

Ah, true. So he might be referring to SLI-ed 6800s... It's surprising to me how much Ati people seem to be downplaying the raw power of this chip when talking about Xbox 360, and playing up it's cost effectiveness or power saving or elegant design or whatever. I wonder why they keep doing that, when the chip seems very capable. Do they actually expect RSX to outgun this by a far, because to me it doesn't seem that will happen if it has 24 pipes and R500 is equivalent to 32...

Given that R500's "32 pipes" have to handle vertices as well as pixels, I think it would be fair to count in vertex shader pipelines when comparing with other cards, no?

And not being WGF2.0 compliant is a no brainer, as of the last WinHEC conference it still wasn't finalised.

 

[ralexand]

So does the RSX have a tessalator also? Is RSX WGF 2.0 compliant in a functional since? And what's the advantage of WGF 2.0?

 

[Titanio]

So does the RSX have a tessalator also? Is RSX WGF 2.0 compliant in a functional since? And what's the advantage of WGF 2.0?

Not sure if RSX has a tesselator, but if not, that's something a SPE could be very good at

I doubt RSX is WGF2.0 in a functional sense either (the specification not being finalised aside). It's beyond WGF 1.0 probably.

WGF2.0 is basically the "real" DX10. Programmable tesselator, SM4.0, virtual memory for the GPU etc. (see: http://www.beyond3d.com/articles/directxnext/).