ATI & XBox

Well, we will know soon enough what to expect specifically in the Xbox2. Microsoft is talking to devs about the capabiilties of Xbox2, namely DX10. I also wouldn't be surprised if a lot of other devs give major imput for the OS in Xbox2.

Dave, do you know if the GPU is supposed to have eDRAM or not?
 
Is anything in particular barring you from telling us?

If "yes", then what would that be?

If "no", then stop being smug and spill the beans dammit! :devilish:
 
Just wondering... Say you have 2 chips that, power wise, are very similar (MHZ, pipelines, ram, transistors, etc...). One is only DX9 compliant and the other supports DX10. How big of a difference would that make?
 
I do know whether or not the X "VPU" ( ) will have a chunk of onchip storage.

Ahh I see, so Xbox 2 GPU from ATI, is going to be configure like the recent announced P20 ? with VSU + VPU combo. Thanks Dave.
 
How much of an advantage would fsaa have if it was stored in say 64 megs of edram or 32 megs of edram. This would certianly remove the bandwitdh bottle neck to higher fsaa modes wouldn't it ?
 
V3 said:
I do know whether or not the X "VPU" ( ) will have a chunk of onchip storage.

Ahh I see, so Xbox 2 GPU from ATI, is going to be configure like the recent announced P20 ? with VSU + VPU combo. Thanks Dave.

Why would the Xbox 2 need ATI to provide a VSU if the CPU's have Vertex Shaders already? Unless you call the IBM parts VSU's instead of CPU's.
 
Why would the Xbox 2 need ATI to provide a VSU if the CPU's have Vertex Shaders already? Unless you call the IBM parts VSU's instead of CPU's.

Who knows, the CPUs might still not be fast enough and still needs help.
 
It's a pretty know fact that XBOX2 GPU will have embedded dram in the game developers community..
M$ has already started a complete support for their Xenon effort, with specs, xdk and related newgroups for selected developers.
 
I guess there is no way that we, the unwashed masses, can peek into those said newsgroups :devilish: that would be fun though.

I'd be severely disappointed if Xbox 2 ended up with no edram on its
VPU (s?) not counting the usual small caches found in every graphics processor / chip.


what does VSU stand for and do? I know it's something to do with the 3Dlabs P20, is it some sort of interconnect for two VPUs?


what I am really curious about is, is the XBox 2 graphics sub system being designed by ATI alone, or is Microsoft participating actively, other than just drawing up desired specifications. I really want to know if MS's acquired graphics assets are being put to use (they apparently were not with Xbox/NV2A) ...want to know what the Cagent people are doing.
 
Interesting... is ?ati? doing some damage control or why would r500/r600/r666 ;) be totally new part? or is this r420 thing mainly old stuff with speed improvements and then very soon some really new architecture with bells and whistles ?

For given timeframe and cost it would seem likely that every vendor will launch similar offering if the cost per device is the same. So it would seem likely that everybody has similar hardware on the market at the same timeframe, assuming similar factories producing devices and similar focus on feature ? How does sony, toshiba and ibm putting a lot of money to producing chips(process technology, factories) count here? is MicroSoft putting some money also to advanced technologies or are they going conservative and producing on 0.9 at 2005?

off topic... how does sony seem to be in the sence that they currently have the much "disliked" multi chip console. Is it advantage when next generation seems to be all multichip? e.g. Has sony&sony developers learnt how to do multichip development and tools. How does this compare e.g. to microsoft providing support for multiprocessor platforms in windows(or dual pci-express cards)

and Off topic2.... if Tim Sweeney is wondering how to program ps3, what does he feel about 3*cpu*dual core + edram + gpu configuration for xbox or has he changed his opinion?
 
Embedded ram is integrate ram inside a chip if thought in a simple way. The good thing is that it can be clocked equally to the actual chip and it is easy to have really wide data paths from processing unit to E-Dram inside same chip. So think of it as really fast ram(don't forget about cache's though that can used to e.g. hide latency from random access to memory(cache->-e-dram-memory-disk-etc) ).

for example sony's playstation 2 has 4MByte of E-Dram that has about (total) of 48GB/s bandwidth.

E-Dram could now days be a lot faster or slower depending why it is put into the chip...
 
Back
Top