Parhelia!!!

[xortam]

I don't doubt one can build a reasonably priced shared memory processor board. We did this back in '84 (not a graphics board). How are you going to address the other issues I brought up superfly? Remember, John want's to run his Window's games on this board, or so we would assume.

[Wombat]

I don't see any difference in the technology advance needed for this configuration than say the technology of today's video cards as seen say from five or six years ago. A Matrox g550 could have been built six years ago, but as you say it would have cost $8000 and may not have fit in a pci slot. Free your mind Wombat.

You're looking at things wrong. Sure, in 5 years a handheld will likely beat the pants of my current computer system, but think of what an entire <I>system</I> can do in 5 years. You'll be clamoring for that instead. Just because today's sedan outperforms the old Gran Prix racers, do we stop longing for today's race cars?

Imagine what could be.

That's my damn job.

I think it would be possible to engineer a video card with it's own cpu,and it wouldn't cost a fortune either,a 0.13 micron version of the p3 measures out to less that 80 sq mm,which makes it very cheap anyhow and integrate it into a shared memory architecture like the X-box uses...
The system's cpu would only be used for HDD,cdrom access as well as sound issues and perhaps still use it's memory for data that isn't bandwith intensive(if needed at all,of course)...

Of course, the X-box IS a PC.
Some of you will remember when Sun came out with the PC on a PCI. K6-2, memory, sound, etc. all on a PCI card, so that your Sparc could do some PC jobs for you. Just remember that it was quite outperformed by an actual PC bought at the same time, and cost a good bit of coin anyway.

Price, performance, environmental constraints: pick any two.

[superfly]

Originally posted by xortam
I don't doubt one can build a reasonably priced shared memory processor board. We did this back in '84 (not a graphics board). How are you going to address the other issues I brought up superfly? Remember, John want's to run his Window's games on this board, or so we would assume.

Well,like i said,the game woudn't have to be stored entirely within the graphics card memory either and would use external components such as the Cd-rom,HDD,sound card and perhaps local system memory for less bandwith intensive issues...


The performance would likely be much higher that any pc with an equivalent processor and graphics chip,when used within the standard pc architecture,look at how some games look awsome on the x-box,even though it's using a fairly low end cpu(celeron 733) by today's standards....


As far as cpu power goes we're not not close to using our cpu's full potential in games precisely because of bus speeds/bandwith/contention issues...


Physics and AI calculations aren't even close to stressing even mid range cpu's and even today,only 15~20% of cpu time is dedicated to those issues,with the largest slice going towards graphics,imagine if developers use all that power for AI and physics....


The real thorny issue would be on the software end,not hardware wise,execpt of course for Intel and AMD losing sales from gamers not needing to upgrade their pc as often as before....

[xortam]

So we're back to the coprocessor design then aren't we? I brought this up as a possibility over a year ago. The question I had was at what interface does it make sense to break up the functionality. I don't think you want to emulate a Windows environment on the graphics card but instead process at the DirectX level or below. The card can be upgraded via a downloadable control program to support changes in the interface. It would be desirable to support more than DirectX. I don't know enough about the 3D graphics subsystem to propose the right interface that would make sense from a performance stand-point. If you try to actually run the game code on the card then you've got yourself a big can of worms in supporting quite a bit of Windows functionality.

[Wombat]

A good bit of the X-box speed comes from it having a video chip that isn't even out in the public yet. A very sizable chunk of speed comes from the very tightly compiled custom DirectX on the X-box. If you know exactly what your hardware is, you can get a considerable increase in software speed.

[xortam]

M$FT is in a position where they can develop a new platform and dictate to game developers to develop for that platform. Video card manufacturers don't have that luxury. Look at all the chaos already over perversions in the APIs. Aren't they all fighting towards DX9 to help alleviate some of this special versioning? Matrox isn't in a position where then can dictate but minor variants in the code base.

[superfly]

Well sooner or later,even if general purpose cpu's will never get integrated into video cards,the graphics chips themselves will end up handling everything anyhow,leaving only game code that's not directly related to graphics,handled by the cpu...


At the speed they're introducing smaller fab processes,i woudn't be to surprised if graphics chips in 3 or 4 years time end up having upwards of 300 million or more transistors in their desing,making a GF4 card look like a voodoo 1....

[roadie]

sorry i just had to laugh at this:

Free your mind Wombat. Imagine what could be. Do not be stuck here in what is.

been watching the matrix recently???

anyway i think this coprocessor thing would just be too expensive, i think refining todays cards technology would be more beneficial until there are serious breakthroughs in game technology

[knirfie]

Hey,
Onewhen playing a game there's a lot more going on than just graphics, there are also things as A.I., sound is also becoming more and more work, input, output (multiplayer), this idea just wouldnt work.
Secondly, by uploading the game to the memory, you'd get restricted to using that amount and nothing more, think of the price, you'd need fast memory, for storing the data, for processor to work with, with the next generation games coming up, you'd need at least somewhere around 500mb of a few hundred mhz memory, accesible at the same time, by the system, the (co)processor, and the GPU.
third, the price would be a problem, Memory, GPU, Processor, all need to be designed, made, etc, think in the price range of around $2000 a card (this would probably be a minimal)

This is a idea i had been toying with when i started studying Electronics and got my first lesson in computer technologies, and believe, it just wouldnt work.

[GT98]

Originally posted by Wombat
A good bit of the X-box speed comes from it having a video chip that isn't even out in the public yet. A very sizable chunk of speed comes from the very tightly compiled custom DirectX on the X-box. If you know exactly what your hardware is, you can get a considerable increase in software speed.

I thought that the NV2 in the Xbox is pretty much the same thing as the Geforce4

Scott

[Nappe1]

Originally posted by GT98


I thought that the NV2 in the Xbox is pretty much the same thing as the Geforce4

Scott

it is NV2A on XBox... and yep it is pretty much same. differencies are different clock speeds, and more efficient memory controller.

so NV2A is something between NV20 and NV25

[Stringy]

As with most good ideas... it's been DONE!
you should have jumped on my Buddies eBay auction a little while ago if you wanted such hardware...

3DO, add in card, did EVERYTHING, it was like adding a console to your PC in one board...
Seeing as how everyone remembers that it existed I'm suprised that it died and is no longer developed for....

I want a PC, not a Console... there are alot of console out there that fill that market need well enough


Craig

[Evildead666]

i thought the graphics processor on the xbox had mBGA memory placed directly next-to or on-top of the graphics chip? it doesn't seem to have a 'card' in it...
i have seen this type of design for laptops, to reduce space, but it also reduces the distance between the memory and the gpu.
this should enable faster data access between gpu and memory. ( in theory).

Also, i also read about some SGI workstations that use UMA design for the higest quality graphics.
Universal Memory Architecture. All the memory is shared between the Graphics/Sound/CPU/System.
If thez graphics need 500Megs, then they can have it...as i remember this computer could allocate up to 1Gb of memory to the graphics.

Of course this type of system would be hideously expensive, since u would need some very fast RAM (but at least the whole system would benefit.).

Parhelia WILL kick ass, but we just haven't figured out how yet...if we had, ati and Nvidia might get wind of it, and try to modify their drivers etc to destroy the impact. It is all so quiet...like a quiet before the storm.
Matrox will let everyone know in one big WHOOMPH!
it will be more of a spectacle than leaking out info every week, no?

just my thoughts...co-processor...think not.
Dual chip design may be good for the pro version, sort of like an SLI card in one with 256Mb ram...

[Dr Mordrid]

It'll be a big WHOOMPH! alright. Jeeeezzzzzzzz....are some people in for a shock.

Dr. Mordrid

[knirfie]

What if we just place the memory in the GPU, that should solve a whole lot of bandwith problems