Ati and Nintendo GC successor GPU

[Ooh-videogames]

GameCube 2
Dear Matt, What have you found out, if anything, about the successor to GameCube?
Matty

Matt responds: A fine question. Well, I don't have enough to run with a news piece, but after talking to several insider pals around the industry here's what I've been able to come up with:


The console is being referred to as GameCube 2 internally
ATI has been in development with a graphics chip for "GCN 2" for more than a year; it's supposedly coming along very nicely
Several games are already being designed with the system in mind, though no hardware is yet available
The hardware is expected to ultimately ship at a $299 price point -- directly against Xbox 2 and PS3
The CPU partner has not been set, though the obvious choice is IBM
2005 release is likely as Nintendo wants to beat the competition to the gates
I hate to lend credit to this ongoing rumor, too, but I figure I can't ignore it. At least one industry insider I spoke with brought up the whole Microsoft/Nintendo union; apparently Microsoft -- more than Nintendo -- is interested in some kind of partnership.

There are a few other smaller bits here and there, but I don't feel comfortable posting them yet. Note that while I do trust my sources on this information (they've come through dozens of times in the past), it's still so early on that anything can happen.

You can take what you want from this. I think the possibility of backwards compatibility are very high. I wonder if Nintendo going to use a derivative of the Mac G5 processor, thats if they go with IBM.

 

[Deepak]

ATI=GC2
ATI->XB2
NV->PS3
NV->XB2
IBM=PS3->GC2

?????

 

[CeiserSöze]

If the chip has been in the works for over a year why did ATi announce the new Nintendo-deal just a few months ago?

 

[London Geezer]

If the chip has been in the works for over a year why did ATi announce the new Nintendo-deal just a few months ago?

WELL.... thats not the point....

the point is:

(IF THE INITIAL POST IS ACTUALLY TRUE, AND NOT SOME STUPID SPECULATION)

how can they get IBM to do their CPU, when Cell is going to be the main project for them from 2004?

 

[Paul]

The CPU partner has not been set, though the obvious choice is IBM

I doubt that.

If IBM wasn't working for 4 years on the PS3 chip than I would say yea, but now.. no just no.

 

[cybamerc]

IBM is a huge company with many, many teams. It is more than capable of working on multiple products and has already expressed interest in working with Nintendo again. This after work on Cell began.

I doubt Cell will be IBM's main priority either. It will be one among many different CPU configurations it will support and develop.

 

[DeanoC]

IBM is a huge company with many, many teams. It is more than capable of working on multiple products and has already expressed interest in working with Nintendo again. This after work on Cell began.

I doubt Cell will be IBM's main priority either. It will be one among many different CPU configurations it will support and develop.

Absolutely.

The IBM PowerPC970 (Used in Apple G5 systems) is ready to go, and would fit perfectly in a new Nintendo system. Already at 2 GHz, should be hitting 3-4GHz in GCN2 time scale. It has a 1 Ghz bus, fully 64 bit and IBM can easily customise for Nintendo's system. It would easily be backwardly compatible with the PowerPC core used in GCN, so plays GCN games cheaply.

Also using the existing partnership with ATI and IBM will mean they get good deal, tools and libraries etc. Which will be a factor in the next generation systems.

Sounds good to me.

 

[London Geezer]

ok i see... still how is a PowerPc970 going to compare to Cell? PS3 will most certainly have a total of 1Tflops spread between CPU and GPU as we've seen...

the thing is, as big and good as they are, the major investment put into Cell is enough for me to censider it the "main" project IBM will have in the future, especially considering that it will be included in EVERY Sony appliance released, not only PS3...

 

[MfA]

IBM didn't exactly do Cell for Sony alone BTW. They have the rights, and are intending, to use it themselves too ...

 

[Panajev2001a]

I do not think that Sony would be that angry if another console manufacturer increased the chances of Cell finding success in the consumer electronics market and paid them some license fees ( they would go to Sony and Toshiba too and not only IBM )...

 

[Entropy]

A 2005 derivative of the PPC970/980 is likely to be roughly as powerful as whatever Microsoft can put into the Xbox2. Nintendo will have a significant advantage being less tied to a PC kind of architecture, so overall odds are that Nintendo should be able to match or exceed the XBox2, given similar hardware costs. Of course, Microsoft could accept substantial losses on the XBox2 as well, but there still isn't too much more they can do, being limited by their PC compatibility agenda.

The real life abilities of the PS3 are much harder to predict.

Entropy

 

[BlackAngus]

how can they get IBM to do their CPU, when Cell is going to be the main project for them from 2004?

Ibm is huge, only a small % of its resources are directed towards Cell. Cell is probabaly a small thing compared to the Power5 RnD and support. Ibm could easily get the big N an IBM 970 series processor. Those wont be that expensive to produce in 2005, possibly even a cut down Power5 processor by then.

IBM didn't exactly do Cell for Sony alone BTW. They have the rights, and are intending, to use it themselves too ...

That would indeed be interesting. Wouldnt that mean that MS and Nintendo (assuming they choose cell) would be paying Sony royalties? =)))))

Personally I doubt that Cell will appear in a second console in the next generation. Only sony this time around.... after that however maybe a different story.

 

[cybamerc]

london-boy:

> still how is a PowerPc970 going to compare to Cell?

Cell is a scalable design and assuming that it will be used in PS3 (which is far from given) we don't know what the exact configuration will be. Moreover, Cell performance will rely greatly on the quality of the development environment. I think there is a very real possibility that Cell will suffer from massive overhead resulting in real world performance far below its theoretical maximum. It will take quite a while before compiler technology is up to speed.

From that perspective going with a simpler (and cheaper) design might actually pay off.

Moreover, if Nintendo, or someone else, wants more power than what IBM's standard selection of Power PC CPUs have to offer I'm sure IBM would have no problem boosting performance by modifying the design. Perhaps by using multiple processor cores Power4/5 style. Adding more AltiVec units could also be a solution (why it could be a regular mini-Cell. Perhaps they could call it "Coalition"). There is a reason why Sony chose IBM as its partner after all.

> especially considering that it will be included in EVERY Sony appliance
> released, not only PS3...

Well... IBM won't have much to do with that. IBM will use Cell themselves in servers at some point, but not exclusively. Development on Power6 is underway and they will still support Intel and AMD. And then there's Power PC of course for consumer products.

 

[Ooh-videogames]

Could Nintendo be using NEC's Cray processer if its really in development? If ATI really started on the GPU a year ago would it still be able to match the Xbox2 and PS3 graphics power and features.

 

[Paul]

Care to elaborate how Cell being in ps3 is "far from given" despite dozens of currrent articles and Sony stating it would be inside the system? I could have sworn SCEI which is the Playstation branch mind you is going to manufacturer the chip.

 

[PC-Engine]

I guess if N wants backwards compatibility then they can go with a dual core G5 or G6. I don't know if that will be enough cpu power for N5 though. Not only that but it will run pretty hot too.

http://www.eetimes.com/sys/news/OEG20030623S0092

 

[Vince]

Ibm is huge, only a small % of its resources are directed towards Cell. Cell is probabaly a small thing compared to the Power5 RnD and support. Ibm could easily get the big N an IBM 970 series processor. Those wont be that expensive to produce in 2005, possibly even a cut down Power5 processor by then.

Actually, STI-Austin was founded by many people from IBM's T.J. Watson Research Center, AFAIK. Many of the Cell researchers came off of working on next generation PowerPC design and such related projects. Infact of IBM's 5 major project leaders for cell, I believe (don't quote me) that 4 of them formerly worked on PowerPC or other high preformance architectures. The other one is an academic (formerly of IBM server group I think), who has spent his life on promoting the cellular computing ideal and is a widely known fanatic (genious?) in the Forth community.

While Cell isn't a massive cancerous growth that's eating up all of IBM's talent - it's hardly the little engine that could.

 

[MfA]

I am reasonably sure the IBM Chuck Moore is not the ColorForth Chuck Moore.

 

[V3]

Could Nintendo be using NEC's Cray processer if its really in development?

Nintendo better off, using a modified of this Power PC, than what NEC has to offer.

If ATI really started on the GPU a year ago would it still be able to match the Xbox2 and PS3 graphics power and features.

I don't see, why it can't.

 

[Vince]

I am reasonably sure the IBM Chuck Moore is not the ColorForth Chuck Moore.

Well then the coincidence's are amazing between Cell and Moore's earlier c18 and 25X chips aswell as the last I heard of him pre-cell we was at the University of Texas - Austin doing think-tank type R&D on silicon based architectures in the second decade of the century.

Lets put it this way, when I heard the name attached - I didn't even think it would be anybody but the phycho Forth proponent. Hell, Forth is even best suited/targeted towards these type of "limited" resource architecture as would be seem in each APU or whatever it's called on cell when you strip down the functional units to the minimum core and want to run threads/microprograms in each unit. No?