MS designing their own CPU and GPU for Xbox2?

[randycat99]

...not to mention all that knowledge base built around making fancy pixel shader effects on nVidia hardware (circa Xbox 1) will go out the window when you have to re-figure how to do that kind of stuff again on new PowerVR hardware.

My question from before still remains- what earthly motivation is there for MS to suddenly switch camps from a well established x86/nVidia GPU architecture for 2005?

 

[Lazy8s]

randycat99:

Realistically, why would a hardware "newbie" like MS suddenly be able to whip up a "better" CPU than Intel/AMD and a "better" GPU than nVidia/ATI in a mere 2 years? That is a tall, tall order no matter how you look at it. Otherwise we would be seeing companies such as Intel/AMD/nVidia/ATI being "rolled over" as a matter of routine. They are on top of the game and stay there because they have a great deal of intellectual expertise on the subject.

A company could design a better chip, but that doesn't mean a chip-making mainstay like Intel would be rolled over. The computer makers like Dell and Compaq have already established relationships with Intel, they sell their computers partly on the strength of the Intel name, they get pricing deals on the chips that only large companies like Intel can afford to give because of mass production and efficient fabs, and Intel leads many of the PC industry's standards and trends.

In the PC market, there are backward compatibilities and legacy issues to be maintained in every product, so a new company couldn't just expect the market to embrace their superior, custom solution just because it was better. For example, Sony couldn't just expect the whole PC industry to drop Wintel and follow their newer EE+GS architecture just because of its benefits. The new chip has to be restricted by the same compatibility requirements to be able to work with the existing tools, software, hardware, and architecture the PC market is geared to. Intel themselves could design a better CPU if they ditched the compatibility requirements and developed an entirely new architecture. But, that's not the way an ongoing, open platform like the PC market works.

And then you even have companies like AMD, whose chips do adhere to the same compatibility requirements as Intel, who have a hard time competing with the giant due to Intel's industry clout and the fact that so many more things are built with the Wintel standard in mind. Intel would naturally defend their market dominance against any new challenger, superior tech or not, by leveraging their support, marketshare, vast production resources, and finances to undercut them in some way.

The Xbox is a closed platform, so Microsoft could technically develop their own custom chip solution. However, I think Microsoft will push hard to maintain the Direct-X roots of the "X"box and keep it in line with at least some PC conventions.

 

[Mulciber]

It's going to be very hard for MS to maintain backward compatibility if they go down the 64bit CPU path or/and chose PowerVR. From what I heard, there are inherent problems with TBR and DX9 that prevents it from getting full WHQL certification. Unless they somehow get Intel and NV together and shrink the CPU/GPU into one.

Whats the odds that MS is working with PVR right now to standardize DX10 around a TBR design to be used in the Xbox2?

It wouldn't have to be backwards compatable with DX9 unless they actually brought the chip to the PC market. So it would only need to be compatable with DX8 and DX10.

hehe, just food for thought. 8)

 

[Panajev2001a]

chances are that PVR would want to bring it to the PC market...

 

[randycat99]

randycat99:
A company could design a better chip, but that doesn't mean a chip-making mainstay like Intel would be rolled over.

Ok, I should have been more specific. I had presumed that the Xbox2 would still be x86 based whatever it uses. So my question was aimed with context of a "better" x86 CPU. I still don't think it could happen within the 2-3 year window.

Now that we are hashing about ideas, is it even plausible that MS would even consider a different CPU architecture and ISA for XBox2? Isn't that even more farfetched than the original premise (I thought) of an outsourced chip but still based around an x86 compliant environment? I didn't even realize that was being premised here, but is it really what you guys are contemplating? Wouldn't that turn everything on end for the XBox2, including the GPU it would be driving?

 

[Saem]

CPU architecture and ISA for XBox2

These are the same thing.

 

[Mulciber]

chances are that PVR would want to bring it to the PC market...

Well, if they can't get enough foundry time then the point would be moot. Thats the reason the PVR 250 was a year late to market.

Then again, with the ability to do nearly anything you want with a pixel in DX10, it wouldn't be hard to emulate any of the DX9 effects perfectly through DX10 with the proper drivers.

 

[Lazy8s]

randycat99:

So my question was aimed with context of a "better" x86 CPU.

Sticking to that context, outdesigning Intel would only be part of the battle. They'd still have to compete against the entrenched incumbant Intel on the business side.

The problem here is that they wouldn't be likely to gain much acceptance, and thus they wouldn't be producing these better x86 CPUs in anywhere near the volume as Intel. Without a large outlet besides the Xbox2 project in which to sell these superior x86 CPUs, they'll have nothing to offset the risk like Intel can with their multiple market dominances.

About the possibility for a custom, non x86-compliant solution for Xbox2 - Since Microsoft's corporate focus exists strongly in both the PC and console markets, it's hard to see them going with a solution that doesn't have at least some crossover between the two. However, the increasing pressure from rival Sony and the fact that Xbox is technically a closed platform might make MS more willing to make concessions and compromises as to the level of similarity between the two platforms' architectures as time goes on.

 

[Magnum PI]

Yeah, outsourcing seems like the only way they could come out ahead with the "custom" approach, but that assumes there is some vendor out there that actually has a better architecture to offer than Intel/AMD/nVidia/ATI. If there is, then why didn't they use them for Xbox 1?

because this approach takes more time, and they didn't have this time for the xbox 1..
so they took parts that were designed for the peecee world, who advantage was that it was ready.

I just don't see who that could be, but I could be wrong. Frankly, I don't see how anyone could come out with something custom in 2/3 years that would beat Intel/AMD/nVidia/ATI in their own game by a large margin to make it worthwhile.

if they want to keep the x86 compatibility, intel and amd aren't the only choices: transmeta, via...

for the GPU i don't see why no other company couldn't design a chip that is as suited for this task than nvidia or ati one.. look at the arcades, i would rather say that powervr can do great things, see naomi2.

performance is only a factor, there is also cost, thermals (you don't want your console to be as noisy as the latest geforcefx..), and needs of a console GPU are different than a PC GPU, ie: the resolution and the framerate of the tv that are very different of the pc.

 

[Magnum PI]

My question from before still remains- what earthly motivation is there for MS to suddenly switch camps from a well established x86/nVidia GPU architecture for 2005?

costs !

 

[randycat99]

The problem here is that they wouldn't be likely to gain much acceptance, and thus they wouldn't be producing these better x86 CPUs in anywhere near the volume as Intel. Without a large outlet besides the Xbox2 project in which to sell these superior x86 CPUs, they'll have nothing to offset the risk like Intel can with their multiple market dominances.

I'd imagine they would follow in the tracks of AMD, at the least initially. OK, maybe they wouldn't roll-over Intel right away, but if someone did come up with a chip that managed to considerably outperform Intel, I'm sure they would engender a bit of limelight to make a name in the industry. Initially, they could work the same channels that AMD survives by plus they would have the "Xbox2" account. That should be enough to sustain them (along with a generous MS dowry). If they truly have a "better" chip consistently over multiple generations, I'm sure gradual adoption could threaten Intel over time.

...but alas, I'm arguing the wrong side (to reflect my interests in the topic). I don't think it is possible to come up with a "better" x86 chip out of the great blue. The monetary development risks are considerable, the avenues of initial acceptance can be treacherous (as you had cited), and the performance targets are nontrivial given the competition (IMO). That is a daunting vision for MS to convince anyone to embark on, no matter what kind of money they are offering.

 

[randycat99]

My question from before still remains- what earthly motivation is there for MS to suddenly switch camps from a well established x86/nVidia GPU architecture for 2005?

costs !

Maybe costs, but considering the risks of adopting an entirely new architecture, I don't think cost would be big enough motivation. With the "legions" of money people say MS is willing to pump into the Xbox-n venture, I don't see how saving $5 or so on an encumbent CPU chip would be worth it (vs. just using established Intel pieces and taking that $5 loss on each console as a result). (I'm assuming MS could get a deal price of $20 for an Intel chip, but I have no idea really.) I'm greatly tempted to believe performance would be the greater motivation.