XENON to have two PPC Cores?

[j^aws]

...
In its next version of the Xbox, Microsoft plans to shift from using Pentium chips from Intel to a PowerPC microprocessor from I.B.M. The chip will have two PowerPC processor cores, but it will not be as radically new as the I.B.M. Cell design that Sony plans to use, said one executive who is familiar with the Microsoft project.
...

http://www.nytimes.com/2005/02/07/technology/07chip.html

 

[cthellis42]

NY Times. Where to go for your up-to-date tech information!

 

[Tuttle]

Sounds like he is talking about the 970 MP, Antares. It is rumored to arrive in Apple Power Macs sometime this year.

 

[one]

Richard Doherty is everywhere. Why don't they hear from others... :?

 

[j^aws]

NY Times. Where to go for your up-to-date tech information!

Hey, I know it's not Spong but nonetheless, it's a decent read!

Anyway this would smell of a 90nm process for the Xe CPU if it was a dual core...

 

[cthellis42]

Hey, I'm not saying I didn't read it, I'm just saying that pulling a loose comment about the number of cores involved probably doesn't say much in general. The mainstream press frequently gets all sorts of information like that wrong.

 

[Titanio]

Hey, I'm not saying I didn't read it, I'm just saying that pulling a loose comment about the number of cores involved probably doesn't say much in general. The mainstream press frequently gets all sorts of information like that wrong.

They go on to mention "an executive close to" Xbox2, so I'd say it probably came up when talking to him/her about it. That "executive" was probably at one of MS's partners (like say, IBM, perhaps).

Of course, nothing's for sure, but certain things seem more likely now, at least to me.

 

[j^aws]

Hey, I'm not saying I didn't read it, I'm just saying that pulling a loose comment about the number of cores involved probably doesn't say much in general. The mainstream press frequently gets all sorts of information like that wrong.

It's quite interesting that the majority of the above article is about CELL but I intentionally did not post it's content just to focus on Xenon, but the SAME article posted later,

http://www.beyond3d.com/forum/viewtopic.php?t=20287

...but focusing on CELL instead, generates something seemingly more credible? Don't get me wrong, I know what you meant, It's just a personal observation and prolly means nothing!

 

[j^aws]

*BUMP*

So is this same story from the NYTIMES validated elsewhere?

 

[Pugger]

So the leaked specs showing 3 PPC cores is a fake? Or has MS revised its specs downwards just as CELL gets a shed load of PR coverage?

 

[j^aws]

So the leaked specs showing 3 PPC cores is a fake? Or has MS revised its specs downwards just as CELL gets a shed load of PR coverage?

Well to some, a 1PE CELL is a downgrade from a 4PE CELL (i.e. a BE)...

Also remember that a dual core/ smaller die would clock higher than a larger tri-core die. So we may still see 4+ GHz speeds if the tech is comparable to CELL at 90nm...

 

[Pugger]

But if they were going to go with a smaller die size for the dual PPC surely the the CELL would go to 65nm. As for the 1PE 8APU none of this has been confirmed so we will have to wait for Sony, although I do agree it does at this point seem very likely.

 

[wco81]

I thought the Xenon CPU was suppose to be 3.5 Ghz?

And the 970FX was already at 90nm?

 

[j^aws]

But if they were going to go with a smaller die size for the dual PPC surely the the CELL would go to 65nm. As for the 1PE 8APU none of this has been confirmed so we will have to wait for Sony, although I do agree it does at this point seem very likely.

There are rumours that a 'new' high clocking POWER based core is the basis for CELLs PU and Xenons PPC core. We need some heat dissipation figures to come out of these presentations...but they would have to balance clock/watts/die size etc...

If you take a PE at 4GHz at 90nm, then shrink it to 4PEs at 65nm, it doesn't mean you can still clock it at 4GHz...there are thermal issues that may force you to down clock. So a 1PE at 4GHz may ONLY clock at 1GHz with 4PEs and ultimately give the same 256GFlops figure...

 

[j^aws]

I thought the Xenon CPU was suppose to be 3.5 Ghz?

And the 970FX was already at 90nm?

The leak said 3.5+Ghz...see above for the core description...

 

[V3]

GwiDan wrote:
I don't think it's unrealistic to expect the Cell to be based on the G5.

Steve wrote:
They are about as architecturally similar as a Ferrari 360 Spider and a bog brush.

The Xbox2's CPU's are related to the G5's though.

http://forum.gamesradar.com/viewtopic.php?t=58962&postdays=0&postorder=asc&start=30

I went there and found that from the guy in the know, interesting.

 

[Guden Oden]

Also remember that a dual core/ smaller die would clock higher than a larger tri-core die.

Um pardon me, but that sounds like complete fiction IMO. I see no absolute reason why a n-core die would clock less than a n-1 core die...

 

[AzBat]

So the leaked specs showing 3 PPC cores is a fake? Or has MS revised its specs downwards just as CELL gets a shed load of PR coverage?

The L2 cache sharing was evidently not fake, but I always had my doubts about the rest of the CPU info. It's possible it's fake or it's possible it's true or even possible that it was early specs that were abandon awhile ago. There's no telling, but I'm sure we'll find out what the truth is once MS spills the beans.

Anyway, dual PPC cores seems more rooted in reality. Whether or not they're dual issue is another matter.

I personally don't see the CPU being an underperformer even it doens't match the leaked specs. I will remind you of some of Deano's more recent comments regarding the Xenon CPU:

What I know of its origins is that its not an existing processor design you can buy or read any public information on, that MS asked and got some specific hard to make features (that IBM had problems with) and its quite powerful

Each core isn't related to an existing core though. If it has only one core it would still be unique from any existing PowerPC chip.

Hopefully all will be laid bare in the coming months if not weeks.

Tommy McClain

 

[j^aws]

Also remember that a dual core/ smaller die would clock higher than a larger tri-core die.

Um pardon me, but that sounds like complete fiction IMO. I see no absolute reason why a n-core die would clock less than a n-1 core die...

It's a simple rule of thumb...obviously there are other variables involved but the heat generated by MORE switching gates for MORE cores clocked at the SAME speed is gonna generate MORE heat!

 

[Phil]

^^ that's only the case if the power voltage is equal on 65nm (which it shouldn't be) though