oh so now Xenon has a TERAFLOP processor? (just CPU alone??)

[Megadrive1988]

ever since J Allard announced that Xenon would have more than a teraflop of computing performance, almost everyone assumed that this was from the CPU and GPU combined, the entire cosnole as a whole, and that most of this floating point performance came from the GPUs processing resources. fudged or not fudged.



but now look at this from the Chris Satchell interview about Xenon using DVD9 has the media:

You have to have absolute killer hardware, and you heard in the keynote about the teraflop processor, so we've got the hardware.

http://biz.gamedaily.com/features.asp?article_id=9165

big difference from previous statements, if true.... a teraflop processor

he very well might have mispoken, or did not mean what it sounded like, but perhaps not ???

that sounds like Xenon has a teraflop processor. (even if thats MSflops) and 'processor' usually refers to the CPU. *not* the GPU, and *not* both the GPU and CPU combined, the system as a whole.


hmmm the only way I guess that Xenon's CPU system could hit a teraflop is if they're using more than one Xenon CPU-module, and plenty of cores.

http://news.teamxbox.com/xbox/7421/Xbox-2-Patent/

Allard talked about the “unbelievable amount of raw computing powerâ€￾ that will be available in the Xbox successor

The thing we're looking at in the next generation is just an unbelievable amount of raw computing power - the architecture will be much more specialised,' he added.

http://www.computerandvideogames.com/r/?page=http://www.computerandvideogames.com/news/news_story.php(que)id=105001



Allard text from Keynote:

http://www.major-nelson.com/blogcast/gdc.txt

This system is a monster; it's going to deliver over a teraflop of targeted computing performance. And for the last three years we've had a thousand engineers working on it, working across nine different locations to bring the system to life.

It's going to be an amazing system. We're going to deliver custom hand-coded silicon, just like the other guys, but we're going to design that for the HD era in mind, the HD consumer in mind and most importantly the HD game creators in mind.

We believe that innovation is critical and to unlock that innovation we need to deliver on familiarity so we designed the system to bridge a gap, bridge from the successful last generation and make sure you can use the same tools, the same processes, the same middleware that you're all familiar with so you're not re-imagining your process, your re-imagining the game.

Now, for the next console we've made a very conscious choice, we made the conscious choice to go to multi-core general purpose silicon, high performance cycles that provide unlimited flexibility, limitless. It's provides you the headroom that you'll need and it gives us the opportunity to deliver value to the customer after we've locked on the hardware specs.

And we're also ensuring that you can take advantage of it by taking a symmetrical multi-core architecture that's happening in the PC world as well, you get to take advantage of all the new techniques and processes that are coming to life both on the Windows platform and on the Xbox platform.

the in-bold parts are what I feel directly and specifically relates to the CPU


btw, does symmetrical multi-core architecture mean an even number of cores like 2,4,8 and not odd ones like 3 cores ?



finally we have this piece from an old GI.biz article that you've probably seen numerous times:
http://www.gamesindustry.biz/content_page.php?section_name=dev&aid=2897

Our sources indicate that the console will use "four or more" of the IBM PowerPC processors, an architecture which will force game developers to significantly rethink the way that games are programmed in order to take full advantage of it.

there! I have made the best case I can come up with in a short amount of time & space, for Xenon to have a teraflop or teraflop+ CPU system.
not counting the GPU. maybe Microsoft meant what it said.


with that said, I don't buy it. I don't believe it. I think Xenon CPU will have 2 or 3, maybe 4 cores, all one one chip. no additional CPU modules with additional cores. no teraflop CPU. more like the 90-110 Gflops that has been most often mentioned here and elsewhere. or still sub-150 gflops in the case of 4 cores.

 

[London Geezer]

Well i'd love to see them proving it. :|

 

[Shifty Geezer]

So they're sticking a few thousand dollars worth of CPU in there then?

 

[Shinjisan]

Have you ever heard of NVflops?
Well it's the same.
They're giving the figure for ALL the system considering hardwired functions performed in the GPU with flops.
It's all PR,they have heard the Teraflops talk from Sony's side and they're finding a way not to look too inferior on the PR specs.
I hope they will release the true,detailed and complete specs also for the people that want to know and they don't try to bypass the true specs with PRish statements on the hardware.

 

[Megadrive1988]

Have you ever heard of NVflops?
Well it's the same.
They're giving the figure for ALL the system considering hardwired functions performed in the GPU with flops.
It's all PR,they have heard the Teraflops talk from Sony's side and they're finding a way not to look too inferior on the PR specs.
I hope they will release the true,detailed and complete specs also for the people that want to know and they don't try to bypass the true specs with PRish statements on the hardware.

originally that was what we all thought -- that they were giving the FP for the WHOLE ENTIRE system - CPU and GPU.

but this new statement from Chris Satchell (and by going over Allards statements again) seem to indicate that Microsoft is, at least TRYING to cop a teraflop from just the CPU alone.

 

[London Geezer]

They're full of crap.
Not even Cell can get to 1TFlop on a single chip without being impossibly expensive, and that's an architecture that does away with some things to ensure it can churn out as many FLOPs as possible. Not sure how 2 or 3 PPC cores in Xbox2 will ever get to 1TFLOP.

 

[Megadrive1988]

They're full of crap.
Not even Cell can get to 1TFlop on a single chip without being impossibly expensive, and that's an architecture that does away with some things to ensure it can churn out as many FLOPs as possible. Not sure how 2 or 3 PPC cores in Xbox2 will ever get to 1TFLOP.

of course not. you're right. 2 or 3, or even 4 PPC cores could not hit a teraflop. but what if they're using several CPU modules, each with several PPC cores?? then maybe they'll reach a teraflop, or what they will stretch into a "teraflop"

but don't misunderatand, I am just trying to make a case for teraflop Xenon CPU, even though I don't actually believe it.

 

[London Geezer]

They're full of crap.
Not even Cell can get to 1TFlop on a single chip without being impossibly expensive, and that's an architecture that does away with some things to ensure it can churn out as many FLOPs as possible. Not sure how 2 or 3 PPC cores in Xbox2 will ever get to 1TFLOP.

of course not. you're right. 2 or 3, or even 4 PPC cores could not hit a teraflop. but what if they're using several CPU modules, each with several PPC cores?? then maybe they'll reach a teraflop, or what they will stretch into a "teraflop"

but don't misunderatand, I am just trying to make a case for teraflop Xenon CPU, even though I don't actually believe it.

If there were a CPU available today or in the next 12 months that could pull out 1TFLOP, we'd know about it, like we've been hearing about how Cell will do that, albeit under a very expensive configuration (4 cores or so?). Even too expensive for PS3, which will come out long after the Xbox2.

So, the Xbox2 being released in the next 12 months, there is simply no way it will have a CPU capable of 1TFLOP.

This PR bullcrap is really not healthy for this market.

 

[Inane_Dork]

Has it ever occured to anyone that it's just a slip of the tongue?

Or is that too simple and logical an explanation for you?

 

[Pugger]

Lets be honest, XB2, PS3 and Revolution are game machines that are going to retail for £200-300. When new machines come out the PR agencies go into over drive and gamers tend to do the same. All the machines will be comaprable none will significanlty outperform the other when it comes to actual game contnent. We all go overboard and its great to speculate.

 

[aaaaa00]

Re: oh so now Xenon has a TERAFLOP processor? (just CPU alon

btw, does symmetrical multi-core architecture mean an even number of cores like 2,4,8 and not odd ones like 3 cores ?

Symmetric, as in "Symmetric Multi-Processor" (SMP) simply means that each CPU in the system is identical, and a program written for that system can run on any of the CPUs inside it.

Since it is fashionable these days to refer to multiple CPUs on one die as a "multi-core" architecture, you can consider "Symmetric Multi-Core" as meaning the same thing as "Symmetric Multi-Processor".

Contrast with "Asymmetric Multi-Processor", where each processor in the system is not identical to all the others, and a program written for one processor may not necessarily run on a different processor in the system. Ex: CELL.

 

[Megadrive1988]

Re: oh so now Xenon has a TERAFLOP processor? (just CPU alon

btw, does symmetrical multi-core architecture mean an even number of cores like 2,4,8 and not odd ones like 3 cores ?

Symmetric, as in Symmetric Multi-Processor (SMP) simply means that each CPU in a multiprocessor system is identical, and a program written for that system can run on any of the CPUs.

Since it is fashionable these days to refer to having multiple CPUs on one die as "multi-core", you can consider symmetric multi-core as meaning the same thing.

thanks for the clarification on that 8)

 

[one]

You know, they can speak about whole teraflop-petaflop crap some time and just before the release will lower it near the real figures as in the Xbox 1 PR.

 

[London Geezer]

In the end, the big numbers never saved them this generation, they're surely not gonna save them next time around...

People will buy the consoles with the good/cool/big games. If they happen to be on Xbox2, they'll buy it, TFLOP or not.

 

[Brimstone]

Maybe some of Microsofts bleeding edge research is finding its way into what we are calling the CPU. Mihai Budiu's, a Microsoft Silicon Valley researcher has produced some intresting stuff.

My research explores a new model of computation: ASH, Application-Specific Hardware. In this model of computation one no longer uses general-purpose processors (i.e. microprocessors) to execute programs. Instead, high-level language programs are directly compiled by CASH (Compiler for ASH) into asynchronous circuits. My work is focused on developing CASH and cycle-accurate simulations for ASH systems.

CASH is described in FPL '02 and CGO '03. My defense [Powerpoint] [pdf] contains some interesting results. See also our ASYNC 2004 tutorial.

My thesis describes in detail the compilation process and compares ASH with other computational substrates. I show that ASH can effectively exploit the instruction-level parallelism in media kernels. On control-intensive kernels ASH is somewhat less effective than a superscalar processor, since it lacks branch prediction and general speculation support, and incurs additional synchronization overheads. ASH is however up to three orders of magnitude more power-efficient than superscalar processors, one to two orders of magnitude better than low-power DSP processors, one order of magnitude better than asynchronous processors, and approaches custom hand-designed hardware. These results are summarized on our ASPLOS '04 paper.

http://www-2.cs.cmu.edu/~mihaib/research/research.html#research


http://www-2.cs.cmu.edu/~phoenix/compiler.html[/url]

 

[McFly]

Do you remember the big hypercomputing story some years ago: http://www.starbridgesystems.com

That ASH sounds a bit similar.

Fredi

 

[Titanio]

He's a software guy more than a hardware guy. Like Allard, the MS marketing department is coaxing him into dropping that "teraflop" number at convenient opportunities, and the result is undoubtedly a slip of the tongue.

 

[Deadly Towers]

I don't see how that comment is interesting considering that even a geforce 6800 is a "teraflop processor".

 

[Shifty Geezer]

of course not. you're right. 2 or 3, or even 4 PPC cores could not hit a teraflop. but what if they're using several CPU modules, each with several PPC cores?? then maybe they'll reach a teraflop, or what they will stretch into a "teraflop"

but don't misunderatand, I am just trying to make a case for teraflop Xenon CPU, even though I don't actually believe it.

It'll never stretch. What does a PPC core manage? Let's say 20 GFlops. They'd need 50 of them!!! OR put it another way, if they want 1 Teraflop from 4 cores, that's a 256 GFlops core. Only chip to do that is Cell which is why we've been hearing about it so much. If IBM had a PPC core capable of Cell's performance, we'd know about it.

No matter which way anyone tries to spin it, 1 teraflop from a CPU isn't going to happen this generation.

 

[Megadrive1988]

I don't see how that comment is interesting considering that even a geforce 6800 is a "teraflop processor".

since when? i dont recall GeForce 6800s getting rated at "a teraflop" even in PR.