PS3 AC/DC Adapter?

Would you like to bet which final console draws more power?

What are we betting?

 

"more power", ok, but "FAR more power" is a bit much...

Why would PS3 consume FAR more power than X360, everything considered? I guess NVIDIA doesn't help, but in the end they'll probably be close enough, although i do expect PS3 to draw more power than X360.

 

How about you present your reasoning behind your statements instead. This is beyond3d, try to live up to the level expected of the site's participants...

 

Cell was designed with one of the main sub goals of being power efficient, so the GPU is probably going to be the greatest heat dissipater in the system.
Seeing as the current G70 is running pretty cool at 110nm, I can't see RSX at 90nm being that much of a problem.

 

power efficient when the entire CPU was not running full out, not unlike note book CPU's. However I think Games will make use of the entire CPU.

 

True but even with that in mind, the SPE's running 'flat-out' will draw less power than two extra PPE-class cores would.

@Bill: Bill, I think you're equating Flops generation with power draw as far as the CPU's go - it's not that simple.

Cool find about the power supply for PS3 in general though.

 

Quite possibly it's more about a general power delivery circuit for CELL than one specifically for PS3. e.g. when you look at the Schmoo plot for the SPEs at clocks like 4+ GHz, with the voltage going up and all, you could pretty easily hit 100 W up there. At 5 GHz, the SPEs are supposed to draw around 11 W. Down at 3.2 GHz, it reads 3 W. Though the chart figures are all rounded (or possibly truncated) which means "3 W" can possibly mean 3.49999 or even 3.99999.

In fact, I think it's possible for CELL to hit 100 W even at 3.2 GHz if the core voltage is pulled up to 1.3 V or so. The design itself can supposedly bin at 3.2 at 0.9V, but obviously, lower quality dies will happen, even if they don't make it into PS3.

Cooling is one factor. There's also the fact that the 360 is full of higher-clock lower-voltage parts than its predecessor, so it is more sensitive to variances and noise. It wouldn't surprise me if there's a load of big capacitors/inductors inside that brick. And it's certainly possible that there simply is that much power draw.

PS3 may draw more power than 360, but it won't be because of CELL. RSX is probably the more likely culprit.

 

And perhaps bluray (as i'm pretty sure it uses a more powerfull laser ) and xdr ram i believe runs at a higher voltage than gddr ram .



Anyway with both systems as process shrinks happen and yields increase power draw should drop

 

I believe they both operate at 1.8Vdd. XDR signaling, however, uses a much smaller voltage swing (0.2V) than GDDR3 (1.0V), and it's the wiggling of voltage along wire traces on the board that really uses up power. That low swing is why XDR can clock up to high signaling rates, so it's going to wiggle the voltage more times, but I would still expect lower power overall.

 

Well technically the PS3 is smaller than the X360.

 

The Cell Processor in the PS3 will consume more power than XENON in the XBox360 as the Cell Processor has more transistors and a larger die size. If XENON consumes roughly 85watts (though I believe this number may be closer to 100 watts) at 3.2GHz using 165 million transistors (which is less than a Pentium 4 CPU consumes) then the Cell Processor should consume over 120-130 watts at 3.2GHz as it uses 235 million transistors. Of course the question would be what voltage is XENON running at and what voltage will the PS3 Cell be using. The bigger problem may likely be the graphics processor in the PS3... in the XBox360 ATI stated that the power consumption of XENOS is roughly 65 watts. A Geforce 6800 Ultra consumes about 100 watts at 400Mhz using 130nm process and the Geforce 7800GTX at 430Mhz and 110nm process consumes about the same (slightly more under load) and at 550Mhz it is stated that the 512MB Geforce 7800 GTX consumes over 120watts. Now providing the RSX in the PS3 is uncrippled with the same number of transistors/pipelines clocked at 550Mhz, but using 90nm process that power consumption should be reduced by roughly 30% which sould reduce that 120watts from the NV50 at 550Mhz to around 85 watts.

Still more than XENOS, and the Cell CPU will certainly consume more power than XENON given clock speeds... I would say the power consumption and heat issues in the PS3 will likely be worse on the PS3 than it already is on the XBox360 (and I can tell you that the XBox360 does put out quite a bit of heat).

In general... the more voltage you use, the higher clock speed you use, and the more transistors you use the more power that will be consumed and the more heat that will be generated. There are ways to reduce the power consumption and heat output, but that usually involves smaller process shrinks... and that next step would be the 65nm process a year or two from now.

Notes: For those wondering what other power requirements of the Geforce 7800GTX at 550Mhz is... it wants at least a 350watt PSU with 22amps on the +12v rail, but with the Cell CPU which consumes more power than a typical PC CPU... that may need to be more like a 400-450 watt PSU. Also for reference... a Pentium 4 640 is also clocked at 3.2GHz, 167 million transistors, and also 90nm process... and it consumes about 100-120 watts of power. The external power supply for the XBox360 is rated at 203 watts of power and 12.5amps on the 12v.

 

They better get to work on those vents!

 

What the.

 

GameMaster I just have to say that this logic is all wrong. The majority of the transistors in the SPE's are devoted to the cool running local storage, and the schmoo indicates 3w per SPE at 1v. Why in the world are you ignoring that and just going off on this tangent? Your derived logic using transistor counts as a scaling point is just totally off.

The SPE's eat 3 watts each - the cores in the XeCPU ~30 watts each. What is there to analyze here? Not to mention, the XeCPU has a much 'denser' heat profile than the Cell does - the SPE's act as sort of 'cool zone' buffers against one another, by virtue of the way they are arranged and the local storage in between them.

 

Don't get all wedded to the schmoo plots from ISSCC. There is a long and storied tradition of "prototype" parts at ISSCC being hot bin/lots. Actual production yielding of the parts, etc will likely result in higher power requirements.

I've previously given a pretty good estimate of the floor power range for the PS3. If you have an issue with it, we can discuss it.


Aaron Spink
speaking for myself inc.

 

There's a key flaw in that logic. Transistors don't eat up power simply by existing in the chip. The eat up power when switching states. How much power a device uses is dependent on how many transistors are in use on average at any given moment. By your logic, the RAM in your machine should eat up more than any processor on earth as you've probably got a couple billion transistors worth of memory.

The PPE on CELL will probably have about the same power density profile as each of the cores on Xenon, but the same will definitely not be true of the SPEs as they're far simpler.

Nonetheless, there's little news about any sorts of changes going on at the SPE level. There was the DD1->DD2 shift for the PPE, which may affect the necessary core voltage to clock up to 3.2 GHz, but I still think that with the fuss that Sony themselves are making over power consumption, the binning may not be that conservative for PS3, and the lesser parts would be redirected to some other purpose. Only thing that may change that would be crappily dismal yields, and there's quite a few months left to ramp those up. If we're to believe the inquirer (I know, I know...)... they're about halfway there as far as getting decent yields on PS3-grade parts.

 

Well, what you did in reality is you tossed out some arbitrary numbers based on...well...nothing really, and now you call that a 'pretty good estimate'? Man, I wouldn't want you to do the calculations for a future space mission or something like that, considering how you perform your pretty good estimates!

 

Is that a no?

Talk is cheap, but when I see people (not you) throwing around ridiculous numbers like HALF the 360 power consumption on PS3..you just wonder what these people are thinking.

It's easy to throw claims around here in a debate that will be forgotten soon..but if we put real consequences on it..you'll think about it a bit more, maybe.

We can bet a signature of some sort.

It's not rocket science. I went over my reasoning about ten times.

The GPU on X360 supposedly draws 35 w. Everything else is similar, if not a disadvantage for PS3.

So, you think the PS3 GPU is going to come in UNDER 35W? There's no way.

And lets not forget most people think the PS3 GPU is based on the Nvidia G70. which draws around 110W. Doesn't mean RSX will, but it's a starting point.

Not to mention everything I read on Cell in Japanese seems to hint it will require it's own circuit breaker in your house more or less for power itself. Ok, overzealous statement, but we can go the other way with it.

"as for the power circuit of the Cell processor, supplying the electric current of maximum 100A with 1V is required. Because it corresponds to such difficult request, it seems that adopts FPA technology "

Why is powering Cell a "difficult request"?

 

You know your statements are very annoying without any supporting information.

Let's assume PS3 is slighty smaller, so what? Does it make you feel better?

 

I agree with others here, you're pulling numbers out of thin air, and based on it nothing but your own guesses.

What do you know of Sony's 90 nm process?

Do you know of any power saving techniques that Sony is using for CELL?

Do you know the voltage that Sony has decided on?

I'm not making any guesses, as there are too many unknowns at this point.