Cooling Revolution.

[MechanizedDeath]

Not necessarily. A fan blowing in may sping at a couple of RPM faster than the one blowing out which would build up dead air in the system and reduce the cooling efficiency. By using ducting and exhaust only fans they are better able to control where the air flows, as well as how much air is moved.

Positive pressure serves a few purposes. For one, it keeps dust out of the case. IMO, dust is a much bigger killer than dead air. If you're pulling in cooler air than what's in the case, then I don't really see a problem in the first place. You will end up with positive pressure, but I don't think the detriment to cooling is significant at all. Some of that pressure is forced out through the exhaust fans anyway. With proper airflow, you'll get the heat removal you want whether you have positive or negative pressure in the case. The problem with exhaust-only systems is the dust comes in and sticks to those heatsinks, reducing the cooling surface area since dust can act as an insulator.

This was beaten to death over the summer. I was one of those claiming the Rev's form-factor was a hindrance to its performance. You can remove heat with a firehose or a straw. Hell, you could remove the same amount of heat with both. But with the same CFMs (all things being equal, the determining factor in cooling), the straw is gonna be so much louder than the firehose. All that turbulent air ripping and roaring over the surfaces inside is gonna sound like a vacuum cleaner.

Rev's chips need to be cooler, period. Smaller process (65nm), low-K/low-clock or a combo of the two. There's just no way that small box can hold as much power as the other two machines without hitting the thermal barrier. Fast chips run hot. It's just a fact of life. PEACE.

 

[Ty]

This is precisely why I kept on asking if the wattage for the 360 increased over the xbox1.

Because if Revolution is going to be in the same ballpark performance-wise, then I'd expect Revolution to be in the same ballpark wattage-wise as well (especially when you consider that both went to IBM/ATI to make their chips - i.e. it's not as if MS said, give us your stupid engineers).

I doubt we're going to see liquid metal coolers in Revolution as some were claiming.

 

[OtakingGX]

I've seen a lot of heat pipe coolers recently, but I wonder about their effectiveness when compared to normal finned heat sinks. The purpose of a heat pipe is to take advantage of the huge heat capacity of a liquid's phase change from liquid to vapor. Vapor naturally rises, where it will be cooled, condensed, and returned to the heat source to draw away more heat.

Integrated heat sink/heat pipes probably do very well as a device is warming up, but I can't see them operating much better under steady state conditions, when the temperature distribution across the heat sink is no longer changing. Take for example the shot of the X360's CPU cooler: Only 75% of the heat pipe is surrounded by fins on the top. Also, if the 360 is sitting upright, then the heat pipe looks like a 'U' and vapor must travel down (which it doesn't want to do) first, then up. There's probably some sort of wick in there, but it would definitely be better to use the natural convective properties of a vapor.

A much better heat pipe design, I believe, is Shuttle's. Now heat is not being conducted through fins and a heat pipe, which uses those same fins for cooling, but instead heat is transfered to a radiator via the heat pipes (like your car does).

Edit: A few posts came in between me starting my post and actually completing it.

One of the interesting parts of Shuttle's design is that you can put the radiator in your intake air stream. Now, instead of trying to cool your chip with preheated air inside the case, you're doing so with the coolest possible air you can. They also set up fans on either side of the radiator in push/pull fashion, so you get very good airflow across the fins.

All things aren't equal, and CFMs are only part of the cooling equation. From what I learned in heat transfer there are many components to moving heat around, you have how much heat you're creating, the k (thermal conductivity) of the material (copper is better than aluminum), h (convective heat transfer coefficient) which is affected by what fluid passes over the cooler and velocity/turbulence, there's also ambient temperature, surface area, and contact resistance.

Some of the easiest variables to manipulate are h, by passing more air over the cooler or inducing turbulence, surface area, by adding fins, and k, by using copper instead of aluminum. Creating less heat is always a plus, and something I'm sure Nintendo has been talking with IBM and ATi about. On ArsTechnica it was speculated that Broadway could put out about 25W. I'd estimate a little more, as Shuttle's cooler can keep a Pentium D, which puts out ~130W, at <60C.

 

[pc999]

It is my impression or a lot of people are assuming that Rev will need the same cooling requirements as a XB360 (will generate the same heat) as it will not be as powerfull (in number at least), so they probably need of much less transsitores in the GPU, memory/speed of the memory, powersuply, CPU speed (decompreesion of data from the disc), and they may even lauch on 65nm...

Would be better thinking on the other perspective, what kind of cooling solutions could Rev suport, IMO.

 

[Edge]

From what I have heard, copper on the chip is best, but aluminum is better for conducting heat to air. So the heat sink on the X360 CPU is done well, with it's copper bottom and aluminum top.

No wonder people's disks are getting hot, having the DVD drive over the GPU like that. Don't understand why they insist on using a full size DVD drive, as they could have used a small drive, allowing the GPU to have a larger heatsink, along with more airflow, and this probably solve the main overheating problem.

I agree the push pull ideal would be smart also, with some fans in the front, but you have to then wonder about realibility with FOUR fans over time.

 

[OtakingGX]

Copper conducts heat better than aluminum. There's no magical effect where aluminum convects heat to air better than copper. Copper is superior in all regards, but aluminum is cheaper: about half the price, so most coolers tend to be made of aluminum. X360's cooler is made the way it is for cost reasons, not some magical performance increase.

 

[PC-Engine]

The x360 mainboard is big because it needs big coolers on both chips, and a lot of support equipment in the form of power regulation, southbridge, video scaler etc. Yes, xenos cooler is also very large, if you look closely it actually has a fair height, it just looks really low profile because it's so damn WIDE! It's pretty much, if not entirely the width of the DVDROM it sits beneath! Anyone who thinks that is a small heatsink is clearly smoking crack.

The GPU heatsink is not big at all. Look at the attached pic on the previous page with the outlines. The thin WHITE square outline around the GPU is the size of the heatsink. It looks to be about half an inch in height too.

This big white thing is part of the cooling system, correct?

http://images.anandtech.com/reviews/system/microsoft/xbox360/no_cov.jpg

Think that's going to fit inside the Revolution there PC Engine?

That big white thing is for the TWO BIG VERTICAL FANS and BIG CPU heatsink which needs to dissipate at least 85Watts. It's not required for the GPU since the GPU is just piggybacking off the CPUs huge cooling apparatus. The GPU only dissipates around 25-30Watts.

I don't know what your're getting here, we have seen the size of Revolution case, isn't expected for it to have low profile heatsink just for it to fit in there ?

My point is Revolution can have the exact same GPU as Xbox360 as the small heatsink for the GPU in Xbox360 proves. As for Revolution's CPU, it will not dissipate 85Watts so it will not need that huge CPU heatsink and all the huge fans and ducting that you see there. For Revolution CPU, it will only need to share a unified CPU/GPU heatsink. Basically take the GPU heatsink and increase its length width to cover both GPU and CPU. If you look at the pic I attached, the yellow square outline is the size of the unified GPU/CPU heatsink I'm proposing. It's about two times as big as the GPU heatsink that's currently in Xbox360.

The lack of heat sink size requires a massive amount of airflow to compensate. My 360 blows air out the back that's almost as hot, and almost as fast as a hair dryer, and pretty much everyone knows about the overheating problems.

First of all the massive dual fans and ducting is mostly for the CPU which is dissipating at least 85Watts! It is not required for the GPU!

Second there is no widespread overheating problems! There are a minor few Xbox 360s that have problems. This is normal for any large console launch. Unless you have actual numbers of Xbox360 units with overheating problems (that were well ventilated) then you have no ground to stand on.

Finally there are millions of 3/4" thick notebook computers running hot 3GHz+ P4s on AC power at full throttle with no overheating problems whatsoever so even if the CPU in Revolution needs to dissipate lots of heat, it isn't going to be impossible to overcome.

We know from Nintendo that the Revolution will be slightly larger than the DVD drive in the picture I posted. By looking at the entire cooling system in the 360 (Heat sink AND FAN ASSEMBLY we can safely conclude that there is no possible way that would fit in the Revolution.

It doesn't need to, but if it did it would use the same apparatus as 3GHz+ notebook computers.

The Revolution must incorporate chips that run significantly cooler than the 360's if they plan on launching at the size that they've stated. PC Engine pretty much just proved that point by showing how massive of a fan system is required to compensate for such a small heat sink.

Actually I proved you wrong again. Read above and also look at the size of the motherboard.

It would be more effective as the powermac G5s employ this strategy. But with the tall heat sink and the DVD drive in the way it's hard to channel the air like that.

It's pretty trivial to channel air using an inlet duct and inlet fan in addition to what's already there on the outlet side.

I just think it was a bad idea to place the DVD drive ontop of the GPU heat sink. It looks so cramped I'm surprised they could get any air current going.

Logic would indicate that, but there are ways around it that MS didn't pursue. One option would be to put a heatshield under the DVD drive. Another would be to use a slim DVD drive. Another is to just use the innards of the DVD drive instead of the whole box enclosure. Another is to enclose the GPU heatsink in a plastic shroud like that found on highend graphics cards.

The problem with exhaust-only systems is the dust comes in and sticks to those heatsinks, reducing the cooling surface area since dust can act as an insulator.

A fan blowing in will also blow in dust. You can remedy the dust problem with a thin piece foam filter, but then you have to regularly clean the filter which I don't think is feasible for a maintenance free console. Dust is always a problem regardless if you use exhaust fans or not.

Rev's chips need to be cooler, period. Smaller process (65nm), low-K/low-clock or a combo of the two. There's just no way that small box can hold as much power as the other two machines without hitting the thermal barrier. Fast chips run hot. It's just a fact of life. PEACE.

As I have pointed out with pictures, the GPU doesn't need to be cooler or less powerful than what you have in Xbox360. The GPU's cooling requirement is miniscule compared to the CPU. The pictures prove this. The argument that Xbox360 is big because of cooling requirements is false according to the pictures. It's using a bigger DVD drive and the motherboard is big. Heck even if Xbox360 was passively cooled it would still be the same size due to the drive and motherboard. The argument that Revolution cannot be as powerful because it would need to be the size of Xbox360 is flawed and the pictures prove this.

The only issue of contention now is the Broadway CPU and what it needs in order to compete with XeCPU. It would likely need to consume less power but that doesn't mean less performance. We know that a single 3.2GHz PPE with an integrated miniPPU block running at only 1.6GHz would compete favorably to a 3 PPE 3.2GHz XeCPU while consuming less power.

Because if Revolution is going to be in the same ballpark performance-wise, then I'd expect Revolution to be in the same ballpark wattage-wise as well (especially when you consider that both went to IBM/ATI to make their chips - i.e. it's not as if MS said, give us your stupid engineers).

Well MS doesn't need to be too concerned with power draw while Nintendo will as they've publicly stated. Also we don't know what Nintendo has in mind and they're not limited to just using PPE cores. There's no stopping them from licensing PPU core IP or coming up with their own generic version of it. A single 3.2GHz PPE is enough for AI while a miniPPU is enough for everything else. AGEIA would love for their IP to be in tens of millions of consoles which would also kickstart their PPU PC addon card business. Let me just remind everyone the PPU's specs.

125 million transistors
182 mm^2 die size
130 nm process technology from TSMC
20 watts power consumption

At 90nm you're looking at around 100mm^2 and 10-15Watts and that's for the fullblown version not a mini version.

 

[Guden Oden]

I've seen a lot of heat pipe coolers recently, but I wonder about their effectiveness when compared to normal finned heat sinks.

It's much greater. In fact, it's so much greater that several of the best heatsinks today don't even have any fins attached to the hot-plate surface interfacing with the processor die.

Vapor naturally rises, where it will be cooled, condensed, and returned to the heat source to draw away more heat.

That is only true for a system where air is undisturbed btw. And for a volume of air large enough to express temperature differences across it. Not neccessarily true of the conditions in a heatpipe.

Integrated heat sink/heat pipes probably do very well as a device is warming up, but I can't see them operating much better under steady state conditions, when the temperature distribution across the heat sink is no longer changing.

But the state is constantly changing. There's cool air blowing across the sink continuously in a normally functioning implementation.

Naturally, the heatpipe won't transfer any more heat if the sink it attaches to heats up to the same temperature as the source (processor die) because there is no cooling of the sink, but neither will a conventional sink function under such conditions either.

If heatpipes stopped working when a temperature plateau has been reached in the cooler, then any CPU cooled by zalman's new "figure 8" heatpipe cooler (don't remember the product name but I'm sure you know which one I mean) would have been destroyed, seeing as only heatpipes attach to the hot-plate.

Also, if the 360 is sitting upright, then the heat pipe looks like a 'U' and vapor must travel down (which it doesn't want to do) first, then up. There's probably some sort of wick in there, but it would definitely be better to use the natural convective properties of a vapor.

There's a partial vacuum in all heatpipes. Vapor is automatically distributed across the entire inner volume due to the pressure differential, it doesn't (particulary) matter if it is pointed up or down, the vapor "wants" to go to the other end anyway. Convection isn't the method which heatpipes function with...

 

[function]

My point is Revolution can have the exact same GPU as Xbox360 as the small heatsink for the GPU in Xbox360 proves.

But it doesn't prove this at all, infact quite the opposite as people have tried to point out.

Look at the area that the GPU heatsink takes up - it's not small - and it's only for gathering heat from the GPU and a couple of other chips. The rest of the chips in there are free to radiate heat into the case space and will have airflow over them as it's drawn towards the ducts.

The GPU heatsink has relatively cool air flowing over it - it's not inline with the CPU heatsink - and those fans will draw a lot of air.

Just becauce it looks like you could fit the GPU heatsink in the Revolution doesn't mean that's realistic given the cooling requirements of the rest of the system, and even if you did (you won't) you can't get rid of as much heat without the same airflow of "unused" air. How are you going to fit those fans in Revolution, and dedicate that much airflow to the GPU?

 

[!eVo!-X Ant UK]

Seems to me like there copying PS2's cooling lay-out, only with a few extra touches.

 

[PC-Engine]

Look at the area that the GPU heatsink takes up - it's not small - and it's only for gathering heat from the GPU and a couple of other chips.

You're telling me to look at the area the heatsink takes up? What you think I'm blind or something? Look at the pic I ATTACHED. Look at the thin WHITE outline around the GPU. THAT is the actual size of the heatsink. That is not big at all. In fact it's pretty damn small for a GPU with that much processing power. As for the height, it's less than half an inch high.

The rest of the chips in there are free to radiate heat into the case space and will have airflow over them as it's drawn towards the ducts.

What does that have to do with anything?

The GPU heatsink has relatively cool air flowing over it - it's not inline with the CPU heatsink - and those fans will draw a lot of air.

Yes and your point is?

Just becauce it looks like you could fit the GPU heatsink in the Revolution doesn't mean that's realistic given the cooling requirements of the rest of the system, and even if you did (you won't) you can't get rid of as much heat without the same airflow of "unused" air.

I didn't say the heatsink will work just because it looks like it will fit. I'm saying the heatsink is small because it doesn't need to be bigger. Besides, the heatsink in Revolution would be even bigger because it would be a single unit used to cool both the GPU and CPU.

How are you going to fit those fans in Revolution, and dedicate that much airflow to the GPU?

Did you even read any of my previous posts? The TWO BIG VERTICAL FANS are there mainly to cool the CPU which dissipates 85Watts of power. It is not there for the GPU. The GPU only needs a fraction of the cooling requirement you see there. I'd say only around 20-25% of the cooling system is required for the GPU. The rest is for the CPU. BTW I don't get what you mean by "unsused air".

 

[ADEX]

Looks like the GPU is getting a full half of the air, the CPU heat sink is tall and relatively thin, the GPU's short and fat. It looks like the the GPU is putting out lower heat than the CPU but not massively lower.

 

[PC-Engine]

Looks like the GPU is getting a full half of the air, the CPU heat sink is tall and relatively thin, the GPU's short and fat. It looks like the the GPU is putting out lower heat than the CPU but not massively lower.

Even if the GPU gets half the air which is debatable, the heatsink has a lot less surface area not to mention no heat pipe compared to the CPU. This points to the GPU needing less cooling than the CPU. ATI has said the GPU only dissipates around 25-30Watts about 1/3 the CPU which is about 25% of the total power consumption of the GPU+CPU which matches the 20-25% cooling system requirment for the GPU.

 

[Teasy]

Some very interesting stuff here. Its suprising just how much space 360's DVD drive takes up (almost half the system). Revolutions drive should be about a quarter of that size.

Also I agree with PC-Engine's idea of extending the GPU heatsink to cover both GPU and CPU. I'm guessing Hollywood will use about 10-15 watts and the CPU will use around 40-45 watts. 360's GPU uses about 25 watts, so if you extend that heatsink to twice its surface area it should easily be able to handle those two chips.

The PPU mentioned is also interesting. I'd always gotten the impression (from people here) that AGEIA's PPU was very power hungry. But 20 watts is quite a bit less then a single 360/PS3 PPE core and that's at 130nm (more like 14 watts on 90nm). PC-Engine, how powerful (calculation wise) is that PPU?

 

[function]

Cooling requirement doesn't relate directly to airflow. You can cool a hotter chip with less or warmer air with a better heatsink.

The 360's GPU has the largest GPU heatsink seen in a console. Here's the original Xbox:

http://www.myxbox.net/pictures/console/xbox_bloc_proceseurs.jpg

and the GC:

http://www.anandtech.com/showimage....eviews/system/gameconsoles/gamecube/step2.jpg

Quite how the 360's GPU heatsink classes as small I don't know. Perhaps if you pointlessly compare it to something on fast PC card.

Also I agree with PC-Engine's idea of extending the GPU heatsink to cover both GPU and CPU.

This is what the GC did. CPUs typically consume more power than GPUs, so if you share a heatsink I'd say odds on the CPU will use the majority of its heat dissipation capacity.

The idea that you could practially put the 360's GPU in the Revolution is not a great one IMO. Little fans make annoying noises, and the faster they spin the more annoying they are.

 

[Teasy]

I think its possible, but certainly not neccesary. I can see Revolution using a GPU as advanced as 360's, just with less of everything. Half the pipes, around half the eDram (maybe 6MB) ect and consuming half the power (12-13 watts). For the CPU perhaps they could use a single PPE core at 3.2Ghz (28 watts) with a cut down AGEIA PPU on the same die (around 10 watts). That system would have a good amount more power per pixel then 360 at standard resolutions and easily enough AI/physics power to compete. Yet its GPU and CPU would only consume a combined 50 watts compared to the 110 watts 360's CPU/GPU uses.

 

[Joe DeFuria]

I'm not sure I understand the Nintendo fan's approach here...usually it's "power of the console is not really important" followed by "and this is how Nintendo will be able to compete with power."

None of these console companies are appreciably better when it comes to getting more computing power at a given cost or power consumption level. The bottom line is, if Revolution is going to be considerably smaller in size (which hints at significantly lower power consumption), and supposedly significantly less costly to manufacature then the console itself will be significantly less "powerful." (Note that I'm not passing any judgement on the console itself.)

Any speculation beyond that is pure guesswork and wishful thinking. Throwing around power consumption numbers for an on-die Agiea PPU? Come on....what's the point of pulling numbers like these out of your ass.

 

[pc999]

Well MS doesn't need to be too concerned with power draw while Nintendo will as they've publicly stated. Also we don't know what Nintendo has in mind and they're not limited to just using PPE cores. There's no stopping them from licensing PPU core IP or coming up with their own generic version of it. A single 3.2GHz PPE is enough for AI while a miniPPU is enough for everything else. AGEIA would love for their IP to be in tens of millions of consoles which would also kickstart their PPU PC addon card business. Let me just remind everyone the PPU's specs.

125 million transistors
182 mm^2 die size
130 nm process technology from TSMC
20 watts power consumption

At 90nm you're looking at around 100mm^2 and 10-15Watts and that's for the fullblown version not a mini version.

How do you know that 1 PPE+cut down PPU is equal to XeCPU, that is wrong unless there is someone using 4 threads for physics and animation, and we dont even know how much can help with animation (besides physics basead animation), still 1 PPE still hot, plus I dont think that there is a lot of chances (but still some) of we see a AGEIA PPU in Rev unless they GIVE the Novodex to Nintendo for all the games (1,2,3 partys) (it may be a good trade off if there is enought ports from/to PC which is possible considering that it can work as a mouse too, but I still have a few doubts) I think it would be more probable a maths processor from IBM (basic as a bounch of VMX units) or ATI (a bounch of pipelines) as they are used to do inexpensive and coll HW, anyways I dont think a PPE would be ideal (hot, expensive) as it is in order and they can make a lot of money from companys how cant afford make a game for XB/PS but still have great ideas (as it happens for DS).

 

[Powderkeg]

You're telling me to look at the area the heatsink takes up? What you think I'm blind or something? Look at the pic I ATTACHED. Look at the thin WHITE outline around the GPU. THAT is the actual size of the heatsink. That is not big at all. In fact it's pretty damn small for a GPU with that much processing power. As for the height, it's less than half an inch high.

Yes, apparently you are blind, or more likely something.

Here we can see the heatsink for the GPU extends from the yellow transistors on top to the yellow transistors on the bottom. It also stretches from the CPU heatsink to almost the black and silver chip on the upper-left.

That means that in this pic:

http://images.anandtech.com/reviews/system/microsoft/xbox360/mb.jpg

the heatsink takes up the entire outlined space around the outer edges of the orange holes.



So yes, you are blind,,,,,,,,,,,,,,,,,,,,,,,,,,,or something.




Now that we've clarified that you are in fact blind, or something, would you like to reevaluate your position based on the newfound knowledge that your previous claims are based on such obviously bullshit claims?





What does that have to do with anything?



Yes and your point is?



I didn't say the heatsink will work just because it looks like it will fit. I'm saying the heatsink is small because it doesn't need to be bigger. Besides, the heatsink in Revolution would be even bigger because it would be a single unit used to cool both the GPU and CPU.



Did you even read any of my previous posts? The TWO BIG VERTICAL FANS are there mainly to cool the CPU which dissipates 85Watts of power. It is not there for the GPU. The GPU only needs a fraction of the cooling requirement you see there. I'd say only around 20-25% of the cooling system is required for the GPU. The rest is for the CPU. BTW I don't get what you mean by "unsused air".[/QUOTE]

 

[Teasy]

Joe I'm speculating on possible console uses for current technology in a thread in the console technology forum.. what exactly is your problem with that? At no point did I say Nintendo would use a PPU or anything else, I said "perhaps they could use" not they will use. Also the numbers I used are all based on real power consumption numbers. Oh and by the way not all Nintendo fans think that graphics don't matter.