Predict: The Next Generation Console Tech

[anexanhume]

No. In a water cooling rig, the heat gets captured in the water by heating it, and released in the cooler by cooling it. In a vapour chamber or a heatpipe, most of the heat is stored as the energy of evaporation, as opposed to just heating the water. This is ~an order of magnitude more efficient than plain water cooling.

Yes, it relies on a change of state, but they are both based on the superior coefficient of heat of water, and evaporation still serves as a proxy for a pump.

 

[MrFox]

Yes, it relies on a change of state, but they are both based on the superior coefficient of heat of water, and evaporation still serves as a proxy for a pump.

I think you're mixing up conduction and phase change. If the phenomenon was simply conduction, we might as well use a solid slab of copper (which is much better than water in this specific application where the distance is about one millimeter).

Here's the swimming pool example: Water and air have the same temperature. When you go into the water, your skin temperature becomes the same as the water around, and your comfort depends on the water's temperature. That's conduction. When you get out of the water, your skin gets MUCH colder than the air around. You'll feel very cold regardless of the initial water's temperature. That's evaporation. If it's windy or dry, the evaporation is faster, and you get much colder. Consider the tiny amount of water that was on your skin and you get the idea how powerful the phenomenon is compared to conduction.

 

[tunafish]

Yes, it relies on a change of state, but they are both based on the superior coefficient of heat of water, and evaporation still serves as a proxy for a pump.

No, the coefficient of heat of water is almost entirely irrelevant. The heat of vaporization of one gram of water is 540 times higher than the energy needed to heat one gram of water by 1 degree celcius.

Let's say you have a heatpipe where one end is 10 degrees C hotter than the other end. That's actually quite a lot of difference for the class of heatpipes used in computer parts. Even in that case, heat of vaporization is responsible for 98% of the total energy transfer. The coefficient of heat is irrelevant.

I think you're mixing up conduction and phase change. If the phenomenon was simply conduction, we might as well use a solid slab of copper (which is much better than water in this specific application where the distance is about one millimeter).

You're even more confused than anexanhume. No-one was claiming that the work is done by convection alone. In a normal water cooling system, the heat is transferred into the water at the block, by convection, but that hot water is then moved to the radiator with the pump. This is inherently more efficient than convection through a solid slab of copper. Heat pipes use phase change instead, making them even more efficient.

 

[Shifty Geezer]

You're even more confused than anexanhume. In a normal water cooling system, the heat is transferred into the water at the block, by convection, but that hot water is then moved to the radiator with the pump. This is inherently more efficient than convection through a solid slab of copper. Heat pipes use phase change instead, making them even more efficient.

I think everyone's a bit confused.

 

[anexanhume]

No, the coefficient of heat of water is almost entirely irrelevant. The heat of vaporization of one gram of water is 540 times higher than the energy needed to heat one gram of water by 1 degree celcius.

Let's say you have a heatpipe where one end is 10 degrees C hotter than the other end. That's actually quite a lot of difference for the class of heatpipes used in computer parts. Even in that case, heat of vaporization is responsible for 98% of the total energy transfer. The coefficient of heat is irrelevant.

Coefficient of heat is still responsible for how quickly phase state change can occur. If your liquid is resistant to heating, heat is forced to flow in the metal shell of the vapor chamber. Evaporation is still the mechanism to move the fluid to a place where it can be cooled. The analogy is still valid, even if phase state change is where a bulk of the heat energy is transferred.

I think everyone's a bit confused.

Convection is the correct term. Conduction of heat through diffusion and fluid flow.


Edit: I should clarify further. My original post comparing the coefficient of heat for water and air is an argument for liquid based cooling, not a specific implementation of it. In the end, you must face that heat will be removed by forced airflow over metal. Forcing liquids or other gases with a superior coefficients of heat in a non-enclosed system is not technically feasible.

So, the question is what is a superior way to get heat to the metal that is ultimately being air cooled? If you're going through a solid medium, you can't beat a nice metal like copper. Thus, you must artificially boost the amount of molecules available to carry heat. Introducing an open, non-solid medium is the way to achieve this. In the case of water cooling, you are mechanically boosting the amount of molecules available to conduct heat. You are also creating a system of total heat conducting molecules that far surpasses the practical volume of a copper slab approach. Pumping the fluid effectively allows more molecules than normal in a given volume to carry heat away. Now, for the sake of argument, you could achieve the same system with a theoretical gas that shares water's coefficient of heat but cannot undergo state change rendering the second option (vapor chamber) unavailable.

A vapor chamber also uses liquid to cool, but it asks each molecule of water to dissipate much more heat by allowing it to go through a state change. The only way water does better than copper is the open space in the interior which allows vaporization to quickly physically transfer the heated molecule to be cooled(otherwise, you're super heating the liquid and not allowing for proper expansion and state change). That is why I say vaporization is analogous to the pump. The liquid requires a quick transfer to a location where it can be cooled not possible in a solid. Yes, it is also true this state change absorbs the brunt of the heat dissipated, but the method is still only feasible because of water's superior coefficient of heat. Similar to the gas and liquid interchange analogy before, you could use a theoretical liquid with the same coefficient of heat as air and it would perform poorly in a vapor chamber. Sure, the ability to change states of said theoretical liquid would be advantageous, but said scenario doesn't even imply a physical chamber design exists such that any volume of said liquid would be able to keep the chip below failure point before it began using state change to cool.

Indeed, a vapor chamber with the physical volume of a liquid cooling method would also be impractical due to the solid nature of the physical exterior of the chamber. The pump method also beats vapor cooling by presenting so many molecules to carry heat than vaporization ever could, even if they carry less heat per molecule. You could probably make a system that matches flow rate such that energy per molecule times number of molecules available from flow in a vapor chamber equals energy per molecule times number of molecules available from flow in a water cooling system. The thing is, due to practicality of design, you can still do much better with water cooling.

So, my original statement was misleading because it implied coefficient of heat was the sole mechanism making a vapor chamber appealing. However, my analogy of pumping to vaporization stands. I hope it is clear now.

 

[Shifty Geezer]

Convection is the correct term. Conduction of heat through diffusion and fluid flow.

Convection requires movement of particles, and doesn't happen in solids. Heat transfer through solids is conduction, and so you don't get convection through a 'solid slab of copper'.

 

[anexanhume]

Convection requires movement of particles, and doesn't happen in solids. Heat transfer through solids is conduction, and so you don't get convection through a 'solid slab of copper'.

I think he obviously meant conduction given he used convention properly initially. But then again, it occurs you were just being difficult as evidenced by one Mr. Winky Face.

 

[orangpelupa]

isn't heat transferred to water is conduction? or maybe my memory is playing tricks at me (its been many years since i got that lesson in elementary school)

im really confused now.

 

[Shifty Geezer]

isn't heat transferred to water is conduction?

Either conduction or radiation to transfer heat into the water, or which conduction will be the predominant method obviously. Then convection causes the heat to move, which moves the heat to the other end in the case of heat-pipes and vapour-chambers, or to the heat-sink or case in the water system. This movement is faster than via conduction through a solid, hence its value. At the other end of the heat-pipe/chamber/water circuit, it's conduction that carries heat away, and it's conduction that dumps it out into the air where it's moved away from the heat source (PC) via convention.

Or in brief, convection is the movement of heat within a fluid/gas. Conduction is the transfer of heat by atomic interactions between particles, passing heat energy from one atom to another in close proximity either in a solid or at the interface betweeen two materials.

 

[orangpelupa]

thanks ^^
your second paragraph that's clear things up on my mind. seems my understanding still ok.

its just my previous sentence is missing 2 words it should be:
"isn't heat transferred from metal to water is conduction?"

 

[Prophecy2k]

Coefficient of heat is still responsible for how quickly phase state change can occur. If your liquid is resistant to heating, heat is forced to flow in the metal shell of the vapor chamber. Evaporation is still the mechanism to move the fluid to a place where it can be cooled. The analogy is still valid, even if phase state change is where a bulk of the heat energy is transferred.

Just to make a correction. The Coefficient of heat is a number that describes the thermal performance of a heat transfer fluid for a given cooling/heating system design. This was the number the previous poster quoted for water and air.

It's actually the Heat Capacity of the fluid (units = kJ/kg °C) which describes how quickly a phase change can occur in a fluid. The heat capacity is the thermophysical property that would more relevant in a comparison between various cooling fluids, particularly when speaking about how quickly said fluids would begin to vaporize. Since liquids will vapourise at a certain temperature (at a given pressure), the heat capacity tells you how much heat can be absorbed by a fluid per mass of fluid for a given temperature change, hence how much heat a fluid can absorb before going from an initial state temperature up to its bubble point (thus point at which boiling/vaporisation occurs).

 

[MrFox]

Thanks all for the explanations :smile:

I'm always confused, but to clarify my comment, since copper is 700 times more conductive than water, i was thinking there was a break even point under which the distance is small enough to make solid copper better than water+movement.

 

[Shifty Geezer]

There will be, but I've no idea what it is. Copper is also more expensive, and I can well see a system like a vapour chamber being chosen over price rather than performance in some cases. But I think in this case you have, say, 200 mm^2 chip surface connecting very quickly to 500 mm^2 heatsink surface via vapour movement, whereas the same 500 mm^2 of heatsink slapped onto the chip would need conduction to spread the heat all the way out. I imagine that's not as efficient, hence the use of vapour chambers, but I can't really say as I'm no expert.

 

[Prophecy2k]

There will be, but I've no idea what it is. Copper is also more expensive, and I can well see a system like a vapour chamber being chosen over price rather than performance in some cases. But I think in this case you have, say, 200 mm^2 chip surface connecting very quickly to 500 mm^2 heatsink surface via vapour movement, whereas the same 500 mm^2 of heatsink slapped onto the chip would need conduction to spread the heat all the way out. I imagine that's not as efficient, hence the use of vapour chambers, but I can't really say as I'm no expert.

I think it's more about temperatures and the mechanism of pure conduction through copper vs convective heat transfer in a fluid.

If you had a big chunk of copper sitting ontop of the chip the copper would heat up and remain very very hot, as the heat is conducted along the mass of the copper towards the end where it is removed through conduction and convertion in a forced air flow. In this way the chip itself would be operating at very high temperatures, which would in itself be counterproductive as a cooling solution.

Instead fluid cooling provides a system whereby as heat is removed from the chip, the hot fluid is transported away, drawing in cold fluid to remove more of the chip's heat. This regulates the chip temperature as you always have a constant supply cold heat transfer fluid to absorb more heat from the generating chip.

In the solid copper solution, since the air-blown surface will also be some distance from the chip, you'd need a relatively large, heavy and expensive chunk of copper for heat transfer across that distance. With a fluid solution, that distance can be much further away, and/or you can save a considerable amount in material costs by effectively having a hollow cavity filled with water (cheapest fluid you can use).

 

[Shifty Geezer]

Instead fluid cooling provides a system whereby as heat is removed from the chip, the hot fluid is transported away, drawing in cold fluid to remove more of the chip's heat. This regulates the chip temperature as you always have a constant supply cold heat transfer fluid to absorb more heat from the generating chip.

Hang on. I was talking about the internals of a vapour chamber. Water cooling is different, as you are physically moving heat away from the chip (convection).

 

[Prophecy2k]

Hang on. I was talking about the internals of a vapour chamber. Water cooling is different, as you are physically moving heat away from the chip (convection).

I wasn't specifically speaking about water cooling, rather a comparison of both vapour chambers and water cooling vs. conduction through a pure solid material. I.e. with solid conduction you have no realistic option for chip operating temp. regulation.

In a vapour chamber you can regulate chip operating temperature by fixing the pressure of the vapour space within the chamber. You usually induce a vaccuum so that the fluid bubble point is low (thus if you're using water, fixing a vacuum allows a boiling point well below 100°C).

The principal difference is still the same. As with copper conduction you have no way of regulating the chip operating temperature outside of manipulating the flow and/or temperature of air at the metal-air interface, the latter of which is generally not possible.

For a given system, vapour chambers & liquid cooling circuits would almost always be the better choice.

 

[IllusionistK]

AMD just announced the 7970M based on a full-fledge Pitcairn core, though under clocked to 850mhz from 1000mhz.

In gaming, the 7870 consumes just 115w, under clocked should put it at 75w or less given the mobile PSU.

It would be cool to see a 7870-based APU paired with an asymmetrical 7970 for next-gen

 

[-tkf-]

So, in regards to the new generation and getting rid of heat and keeping noise in check, is it feasible that with the same amount of space a larger TDP than the original PS3/360 is possible?

Didn´t one of the 360 revisions feature a heat pipe addon?

 

[archangelmorph]

So, in regards to the new generation and getting rid of heat and keeping noise in check, is it feasible that with the same amount of space a larger TDP than the original PS3/360 is possible?

Didn´t one of the 360 revisions feature a heat pipe addon?

Sure but I would assume bigger, hotter chips means more expensive chips + more expensive BOM given the increased cooling requirements...

 

[Brimstone]

According to the people that designed the Nanowick at Purdue, a laptop heatpipe cools 50 watts per centimeter, and the Nanowick design cools 550 watts per centimeter. In one article they mentioned using a hydrofluorocarbon and achieving 1000 watts per centimeter.


How much it would cost to make such a cooler for a laptop or console is the major question.


I suspect just as the next-gen chip designs will achieve over 10x times the performance of the XB360, the XB720 case and cooling solution will be 20x better.