*spin-off* Feasibility of Water Cooling Radiator Setups

[TheChefO]

Another factor in the pro-watercooler camp:

Rising base metal costs.

Copper prices have roughly doubled since 2005, while liquid, plastic tubes, pumps, and fans (needed for either setup) are roughly the same cost they were back then.

The cost difference is shrinking substantially. Not to mention with liquid cooling becoming more mainstream, their price has come down.

So rising costs for heatpipes and lower costs for liquid cooling.

Not sure if it would make financial sense to replace heat pipes with liquid cooling yet, but the raw materials costs are on a trajectory to cross at some point.

 

[almighty]

Water cooling a console could be done with no problems at all. The benefits would be less noise and less heat in general so parts can be run at higher clocks and less heat equals less component stress.

The question is not 'can it be done?' But would it be worth it over existing heat sink designs.

 

[Grall]

The radiators in say Corsair's closed systems can hardly be called thin ...

These radiators are deep because they can be. Space is not a premium (to such an extent, or at least not at fan locations) in almost every PC casing on sale to consumers. A theoretical water-cooled console would have its cooling solution engineered to fit within space constraints.

I'm not sure it would neccessarily be more expensive than a heavy-duty air cooler either, look at the lump of metal Sony strapped into the first iterations of PS3s, they had five long heatpipes soldered to numerous smaller heat columns that connected the pipes to a deep stack of plate metal, along with a giant 160-ish mm blower. That thing surely cost an absolute shit-ton of money to manufacture, comparatively speaking.

One drawback with watercooling is you need an additional mechanical component, the pump, which is another possible point of failure. A heatpipe is a passive object of course, which will function in perpetuity without need of servicing or maintenance.

 

[MfA]

These radiators are deep because they can be.

The same is true for tower coolers.

I'm not sure it would neccessarily be more expensive than a heavy-duty air cooler either

As I said, I think it very unlikely water cooling would get substantially better cooling performance for the same airflow and heatsink surface area as a good heatpipe solution ... but I think it's plausible watercooling could actually be cheaper.

One drawback with watercooling is you need an additional mechanical component, the pump, which is another possible point of failure.

I don't think for low flow solutions it's that much an issue, hell you might consider redundancy.

 

[3dilettante]

I'd be interested in a comparison in the assembly process of a water-cooled console versus one with an air cooler.
As the picture shows, a water cooler is not an integrated hunk of metal that is bolted into one place, but a floppy thing with multiple anchor points and an additional power feed.

Other considerations may be ancilliary, such as storage concerns or failure modes.
A heat pipe contains a very limited amount of fluid, and a multi-pipe design might be able to tolerate a puncture or failure over time of one of them without total loss of function.

The amount of water in a radiator and the lines is higher, and a structural failure in the radiator or hose means complete failure. Retail and wharehousing considerations may also be different, since a failed unit can leak into the box and neighboring boxes.

Water-cooled enthusiast graphics cards are not mass-produced items, and additional labor and manufacturing costs are contained within the higher purchase prices of those items.
The console manufacturers will be worrying about manufacturing a cumulative number of high-end consoles in the tens to hundreds of millions over the course of a product run, and increases in the cost of manufacture add up.

 

[Prophecy2k]

In a console space is limited, as is power.

Something like a water cooling circuit over heat pipes requires far more in terms of added cost to the overall system than you seem to realise almighty.

First consider how heat pipes work (it's a similar principle):

- Heat pipe consists of copper tubing duct containing heat transfer liquid at low pressure (very important - and i'll come to why later).
- Fluid picks up heat via a phase transition from liquid to vapour at the heat source (i.e. processor). (Remember your thermodynamics? Every fluid has a certain capacity to contain heat - Heat Capacity. Fluids however upon changing phase at their bubble point or boiling point soak up substantiably more heat with said phase transition than they would simply sensibly heating from low to high temperature)
- Heat transfer fluid vapour, which has a much greater specific volume than in its liquid phase, travels along the length of the heat pipe down a pressure gradient (therefore no need for a pump to put energy into the system to transport the fluid) to the heat sink (i.e. fans) where the transfer fluid condenses back in to a liquid releasing heat from the system.
- The liquid then returns to the heat source (i.e. processor) via capillary action, which again is important as it is a physical transport phenomenon which negates the need for any external energy being needed to transport the heat transfer fluid.

Note - the benefit of having a fluid in a stiff and rigid closed system like a heat pipe means that you can have your cooling fluid at a very low pressure, thus much lower boiling point. This means that you are maximising your heat transfer fluid by allowing most (if not all) of the heat "take-up" to occur via phase change (therefore greater capacity for heat storage per unit of transport fluid) than if you were allowing some sensible heating for the fluid to reach it's BP.

So how does a cooling water circuit work? - I won't explain that as i believe it's obvious to you, however the principal differences between the two methods are as follows:

1) CW circuits require energy added (i.e. pump) to provide the pressure differential to drive the transport fluid around the circuit. Heat pipes don't. Therefore, lower energy requirement.
2) CW circuits are all single phase systems, therefore don't take advantage of the fluid's considerably higher capacity for heat absorption via phase transition, because you're only taking in heat via temperature change. Heat pipes are two-phase systems and so their heat transfer performance is substantially better (like incredibly so) than any single phase system that relies solely on sensible heating/cooling.
3) Given the limited space in a console box, with cooling water systems, the smaller your transport capillaries become the more power you need to put into the fluid to drive it around the circuit because of the higher pressure drop per unit length of capillary duct (it's actually an exponential increase as you go down below a certain capillary diameter). This is because water is a relatively viscous fluid and thus you have to put in energy to overcome the frictional drag of the water on the inside walls of the capillary. Heat pipes use only physical transport phenomenon and thus require no extra energy input to transport the fluid (hence infinitely more energy efficient).
4) The relative cost of materials for heat pipes, especially given the sizes of those used in consoles, is far far less than the cost of using a cooling water circuit. Not only do the circuit tubing and small energy effiecient high head pumps imply relative high costs, but there's also additional costs implied by the need for sufficient power supply for the pumps, as well as added overall system weight and size dimensions (implying greater shipping and distribution costs).
5) There are also failure and reliability considerations. As heat pipes for the most part won't fail, as they have no moving parts and are internally at a lower pressure than atmopheric pressure, thus more resilient to greater changes in ambient temperature conditions. CW circuits come with a whole host of potential reliability and failure issues, from pump failures (which if you know anything about process engineering, you know that pumps are the most prone things in the world to fail), to capillary failure etc etc.

So in conclusion, i'm afraid that your hopes to see a water cooled console at any point in the future is next to impossible. Economic, power, safety and reliability constraints wouldn't never make it worth it for a manufacturer to implement such a system, when heat pipes do an infinitely better job in a small enclosed space than CW ever would.

The only reason I think you equate CW circuits to being better at cooling in absolute terms is PCs. However in PCs you have your processors sat on boards inside massive cavern spaces when compared to consoles. Heat pipes are limited by their length (as the capilary action phenomenon that allows the condensed fluid to travel bakc to the heat source will only occur over a certain distance) can only get the heat from the cores to the edge of the board. So you still need to get the heat out of the system, and that system being so large is a comparatively looong distance. A CW circuit in a PC provides a direct path from the chip to the external shell of the PC unit and so benefits the system by providing such. In a console with much less space internally (especially well designed ones like the PS3's solution), you can engineer the path itself such that the path to your external air fans has the lowest pressure drop possible, therefore reducing the power requirement of your external air fans, by comparison a cw circuit with its own additional power requirement and complexity simply isn't needed.

Hope that helps ;-)

 

[almighty]

I've built enough water loops over the years to know the costs and the ins and outs thanks. So much so that I don't even water cool my PC anymore.

Water cooling is better then heat pipes no matter the heat load...period.

 

[MfA]

Water cooling units generally use hoses, but I don't see why they couldn't use hard plastic fittings.

 

[Shifty Geezer]

My first thought at this suggestion is a peristaltic pump. You could run it off the same fan motor and save the complexity and risk of a conventional pump. I don't know how they cope with wear though, and I guess the tube would give in eventually. Or a need to increase fluid viscosity above just water (if that's needed) might adversly affect performance.

Fundamentally though, I don't understand why water is more effective than not when ultimately you need to get the heat out of the case, which means getting it into the air, which means a thermal interface between material (heatsink) and air and needing an airflow. Water has a high energy capacity so can store a lot of heat as it's pumped around, but the rate at which is dumps it back into the air is no different to any other medium. So why would a water system transporting heat to a HS+ 120mm fan be more effective than heatpipes transporting heat the the same HS+120mm fan?

 

[Prophecy2k]

I've built enough water loops over the years to know the costs and the ins and outs thanks. So much so that I don't even water cool my PC anymore.

Water cooling is better then heat pipes no matter the heat load...period.

I'm a chemical process engineer by profession. I design cooling water loops for entire process plants ;-)

What you said (bolded) is simply not true. It's thermodynamically impossible for a cooling water system on such a small scale to remove as much heat from a core per unit of cooling medium.

 

[Prophecy2k]

Fundamentally though, I don't understand why water is more effective than not when ultimately you need to get the heat out of the case, which means getting it into the air, which means a thermal interface between material (heatsink) and air and needing an airflow. Water has a high energy capacity so can store a lot of heat as it's pumped around, but the rate at which is dumps it back into the air is no different to any other medium. So why would a water system transporting heat to a HS+ 120mm fan be more effective than heatpipes transporting heat the the same HS+120mm fan?

This is the basic gist of it, however there is the internal processor fan and then the external system shell fan. Heat pipes will carry the heat to the internal processor fan, and then it's down to the air to carry it out of the system shell entirely. On a console the core processor fan can be so close to the external shell that the cost and complexity of a cw system would be redundant, or if the unit is small enough it can be engineered in a way such that there is only one fan which is the external shell device. On PC water cooling is beneficial because unlike heat pipes, your water circuit can remove the heat directly from the core (albeit less efficiently that heat pipes) and carry it all the way to the external shell fan, which in some PC systems can be a considerable distance away.

Either way, like you said shifty heat removal to ambient air at the external shell interface is the limiting factor. But unless you game in a freezer, there's really no solution for this.

Re your suggestion regarding peristaltic pumps, it's an interesting idea. Although i'm not sure of the limitations of such devices, as i'm not all that familiar with them. Off the top of my head though iirc they work best with high viscosity, high density fluids like oils or slurries, and i'm not sure you would be able to get a device that could provide a sufficient pressure differential for your low capillary diameter cooling circuit. I'm not all that sure though, so i'll check that out a bit more ;-)

 

[Prophecy2k]

Water cooling units generally use hoses, but I don't see why they couldn't use hard plastic fittings.

The problem you have then, and especially in a console application is routing and pressure drop where you have bends in the line. For flexible hoses you can route a system in a tight enclosed space with rather relatively large radius bends and thus less resistance to the flow of water around the circuit. The lower the resistance to flow, the lower the pressure drop around the circuit and thus the lower the power requirement on the pump (as the pump needs to provide the pressure differential to drive the fluid around).

With rigid plastic fittings in such a cramped space you'd invariably need to use very small radius bends (i.e. your 90 deg junctions) so you have a lot more pressure loss in the system and would need a more powerful pump.

 

[AlphaWolf]

In a console (custom box) there's no reason they couldn't use hard custom fitted tubing with whatever angles (or curves) they wanted. It still fails to solve many of the issues a water cooling would face in a console that is shipped all over the world.

 

[almighty]

All the tests that I've seen of water cooling vs air cooling on PC the water cooling has ALWAYS given the better results.

The water in a loop is only normally 4-5 degrees C hotter then the ambient temperature during load which is why it works so well.

Under the same conditions a heat pipe would be much hotter over ambient.

My AMD 5850's had monster third party coolers in them with three 8mm thick heat pipes and a massive fan. My 5850's loaded in the mid 90 degrees C.

I then bought some £20 water blocks for them and my load temps quite literally more then halfed despite these blocks being much much smaller then the heat sink they replaced.

 

[Shifty Geezer]

All the tests that I've seen of water cooling vs air cooling on PC the water cooling has ALWAYS given the better results.

The water in a loop is only normally 4-5 degrees C hotter then the ambient temperature during load which is why it works so well.

That's what I'm thinking makes the difference. Basically the water works as a heat reservoir, absorbing more heat energy. I guess as it heats up, the speed at which heat is lost to ambient increases, offsetting the decrease in heat absorbtion from the chips. I'm sure there's all sorts of numbers that can be crunched to explain what happens. But if so, surely the size of the reservoir is important? This reminds me of the oil-bath PC that Al linked to a while back. That cooled passively, using the whole surface area of the PC to disipate heat from the oil. My ill-informed gut feeling is that you need a decent amount of water for it to work.

 

[almighty]

That's what I'm thinking makes the difference. Basically the water works as a heat reservoir, absorbing more heat energy. I guess as it heats up, the speed at which heat is lost to ambient increases, offsetting the decrease in heat absorbtion from the chips. I'm sure there's all sorts of numbers that can be crunched to explain what happens. But if so, surely the size of the reservoir is important? This reminds me of the oil-bath PC that Al linked to a while back. That cooled passively, using the whole surface area of the PC to disipate heat from the oil. My ill-informed gut feeling is that you need a decent amount of water for it to work.

You don't need a reservoir, I've run loops without them and seen the same ambient temperature deltas as loops with them.

I don't water cool anymore, Got bored with it.

I now use a single stage phase change cooler, My CPU idles at -45c and loads at -20c

 

[hoho]

You don't know a lot about water cooling do you? No matter how many heat pipes you stick on a core a water loop will ALWAYS provide much much better temperatures.

How is the airflow for the heatpipe setup compared to heating your water radiator has?

You seeing lower temperatures with water doesn't mean it's because water magically cools stuff better. It's because you are cooling the water better. You still need to get exact same amount of heat energy out of a small contained area. It doesn't matter if you are storing it in heatpipes+radiator or water+radiator before you blow some cooler air on it.

How good were you in physics in highschool/university?

 

[almighty]

How is the airflow for the heatpipe setup compared to heating your water radiator has?

You seeing lower temperatures with water doesn't mean it's because water magically cools stuff better. It's because you are cooling the water better. You still need to get exact same amount of heat energy out of a small contained area. It doesn't matter if you are storing it in heatpipes+radiator or water+radiator before you blow some cooler air on it.

How good were you in physics in highschool/university?

Air flow was good, Was one of the best coolers around for the 5850.

The base of an air cooler is cooled by air pushed by a fan, This air can be anything from 10-20c above ambient temperature.

The base of a water block is cooled by fluid that is only 5-6c above ambient temperature.

I don't do physics, Instead I rely on my 10 years of experience with water cooling and big ass air coolers.

Temperatures are always going to be better on water as the ambient temperature deltas are always going to be cooler then what you can ever achieve with air cooling.

Water is just faster and more efficient at removing and transferring heat.

 

[AlNom]

That's what I'm thinking makes the difference. Basically the water works as a heat reservoir, absorbing more heat energy. I guess as it heats up, the speed at which heat is lost to ambient increases, offsetting the decrease in heat absorption from the chips. I'm sure there's all sorts of numbers that can be crunched to explain what happens. But if so, surely the size of the reservoir is important? This reminds me of the oil-bath PC that Al linked to a while back. That cooled passively, using the whole surface area of the PC to disipate heat from the oil. My ill-informed gut feeling is that you need a decent amount of water for it to work.

Yup. So everything has a heat capacity, which is the amount of energy it can absorb before its temperature rises by a degree C (or Kelvin). Everything also has a thermal conductivity, which is essentially a description of the rate of energy transference through a medium, this is important for heat dissipation.

Water has a heat capacity of ~4.2kJ/(kg Kelvin), copper is significantly lower @ 0.385, and Al is higher @ 0.897. You can clearly understand how the amount of material is important - less material, higher temp per thermal energy)

So why bother using copper? Because of the thermal conductivity (W/meter K). This is why you see copper cores at the center of the heatsink solution or copper heat pipes. (Cu is 401, Al is 237 or lower if not pure Al, and water is a pretty awful 0.6).

Ideally, you want to remove heat from the chip the fastest, transfer it to another material with a large heat capacity, and then finally get rid of the heat to the air, which is where surface area, ambient temperature and external factors (i.e. fan speed) play a role.

In either case of water cooling or metal heatsink, you want copper to start with. From there you're going to have to balance out the amount of water or the amount of aluminum to which you're going to shunt that heat. Clearly, water's high heat capacity means your temps aren't going to rise much (per kg). On the other hand, Aluminum will naturally be pretty hot relatively. The last step really is transferring heat to the worst conducting medium - air.

In the case of aluminum, it's a pretty straight forward step - the greater the surface area, the greater capability to shunt thermal energy away. With the water cooling, the heat within the water has to transfer to a radiator because, let's face it, we don't have a pool of huge surface area to expose to the air. But we know that heat conductance is pretty shit for water, so getting rid of the heat will depend on the surface area contact between the water and the radiator fins, of which there should be lots, because there's the final step of removing the heat from the circuit and into the air.

So... what does this mean? Well, we should probably stick to a particular weight limit because the company is going to have to ship millions of these things. For both, we might just assume the same size fan and keep that out of the equation just because it'll have to be a limited size for a console anyway.

An upper limit of 1kg for either cooling solution might be possible. That's 1kg for water + radiator + tubes + processor blocks, or 1kg for the copper/aluminum heatsinks.

The ultimate heat capacity of either solution should be easy to ballpark, but you might consider that the amount of water will be fairly low when you factor in the rest of the components as well as the amount of tubing that can fit within the console (aluminum radiator, copper processor blocks, tubes, the pump).

And that's one big question isn't it: how much tubing can you fit in there such that everything adds up to our designated maximum weight limit? I'd say the copper blocks are easiest to work out for weight given the surface area it needs is going to depend on the processor's heatspreader. Then it just only needs enough for water to pass through within. How much may depend on how powerful the motor pump is since you need to push the water back out of the confines of the copper block. How big does the radiator have to be? Well, I suppose you don't want the fugly piece of shit to be too obvious on the console. I don't know... how about no large than the dimensions of the console so that it's not "hanging out" nor just asking for it to break off by accident.

It's easy to figure out how much aluminum and copper you want for the conventional solution. The easiest is perhaps a copper core that matches the surface area of the heatspreader on the CPU or GPU, and then have the rest of the heatsink be aluminum. You can work out the height needed from the densities of aluminum and copper. Whatever the thing actually looks like (after molding the fins and whatnot) is irrelevent. We're looking at keeping the weight consistent.

Lastly, there's the cost, and I gather this is why Mfa brought it up, but just looking at the components and amount of the key materials (copper, aluminum, water), the BOM at least could very well be lower for water cooling, but that does not include the increased costs for assembly nor any consideration for reliability (hard to put a dollar figure on faults per thousands or millions of units).

And that's all I have to say for now.

 

[hoho]

The base of an air cooler is cooled by air pushed by a fan, This air can be anything from 10-20c above ambient temperature.

The base of a water block is cooled by fluid that is only 5-6c above ambient temperature.

And how will you manage to get that cool air inside a tiny console box?

I don't do physics, Instead I rely on my 10 years of experience with water cooling and big ass air coolers.

Experience also told people heaveir than air stuff can't fly, unless you are a bird

Water is just faster and more efficient at removing and transferring heat.

When the distance that water has to travel is just a few cm's I'm not that sure and the main problem, actually cooling that water down, still remains just as hard as with regular air coolers.