Water-cooled PCs and heat dissipation

[El Greco]

I've been interested in watercooling lately, and -as in everything else- I see a lot of unsupported claims. For those of you unfamiliar with watercooling, here's the gist of it:

People have been using water to cool the heat-generating components of their PCs for various reasons, mainly better temperatures and quiet operation without fans. The main components of a watercooling loop are:

- The "blocks", usually made of copper or aluminium, which can hold about 50ml of water and are placed on the processor, graphics card chips, and any other PC component that gets hot.
- The pump, which pumps water to the blocks through a system of flexible tubes.
- The reservoir, which serves the purpose of easily filling and refilling with water.
- The radiator, usually made from aluminum, which forces the water to run through narrow paths and thin fins in order to dissipate water to the air through the metal.

This is a typical setup. Many people however have done away with the reservoir, using just a T-shaped connector somewhere in the tubing in order to fill the system with water. Those people don't notice a drop in performance. This is understandable, since most reservoirs don't hold much water anyway - more like 100-150ml. They are made small because many people like to mount them inside their cases.

Omitting the radiator though, is almost unheard of. I say "almost", because in a way this is what Zalman did with the Reserator (http://www.advanceddesignky.com/pictures/RESERATOR1PLUS.jpg) (from Reservoir and Radiator). The Reserator capacity is much larger than that of other reservoirs - it can hold about 2.5 lt. Obviously, the lack of a classic radiator with large dissipation area is compensated for by the increased water mass.

My question is this: If we keep all other variables (environment temperature, pump, waterblocks etc) stable, can we predict how much we need to increase the mass of water as we decrease the radiator dissipation area ? Let's say for example that we have a Cape Cora (http://www.madshrimps.be/?action=getarticle&articID=361) radiator with a total surface area of about 15000cm2, and we want to substitute it for a Reserator, which will hold 2,5 lt more water, but has only about 1250cm2 dissipation area. Can we predict which will have a better cooling performance ?

[RecoveringYuppy]

My first (not too well thought out) thought: Total energy dissipation is proportional (loosely, it's not exactly linear) to surface and the temperature difference between the surface and surrounding environment. So with a lower surface area you're only going to get comparable energy dissipation from the smaller unit at a higher temperature. So the larger reservoir just buys you a longer time to get to that higher temperature. So "comparable performance" would depend a lot on how long you let your equipment run at a time.

I would think that in continuous operation the smaller surface area will have to run at a higher temperature to get the same dissipation.

[Michael Redman]

It seems to me that the larger mass would work better until it came up to temp, but once it did, only surface area would matter.

[El Greco]

Let's put it in another way too: Assume a modern PC radiates about 200 Watts of heat. There must be some mass of water that will be able to maintain a good temperature of the critical PC components with the only dissipation happening through a perfectly cylindrical aluminium tank (no fins at all). Can we calculate what that water mass would be ?

[Ivor the Engineer]

All things being equal, the larger surface area should have the better cooling performance over a long time period.

To cool the components, thermal energy has to be dumped into the environment eventually, so it is the ability of the heat-sink (i.e. radiator or reservoir) to do this that limits the performance.

The reservoir system will likely have a larger thermal mass though, so for short time periods it may remain cooler than a system using radiator with a large surface area but a smaller volume of water.

(That's my guess anyway;))

[El Greco]

All things being equal, the larger surface area should have the better cooling performance over a long time period.

Better compared to how much more water ? If we had a reservoir that could hold 1000 lt of water, what dissipation area would be needed to equal that ?

I mean, doesn't it depend on the total water mass ?

[Terry]

I mean, doesn't it depend on the total water mass ?

Not in the steady state, no.

[RecoveringYuppy]

Comparing the water mass to the surface area is an apples and oranges comparison. The mass of the water determines how long the temperature will take to rise. The surface area determines the rate of cooling (for a given temerature difference).

[Michael Redman]

Unless the finned coolers on the market are massive overkill, I would think a smooth cylinder would need to be very large to dissipate enough heat.

[El Greco]

Not in the steady state, no.

Why not ? The Reserator is a proof that it works. If it doesn't depend on the water mass then I could run the Reserator with just a little water and have the same results.

Even water that isn't in direct contact with the metal will dissipate heat to cooler water in the perimeter. So the steady state of a system with a large water mass may be achieved at a lower temperature than the steady state of a system with large dissipation area.

[Hellbound]

You have to consider these measures:

1. How much heat energy does it take to raise the temperature of the reservoir 1 degree? Using water, this would be based on it's specific heat. Same for both types of systems.
1a. How long does it take the computer to produce this amount of heat energy?
2. How much heat energy can the system radiate to the environment at given environmental conditions (typically 75 degree air). This will be differnet for each system, and I'm not sure how to calculate this from surface area (I'm no engineer :)).
3. At what reservoir temperature does the answer to question 2 equal the answer to question 1a. This should give you the temperature of the system at the steady state for both systems.
4. Now, figure out how long it would take to get the reservoir to the steady state temperature. This will be a different value for each system, as well.
5. Using this, you should be able to figure out how long the large reservoir system would remain cooler than the radiator system.

In response to your past post, which was done as I typed this one, yes, you can transfer heat to the cooler water. But that water must also transfer the heat to the environment, or it will get hotter with time. You account for this in step 1 above.

[El Greco]

Comparing the water mass to the surface area is an apples and oranges comparison. The mass of the water determines how long the temperature will take to rise. The surface area determines the rate of cooling (for a given temerature difference).

So what you are saying is that the Reserator can be run with just 100ml of water? I very much doubt that.

[El Greco]

5. Using this, you should be able to figure out how long the large reservoir system would remain cooler than the radiator system.

Isn't it possible that this time could be infinite, depending on the specifics ?

Let's assume both systems have already reached the steady state. We don't care about how much more time will take with the large reservoir.

[RecoveringYuppy]

So what you are saying is that the Reserator can be run with just 100ml of water? I very much doubt that.
No, I'm not saying that. I'm saying the water mass determines how fast the temperature of the water will rise. If you put less water in to it will reach the steady state that several people have mentioned faster.

(There may also be some problems with the heat flow through airspaces caused by not filling up the reservoir but that seems a bit beyond this conversation now).

[RecoveringYuppy]

Let's assume both systems have already reached the steady state. We don't care about how much more time will take with the large reservoir.
The steady state temperature will be higher for the box with less radiator area.

[Hellbound]

El Greco:
Isn't it possible that this time could be infinite, depending on the specifics ?
No. You'd need an infinitely large reservoir. It will have a steady state somewhere.

Let's assume both systems have already reached the steady state. We don't care about how much more time will take with the large reservoir.
Then the system that dissipates more heat to the environment will be cooler, period. The radiator system wins.

There is a physical limit: eventually, as you make your reservoir larger, you're going to have the surface area closer and closer to that of the radiator system just because you need a bigger container for it :) Additionally, to get technical, you'd need to consider the material the container/radiator was made from, as well, as differnet materials would have different thermal coefficients and stuff (not sure if my terminology is correct-but differnet rates fo heat transfer to air).

However, depending on the amount of heat input you're talking about, the time for the reservoir to reach steady state may be measured in days or weeks for the large reservoir (compared to the small one). So while the steady state of the reservoir system is going to be higher, no question about it, the time it takes to reach that state may be so long that it's functionally not important. Especially considering that modern computers will usually go into sleep and power-save modes, and have varying levels of function that will modify the amount of heat generation (and allow the less-efficient reservoir system time to "cool off").

So while the radiator system is definately more efficient and with a lower SS temp, the reservoir system may be functionally better, depending on hwo the equipment in question is used.

[El Greco]

No, I'm not saying that. I'm saying the water mass determines how fast the temperature of the water will rise. If you put less water in to it will reach the steady state that several people have mentioned faster.

Eventually the Reserator does reach the steady state. It's running 24/7. We don't care about how long it takes to reach the steady state. We care about what happens after. Will running it with 100ml of water result in the same final temperatures ? (assume that the dissipation area remains the same, ie I the water makes contact with all of the Reserator surface).

[Hellbound]

Will running it with 100ml of water result in the same final temperatures ? (assume that the dissipation area remains the same, ie I the water makes contact with all of the Reserator surface).

Yes to your question. IN reality, we both now that with a lower amount of water, the reservoir won't be full, so the contact is less, etc, etc, etc. But assuming you had a smaller reservoir, the steady state temp depends only on the temperature at which heat input to the system equals heat dissipated by the system.

I'd also debate that the reservoir reaches a steady state, even running 24/7...but I'd have to actualyl break down and do (ugh) math to figure it out for sure ;) This is because systems will run at varying power levels (and, consequently, varying heat levels) depending on their current task and status. The only way it's running at steady state is if you have a continuous, consistent load running on the system 24/7 with no variation.

It sounds like this is simply two approaches to the problem. Traditional systems are designed to run in the steady-state mode, so large radiators help lower the steady state temp. This would be useful for systems under a constant load, with little-to-no downtime, such as server systems or supercomputers or something that runs batch jobs or processes overnight. They can also work well with varying loads. The reservoir system sounds like it might actually keep temps lower for a system that sees intermittant or varying loads, where the system has cool-down time between times of high-activity. A home gaming system, or a graphics design workstation that's shut off (or not used) after-hours might be an example.

[RecoveringYuppy]

Eventually the Reserator does reach the steady state. It's running 24/7. We don't care about how long it takes to reach the steady state. We care about what happens after. Will running it with 100ml of water result in the same final temperatures ? (assume that the dissipation area remains the same, ie I the water makes contact with all of the Reserator surface).
You're asking about a comparison of two Reserator's running with different amounts of water? Same final temp, different times to reach it.

[El Greco]

Yes to your question. IN reality, we both now that with a lower amount of water, the reservoir won't be full, so the contact is less, etc, etc, etc.

Isn't the contact the same per gr of water ? And won't a smaller water mass cool faster anyway ?

[El Greco]

You're asking about a comparison of two Reserator's running with different amounts of water? Same final temp, different times to reach it.

Ok, now I'm intrigued. I'll try it out and you'd better have a good explanation if it doesn't work :D

[Hellbound]

Isn't the contact the same per gr of water ? And won't a smaller water mass cool faster anyway ?
Well, assuming the same reservoir container size, then the top part of the water (that touches the metal/radiator when full) touches air now. That goes a lot beyond what we want to get into (or what I'm capable of).

In any case, RY ansered it much more conscisely than I, above.

Also, I edited my previous post, so I don't know if you saw all my comments. Just wanted to point that out :)

The smaller water mass will cool faster, but it will also heat up faster. That's what you have to keep in mind. You can't consider the computer->water and water->radiator->environment exchanges as isolated...the steady stae is the state where the computer->water transfer equals the water->radiator->environment transfer.

[El Greco]

And if it is so, I'm really puzzled as to why they made the Reserator so big and bulky :con2:

[Hellbound]

Ok, now I'm intrigued. I'll try it out and you'd better have a good explanation if it doesn't work :D

I would NOT do this!!!

IT may fry your system, especially if my thoughts about it's intended function are correct. I supect it's made to work at varying temperatures, the idea being that it never reaches the steady state temp but always fluctuates below it (rising temp slowly during high use and cooling slowly during low/no use). By lowering the amount of water, you could very likely change a time to steady state (TTSS) from days to hours, or even minutes. Again, I suspect the system is NOT designed to run at tthe steady state temperature (which seems to be the major difference between it and the radiator systems).

I'd actually crunch the numbers, or see if you could get someone to crunch them, to get some ballpark figures on this before trying anything like it.

[El Greco]

Well, assuming the same reservoir container size, then the top part of the water (that touches the metal/radiator when full) touches air now. That goes a lot beyond what we want to get into (or what I'm capable of).

In any case, RY ansered it much more conscisely than I, above.

Also, I edited my previous post, so I don't know if you saw all my comments. Just wanted to point that out :)

The water doesn't touch the metal, the tower is 3/4 full.

The testing happens like this: People run benchmarks that stress all heat-generating components as much as possible and they continue for as long as the temperatures keep rising, usually several hours. After a point there is no further increase in temperature no matter how many hours (or days) you keep going on, so we can assume that practically we have reached the steady state.

[El Greco]

Don't worry about the system, temperatures are being monitored and even if something goes wrong it just shuts down. All modern PCs do that.

[Hellbound]

And if it is so, I'm really puzzled as to why they made the Reserator so big and bulky :con2:

Well, I've tried to get the idea across in my last two posts, but the thread has moved fast, so you may have missed them :)

So, as analogy, let's use a car. They have a reservoir and a radiator.

Cars are actually designed along similar principles to the Reserator, if my theory is correct. A car cooling system is not made to keep the car cool if you are driving it consistently at maximum speed for long periods. The fan can't get rid of the heat the motor is producing, and the car will overheat.

However, no one drives like that, even speed demons. We stop for traffic lights, slow down in residential zones, speed up to pass, slow down to avoid things, etc, etc, and generally run at a speed much lower than the maximum the car is capable of. For short periods, the cooling system can cope with excessive speed by letting the reservoir heat up a bit. Later, when the car slows down or stops, it gets rid of that excess heat.

However, if you run with, say, your radiator half-full, then the time the system can support hese higher activity levels goes down, because the reservoir can't hold as much heat without reaching critical temperatures.

Think of the reservoir as a "heat jar". The computer adds heat to the jar, and the radiator takes heat away. With a very large radiator that takes away a lot of heat, evena very small jar can cope with the heat the computer puts in. With a small radiator that doesn't take away as much heat, you can very quickly hit problems if you get too much heat in. However, if the heat being added to your jar varies (more sometimes, less other times), then your radiator only needs to be able to take out the average amount of heat that gets put in. The larger your heat jar (reservoir), the longer the system can run in a high-heat input state before the jar fills up and overflows.

Two advantages I can see immediately to such a system are:
1. Less energy used for the fan/radiator/cooling system. This could be designed to use a smaller, or quieter fan, for example.
2. More resistence to a failure of a fan or other active cooling measure. The heat jar is larger, so may even be able to give you time to shut a system down safely even if active cooling measure should fail (as oppised to traditional cooling systems, where the time from cooling failure to CPU death is measured in seconds at best).

[El Greco]

Well, I've tried to get the idea across in my last two posts, but the thread has moved fast, so you may have missed them :)

I think I understood it, that's why I mentioned how the testing is done: Since we keep stressing the computer until we see no further rise in temperatures, isn't it safe to assume that we have reached the steady state under maximal load ? And if we know the temps of the maximal load steady state, shouldn't we observe the same temps (only faster) with much less water ?


And while we're at it, let's move to the second gray area: Copper blocks are largely considered better than aluminium ones, because of the greater thermal conductivity of copper. With the same logic as above, isn't copper just going to warm water faster but at the steady state be exactly as effective as aluminium ?

[ben m]

And while we're at it, let's move to the second gray area: Copper blocks are largely considered better than aluminium ones, because of the greater thermal conductivity of copper. With the same logic as above, isn't copper just going to warm water faster but at the steady state be exactly as effective as aluminium ?

There's a small temperature difference between the chip surface and the water---otherwise heat would not flow between them via the metal. The conductivity of the metal determines this temperature difference.

For the same heatsink, etc., the equilibrium temperature of the water will be the same no matter how the heat got there; the Cu/Al difference tells you how the heat gets there. You can put 200W into 60C water by either (a) connecting it to a 70C 200W heat source via 1/4" of copper or (b) connecting it to a 80C 200W heat source via 1/4" of aluminum ... or, for that matter, (c) connecting it to a 500C, 200W heat source via fiberglass insulation. (Numbers made up, just for illustration.)

Ditto for the heatsink---heat needs to flow through the metal to get from the water side to the air side. You'd prefer this link to be high-conductivity rather than low-conductivity.

[Hellbound]

Okay, the Reserator uses Anodized Aluminum for the material, with a radiative area of 1.274 m2 (which is about 12,740 cm2: 10,000cm2 to a m2).

The Cobra 622 review states "When this is multiplied six times combining all six sections this equates to an approximate total surface area of 15,328-square centimeters".

So, sadly, both have similar cooling areas, so the premise you based the discussion on is false. Sorry, I didn't think to check the m2 to cm2 conversion :)

This means the difference in steady state temps is going to be small.

ETA: The large reservoir is going to provide more resistence to temperature change, so I suspect it will give a smoother temp curve over time than small reservoir systems.

[Michael Redman]

And while we're at it, let's move to the second gray area: Copper blocks are largely considered better than aluminium ones, because of the greater thermal conductivity of copper. With the same logic as above, isn't copper just going to warm water faster but at the steady state be exactly as effective as aluminium ?

The water will also warm the copper faster at the other end, so the heat the water takes away from the source can be dumped into the air faster.

[Hellbound]

The water will also warm the copper faster at the other end, so the heat the water takes away from the source can be dumped into the air faster.

Edited to remove, I though you were still talking about the reservoir sizes rather than copper/aluminum blocks :)

On the blocks, I'll second that. If you get to the nitty gritty, you have a whole chain of heat transferrrences:

CPU->Block->Water->Radiator->Fins->Air
| | | |
V V V V
Case Case Tubing Air
air air | |
V V
Case Air
air

And probably more besides. Anyway, the block material facilitates the CPU->Block and Block->Water pieces.

[El Greco]

Okay, the Reserator uses Anodized Aluminum for the material, with a radiative area of 1.274 m2 (which is about 12,740 cm2: 10,000cm2 to a m2).

That's a mistype from your source; it's 1274 cm2. 1274 m2 would be awfully large, about the area of a football stadium :)

[Hellbound]

El Greco:

That's 1.274m2.

A size comparable to other sinks. It's 10,000cm2 per m2, not 100cm2 per m2. You (or someone) is using an incorrect conversion factor. My source is the same site you linked to for the picutre:
http://www.advanceddesignky.com/viewitem.cfm?id=3856

[El Greco]

El Greco:

That's 1.274m2.

A size comparable to other sinks. It's 10,000cm2 per m2, not 100cm2 per m2. You (or someone) is using an incorrect conversion factor. My source is the same site you linked to for the picutre:
http://www.advanceddesignky.com/viewitem.cfm?id=3856

Ah I see. You mean 1-point-274, not 1,274. You see, the European system... :D

But it's not a matter of comparison between those radiators anyway. I was more interested in the role of the water mass.

[Hellbound]

Well, then, the water mass will jsut make it more resistent to tempeerature change in either direction :).

What was confusing me was the huge differnece in surface area of your first post, which was why the only way I could think of that this would work is the model I was describing.

However, now that the surface area issue is resolved, it makes more sense. The SS temp will be similar. The water will just help keep it more level.

[El Greco]

Ok, so far we have that water mass doesn't matter to the final temps and copper is better. Now something else: Because people prefer copper blocks and most radiators are from aluminium, there have been many concerns and reports about copper suffering corrosion due to the battery effect. I have several questions about this:

1) Can anodized aluminium prevent this ? What about gold-plated copper ?
2) Are there any non-commercial substances that act as anti-corrosive and can prevent this ?
3) Can propylene-glycol (which is added for its anti-freezing properties) also prevent corrosion ?

ETA: I've been thinking about ascorbic acid and zinc oxide but I don't know how effective those would be and what concentrations would be required.

[beelzebub]

Sorry for butting into this conversation, but aren't you all barking up the wrong tree? The real problem here, is to move heat from one location (CPU) to another one (The room). The mass of coolant is largely irrelevent, the speed of transfer, however, is most certainly relevent. If you have a small quantity of coolant (read thin pipes), then you need to move it through the system quickly - a large amount of coolant (read thick pipes) could circulate slower. The amount of heat per unit time removed, is dependent on the volume of coolant passing over/through the block per unit time (proportional to the coolant's specific heat). The key is that the radiator must be able to transfer all the heat, generated from the CPU, from the coolant to the air as quickly as the heat is being generated.

CPU => Coolant => Heatsink => Air

Relying on the mass of the water alone is not going to work - unless you can get rid of the heat, the water temperature will quickly rise, and it's ability to remove further heat from the CPU will degrade.
Experiment!! - My 2kW kettle just took 3 mins to boil 1 litre of cold tap-water. If a CPU dissipates 100 watts, then 1 litre of coolant would boil after only 1 hour, if there was no radiator. 10 litres would boil in less than half a day, and so on. Clearly, quantity of water is not what is required. Speed of heat transfer, on the other hand, is.

[El Greco]

Clearly, quantity of water is not what is required. Speed of heat transfer, on the other hand, is.

This is another long subject, the flow rate (and then we have the diameter of the pipes). But if we have a pump that pushes 300 lt/h vs another one that pushes 600 lt/h, doesn't this mean that the first one will transfer the heat away from the CPU at a slower rate but will also allow the warm water to cool for more time in the radiator ?

[Taffer]

I've always been tempted to make a passive pump water cooling system... Problem is, my GPU is an overheating wanker.

ETA: If you have the cash to spend, you could always get this (http://www.zalman.co.kr/eng/product/view.asp?idx=64&code=020). :D

[Michael Redman]

Sorry for butting into this conversation, but aren't you all barking up the wrong tree? The real problem here, is to move heat from one location (CPU) to another one (The room). The mass of coolant is largely irrelevent, the speed of transfer, however, is most certainly relevent. I don't know from personal experience, but I would guess that the rate of flow for these systems exceeds the point where this is an issue. It only helps to push the hot water as fast as the radiator can cool it, and that seems to be the thermal bottleneck. pushing the water faster, as El Greco points out, simply return hot water to the hot components.

If you have the cash to spend, you could always get this (http://www.zalman.co.kr/eng/product/view.asp?idx=64&code=020). :D

I like that. I've been messing around with my box today, and thinking that there's plenty of radiating surface right there.