Water Cooling on the cheap (really cheap)

[Pablo Escobar]

I fancy giving WC a go but don't want to spend a fortune.

Don't know if anyone has tried this b4 but how about using a Central Heating pump to pump the water round. U can pick them up for about a tenner they pump upto 1 L / Second and push a 5m head....

Plan to stick the pump, a couple of reclaimed car radiators & fans and a big tank/res in the roof of my house and run a couple of pipes into my PC room.

Question is do you know of any waterblocks that would be capable of running with a Litre / second... Sounds like loads to me.
Would a simple pass- through design (no funky channels) still shift the heat considering it's moving that amount of water (cause I'm thinking of making one).

Am I mad?

Will it work?

[Random Nonsense]

been there done that All waterblocks will restrict flow a LOT in this kind of set up. Also the pump at full speed is as hot as another CPU. If you have the cooling capacity go for it

[Kaitain]

Finally we've got the message home... you need flowrate. More flow = better cooling to a point. To a point. People with elementary fluid dynamics knowledge have written equation after demonstration after equation to prove that the higher the flowrate, the higher the theoretical maximum cooling rate.

Well, dealing with abstract numbers like that is one thing... dealing with real numbers is another.

Say you lumped a kilo of water directly onto the CPU - kind've a "beehive cooler" on acid. Water has a specific heat capacity of about 4.2 kJ/(kg.K) at 20ºC, so let's assume that's our baseline temperature.

You have an Athlon CPU overclocked pretty high... it puts out a nominal 100W as heat at full load. Let's assume this is constant, even heating. Now, for that kilo of water to heat by 1ºC (= 1K) you need 4.2kJ = 4200J of energy. At 100W = 100J/s that's 4200/100 = 42s to heat that kilo of water. That's static water.

Now assume your average waterblock has a capacity of about 10cc. In truth it's probably more than that. 10cc = 10g = 0.01kg at 20ºC. 0.01*4200= 42J/s... which means in about 0.4 seconds is needed to raise the temperature of the water in the block - if static - to 21ºC. Now flow the water so that it changes every 0.4s... that's 25cc per second or 1.5 litres per minute at ideal heat transfer conditions. Now, realistically, you'd have less than ideal heat transfer from block to water... the exact efficiency is subject to a whole subset of different equations (suggest you look up the Sieder-Tate equation) but assume it's only 40% of ideal as a very conservative guesstimate. Even then it's only about 3.75 litres/minute. So my question is, why on earth are people investing in central-heating pumps, when fish-pond pumps and proper design are probably better suited to the task?

[Pablo Escobar]

Originally posted by Kaitain
So my question is, why on earth are people investing in central-heating pumps, when fish-pond pumps and proper design are probably better suited to the task? [/b]

Kaitain,
I can understand your reasoning about the low flow rate required based on your equations but my objective here is not to "see how much flow I can get" it is merely to make a WC kit as cheaply as possible.

As you probably know a decent eheim pump (seems to be the standard) is going to cost the best part of £40 + £40 for a rad + £40 for a block + extras. I'm guessing my "scrap yard" approach would cost me roughly £40 all in. I was just wondering whether:

a) anyone else had tried it.

b) will it work.

Thanks for taking the time to reply though.

[drow_elf]

if your attic gets hot- so will your cpu (since the ambient temp of the rad)

on second thought, though, it might not be a problem if the sun never shines; who wants to move to alaska w/ me?? :-D

another thing to look at is the noise (if any?) generated by the pump and the amount of energy it uses and heats the system with (are you really accomplishing anything if the pump adds lots of $$ to your bills in the long run and doesnt work well because the pump heats the water the same as the cpu?) lastly, will it "blow" up?? (thats a lot of pressure!)


however at that price i would still give it a go after considering the above (since its "free" you might as well try huh? and if it fails you already have most all of the bits you just have to change pumps and move some tubes)

[David Shh!]

Well, how did it go?

Given the pressures involved in moving 1 litre / second, you may want to include an expansion tank. Hotter water = more water, closed system = burst system.

If the car rads were kept in the dark and could be kept wet on the outside, they would lose heat more efficiently, as the external water would evaporate, taking away those nasty heat joules.

Alternatively, a few dirty great fans on the rads would keep it nice & icy, but noise + cost + running costs...

[Áedán]

Most of the central heating pumps I have come across are capable of multiple speeds - usually with a small switch on the side of the pump. If you can drop the flow rate, the static pressure will drop too, which helps to make it more manageable.

The biggest problem is that the high flow rate, and high pressure output means you could face difficulties with leaks and hoses coming lose. 5M of head is about 16.5PSI (1.1bar) worth of pressure. Most pond pumps don't manage much above about 2.6PSI, which is a much more manageable pressure.

In terms of heat, my central heating pump operates at about 40W worth of power. I suspect that most of this ends up as heat!

It may be possible to build a bypass loop around the pump to dump the excess flow back to the pump's intake. This would require balancing of the flows however. This removes the concern about the flowrate through the blocks

On the other hand, if you're using a full radiator, rather than just a heater matrix, you have a much higher cooling capacity, and the ability to handle fairly high flow rates.

                   Expansion 
                     Vessel
                        |
  -----------------------
  |          |          |
  |          |         Pump
Valve        |          |
  |        Bypass     -----
CPUBlock   Valve      |   |
  |          |       Rad Rad
  |          |        |   |
  |          |        -----
  |          |          |
  -----------------------

Perhaps something like the above? This isn't really a design, just an idea.

Áedán

[David Shh!]

...and if the CPU valve was thermally controlled, and the bypass valve was opened when CPU valve detected higher pressures...

[Áedán]

It'd be static state stuff, just to balance the flows! The bypass valve is just to ensure that excess is bled off, rather than force fed through the waterblock(s).

It also allows the pump/radiator combination to act as a form of reservoir, as the bulk of the coolant will be making multiple passes through the radiators, rather than actually cooling the CPU.

A couple of nice stop****s sould work here. (What are the odds that s t o p c o c k s gets filtered?)

Áedán