There's apparently a "liquid cooling" snobbery subculture similar to audiophool subculture, that takes pride in spending as much money as possible on their "rigs" - you can't possibly hope to cool anything effectively unless you buy a $170 pump, pure virgin copper heatsinks with waveguide micro-channel internal fins for $80 per square inch, etc.
do you need that power hog for those tasks?
Via.tw makes fanless boards.
-- This email has not been checked by half-arsed antivirus software
why do you need more room? heat pipes don't need a reservoir. both systems need a radiator.
-- This email has not been checked by half-arsed antivirus software
Ya, it needs to be capable of doing real-time audio/graphics stuff where the software is Windows only, so needs to be x86, and a fanless Celeron or Atom won't cut it.
I'm going to take a stab at it with say a 3 core desktop processor slightly undervolted/underclocked, might be able to get away with th job that needs to be done using a low-profile heatsink and near-silent low RPM fan.
There's not a lot of money on offer (art, you see) but the project seemed fun, objective is to get it mostly done with stuff I have on hand already ;-)
You need more room because heat pipes transfer the heat by movement of a va por. The water cooled systems use a liquid. To get the same weight of coolant e xtracting the heat, you have to move more volume. The flow rate in a heat pipe is faster, but the tubes still tend to be much bigger. Also the tube s carrying the water can be flexible and be bent to fit. Heat pipes tend t o be rigid. There may be heat pipes that are flexible. but I am not aware of them.
Dan
vapor.
But the vapour is a lot less viscous than a liquid, and because you are mov ing heat as latent heat of vapourisation/condensation, you move a lot more heat per unit mass.
The vapour travels very rapidly through the heat pipes - I can still rememb er being impressed by it whistling through my vacuum line when I was doing vacuum distillation.
You do have to make sure that heat pipe properly evacuated when the heat tr ansfer fluid went in - heat pipes had a reputation for being a bit anemic a t low temperature differences, but if you test them at goods inward you can send back the one's with any uncondensable gas (as was done with the heat- pipes referred to in my 1996 paper).
extracting the heat, you have to move more volume.
Wrong. You are moving the heat as latent heat of vapourisation/condensation , so you need to move a lot less mass to shift the same amount of heat.
For water the latent heat of vapourisation is 2,230 joules (533 calories) p er gram. If you are shifting heat with circulating water, and can afford 10 degrees Kelvin difference between the device being cooled and the fins rad iating the heat, you move 10 calories per gram - rather less.
ch bigger.
Not that I've noticed.
at pipes tend to be rigid. There may be heat pipes that are flexible. but I am not aware of them.
Plumbers are good at bending copper pipes to fit, and copper pipes are most ly what is used for the vapour path in heat-pipe systems. Thick-walled bell ows can be used if you need real flexiblity. Make the pipe long enough, and copper pipes can accommodate quite a lot of movement. I've done it with gl ass tubes in vacuum systems too.
I was nervous of copper pipes coupling vibration in the 1996 work, but when we tried it the tubes were long enough to be floppy enough not to present a problem.
-- Bill Sloman, Sydney
The basic number is that 1 GPM has a theta of 0.004 K/W. That tells you the required gross water flow.
Here's a calculator:
It doesn't matter whether you run a given flow through one cold plate or many, or whether they are in series or parallel for the water flow, but if in series obviously the last one to get the water is the hottest.
In my case, I can have a small cold plate bolted somewhere on my machined baseplate, a small intense spot of cold, and my problem is to transport heat from all the hot electronic parts to that cold spot. The machined baseplate should be pretty much isothermal, so the serious thermal resistances will be local to individual parts. I care a lot about capacitance to ground for a few hot parts, so there's the compromise between cooling and capacitance.
It's a little harder to do the water flow math. Pressure drop goes up about as the square of flow rate, and pumps are nonlinear, so it isn't well modeled by electrical circuit analogies. I'm sure here are people do do this routinely, but not me. I'd just try it.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
A water system can use long runs of cheap plastic or metal hoses, cut to length as needed. And it can have various paths and branches from one central pump/radiator location.
Water piping can be connectorized too, so a box can be replaced easily in a system.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
That was not immediately obvious to me, but probably should have been. :-)
The large cold plates at Digi are pretty darn pricey, but I can get a good deal on a number of the smaller square aluminum block plates series-connected. The experiments people have with them on YT seem promising.
Thinking about buffing off the powder coat to a smoother surface and "tiling" them in a spiral pattern around the roof of the enclosure such that the last plate is centered in the middle, securing them with thermal epoxy. The materials are inexpensive enough that I can try something different with another enclosure if it doesn't work out.
If the CPU plate is going to be copper and the radiator sinks aluminum there's probably galvanic corrosion issues to worry about, maybe use de-ionised water or some type of glycol-based coolant fluid
I'd expect that you can conduct 65 watts from the top of the CPU to the top of the metal box, purely passive cooling. It would need an aluminum or copper heat spreader plate. Water is messy.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Seems like a heat pipe could do that.
My local salvage yard has heaps of them, $1.25/lb for aluminum, $3/lb if copper. Mostly from Dell desktops, ISTM.
Cheers, James Arthur
A heat pipe just moves heat. It still needs to go somewhere.
Sadly, heat pipes are mostly rigid. That negates a lot of their potential.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Right, 65W from the CPU to the box's panel. Making less heat in the first place would be a big help, too.
James
ote:
g dissipated to pump the coolant around.
ng dissipated can be used to circulate the water. I made a water cooled he at sink and used it with the reservoir higher than the heat sink. It did no t need any pump to circulate the water. I am not saying that heat pipes ar e not a good solution , but you do need more room in the computer case. An d water cooling is easy to implement at low cost if you have a reasonable a ssortment of tools.
e a=srch
)
So the radiator for this heat will be the top of your box? What's it made of? I had issues heat sinking to this steel case in a box. If you've got the box you might try sticking a power resistor on it and measuring how hot it gets at 65W. I've never used water, but it sounds messy, boiling would not be so good either.
George H.
The enclosure is going to be pretty packed as it is; routing rigid heat pipes up to the roof is going to be an exercise in topology as there are going to be components (disk drive, etc.) mounted on a riser on standoffs floating in space above the motherboard.
Aluminum micro-ITX enclosures seem to come in two general types, boxes that are a little too small for comfort, and boxes that are larger than the form-factor the customer wants. :|
Well...wait until I mention the part where I explain that furthermore I hope to horizontally mount a thin PCI-e expansion card out-of-socket on a riser plate just above the CPU and motherboard using a 16x - 1x cable leading to the single on-board slot...
Ya I've got some 50 watt power resistors around here somewhere, at least. Let's see if I can fry eggs on it
The box itself is made from around a kilo of anodized aluminum.
a vapor.
oving heat as latent heat of vapourisation/condensation, you move a lot mor e heat per unit mass.
mber being impressed by it whistling through my vacuum line when I was doin g vacuum distillation.
transfer fluid went in - heat pipes had a reputation for being a bit anemic at low temperature differences, but if you test them at goods inward you c an send back the one's with any uncondensable gas (as was done with the hea t-pipes referred to in my 1996 paper).
nt extracting the heat, you have to move more volume.
on, so you need to move a lot less mass to shift the same amount of heat.
per gram. If you are shifting heat with circulating water, and can afford
10 degrees Kelvin difference between the device being cooled and the fins r adiating the heat, you move 10 calories per gram - rather less.remember density of water is ~1600 times that of steam ...
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.