Fanless Heatpipe CPU Cooling System by FMAH

[mcgradys]

That idea rocks! I've been trying to find a way to do something like this for awhile. First, a question; will the heatpipe work effectively if it is bent like an "L", where the bottom is still the hot? Next, why not mount your power supply such that it will suck air over your heatsink. Maybe make some sort of tunnel to aid flow.

How much would the heat pipes be if you had to buy them (I am assuming they're engineering samples)? Can the heatpipes be flattened for use as the CPU contact surface? Perhaps a largfer pipe needed, but you get where I'm going.

Again, nice setup.

Sean

[dago]

IIRC my lectures (3-4+ years ago),

The use of 'black' for thermal properties is just an image of the 'black' colors, in terms of radiating/absorbing capabilities and isn't meant to be taken in the same way.


Less sure, but, I think there may be != type of painting and that you have special (appropriate) paitings, the same than for heating systems.

In this case, you can even paint your case in yellow or pink, it'll be the same !

(I may try to find back references later ....)

[Bean]

Thanks for putting me back in reality Mike. (In hindsight, I'd sure like to edit that post LOL)

would it be a good idea to insulate the hot baseplate to prevent any loss to the surroundings? I know some will prob say if thats true it would have been implemented already. But since your experimenting anyway it may be easy to do.

[beanjamman]

Would heatpipes even be necessary if the cpu were mounted on the underside of the motherboard? That way the cpu could be in direct contact with the massive heatsink (which would be sort of a combination motherboard-tray/side-of-computer-case/heatsink). I've been wondering for a while why none of the small form factor manufacturers that also produce motherboards (Shuttle, Biostar) have done that - it seems like a simpler, cheaper, more elegant solution than using heatpipes. Of course, the heatpipe/huge-heat-sink-outside-the-case solution is way better than using a (relatively) small heatsink with a fan inside a hot case and doesn't require the fabrication of a new type of motherboard...

[McF]

that sounds like a great idea indeed! I wonder if it would be doable to resolder the cpu socket on the other side... would have to use a lot of wires though (370 or more?)

[fmah]

....will the heatpipe work effectively if it is bent like an "L", where the bottom is still the hot? Next, why not mount your power supply such that it will suck air over your heatsink. Maybe make some sort of tunnel to aid flow.

I think the fewer bends the better. This makes it easier for the liquid and gas to flow back a forth. In a case the power supply will probably be in the typical position so it may draw some heat out from above the copper blocks.

How much would the heat pipes be if you had to buy them (I am assuming they're engineering samples)? Can the heatpipes be flattened for use as the CPU contact surface?

I'm not sure about cost, and you really can't flatten them, at least not without making it hard for the fluid and vapor to flow in that part of the pipe.

[fmah]

Thanks for putting me back in reality Mike. (In hindsight, I'd sure like to edit that post LOL)

would it be a good idea to insulate the hot baseplate to prevent any loss to the surroundings?

Well I'm not sure, but for now the copper blocks are being cooled a little by the surrounding air, which is fine. A little too much complexity to insulate them.

[fmah]

Would heatpipes even be necessary if the cpu were mounted on the underside of the motherboard? That way the cpu could be in direct contact with the massive heatsink (which would be sort of a combination motherboard-tray/side-of-computer-case/heatsink).

This is typical of audio power modules, the hot parts are located to mount to the heatsink. So basically the circuit board mounts to the heatsink, instead of the heatsink mounting to the circuit board, so to speak. You couldn't really do this to a normal motherboard since flipping it over would have the pins in the wrong place. You could put in a rewiring, but that would probably affect the signal timing.

[MikeC]

Would heatpipes even be necessary if the cpu were mounted on the underside of the motherboard? That way the cpu could be in direct contact with the massive heatsink (which would be sort of a combination motherboard-tray/side-of-computer-case/heatsink).

This is typical of audio power modules, the hot parts are located to mount to the heatsink. So basically the circuit board mounts to the heatsink, instead of the heatsink mounting to the circuit board, so to speak. You couldn't really do this to a normal motherboard since flipping it over would have the pins in the wrong place.

The retooling cost for mobo makers is probably prohibitive unless fairly substantial demand was assured. Also, unlike in audio power amps where power transistor casings are quite robust, CPUs in general are fairly delicate, so you couldn't just clamp it direcrictly to the HS, I would think, not at the risk of damage due to thermal cycling -- exnapnsion/contraction of the metals, etc.

[Marvin]

I'm not sure about cost, and you really can't flatten them, at least not without making it hard for the fluid and vapor to flow in that part of the pipe.

I found out that there are allready flat heatpipes made, here. Dont know anything else about it but it should be possible to use those in direct contact to CPU, if one would make adequate mounting gear to those. Different questions are, how can they be bent etc.

But imagine a solution where you would have a heatpipe(s) in direct contact with CPU, would have to be sure it is enough to cool those monsters...


EDIT: fixed the quote

[puff]

Nice design man, interesting how you got your heatpipes. I've been designing a heatpipe HD enclosure for awhile now, prototyping began a couple weeks ago but the only way of getting 5mm heatpipes was buying that zalman HD cooler, they are a total pain to rebend correctly (need straight pipes).

Ive been thinking of adapting a large radiator on the right side of a large ATX case and running heatpipes to it, but finally I decided to spend my time on an HD enclosure (HD is pretty much the last remaining source of noise, as the SD2002 is whoefully inefficient for already quiet drives imho).

I've also been thinking of bolting an aluminum 40cm*50cm radiator used for water-based heater setups (picture on the page not accurate, it's squarer) on the right side of the case, and using it as a passive radiator for watercooling. Looks good, and it should have huge dissipation. I'd need to get adapters for the tubing, though.

I'm not sure about cost, and you really can't flatten them, at least not without making it hard for the fluid and vapor to flow in that part of the pipe.

A flat heatpipe (5 mm diameter flattened to 4mm height/~6mm width) would somewhat help me reduce accoustic bridging, and from my highly scientific test of using a heatgun against a heatplate on one end of the pipe seems to work. I finally decided against running the pipes out of spec, but HDs dissipate quite a lot less heat than CPUs so I could've probably gotten away with flatter pipes.

[zenyatta]

I hate to be a nitpick but doesn't the liquid in the heatpipes make noise when it boils? Mind you, nice warm bubbling beats fan noise anytime

z.

[puff]

I hate to be a nitpick but doesn't the liquid in the heatpipes make noise when it boils? Mind you, nice warm bubbling beats fan noise anytime

z.

No, it doesn't. It evaporates more than boils, there's not that much liquid and the temp isn't that high.

A heatpipe doesn't make noise at all and practically doesn't conduct it. The mesh/liquid inside make it a poor resonator, try hitting one with a screwdriver.

[McF]

so where do you guys get those heatpipes from? Any chance to buy some in europe?

[Semm]

There's a place in Austria that does 'em...I don't remember what site it was, but some guy built a heatpipe-cooled computer in Germany, and he mentioned it. I think it was teschke.de. Might email the guy.

Semm

[Marvin]

I sent that link up there somewhere that is a German firm, http://www.alutronic.de/e/indexe.htm. They have sales rep in Holland...

[McF]

thanks! it looks promising. they got heatpipes in different sizes, even nice 2mmx5 mm flat ones - could be used to attach to a standar cpu heatsink with a bottom plate thick enough

[pangit]

Would heatpipes even be necessary if the cpu were mounted on the underside of the motherboard?

Another thing you could do, which I think Numano3 tried, is to try to get rid of some of the heat through the bottom of the CPU/Mobo. This requires several layers of thermally conductive/electrically insulated gel sheets, to connect the underneath of the CPU to the big heatsink. And would require transferring heat through the (not very conductive mobo), but at the least it may help a little bit, rather than being a single solution on its own.

I've kind of been thinking about doing an experiment like that myself, by cutting a square hole in my mobo mounting plate directly under the CPU, taking off the side panel and mounting a spare HS directly on the underside of the mobo (with a layer of gel sheet between). As well as having a few layers of gel sheet between the CPU and mobo. It should help a bit but might look rather ugly.

[qnliu]

Has anyone heard of a heatpipe-based cooling system where the heat is dissipated by a heatsink/radiator mounted on a case fan (or a power supply fan)? A system like that would be much lighter and smaller than fmah's purely passive system, but have similar performance. It would be very quiet as long as the case fan is a large low-RPM fan like the one on the sonata.

I used to have a water-cooled system like that, its overclocking performance was great, but eventually I gave it up for quiet (Zalman) air-cooling because it made the case very hard to work with, required bi-annual fluid fillups and after running it for two years straight I just didn't trust it to last forever Replacing the water-based heat transfer with heatpipe-based seems just perfect: maintenance-free, zero noise (no pump), minimal space occupied, high performance because the CPU heat is dumped directly out of the case.

[Rusty075]

Has anyone heard of a heatpipe-based cooling system where the heat is dissipated by a heatsink/radiator mounted on a case fan (or a power supply fan)?

Yes, all the Shuttle's use that sort of an arrangement.

[qnliu]

Hmm, nice. I hope someone will make a generic system with longer heatpipes for other cases (pretty sure those are too short for the sonata).

[Rusty075]

Thr problem is that there can't be a generic system like that, it has to be case-specific.

[cogman]

One thing I would like to point out, If you eliminated a lot of the copper, or just went to the pure heat pipe, the heat transfer would be a ton better, though copper transfers heat well, it is no heat pipe (which I believe transfers at a rate like 20x that of copper) and really is just a layer of insulation the way you have it set up. I Like the design but I think that that is a major thing that could improve the ability of it to disapate heat 100x better. maybe melt the copper down and replace the aluminium sink with your own crafted copper heatsink .

[fmah]

Thanks. I'm investigating various things. I will probably make the base thickness around 1/8" or 1/16". The main issue, from my further testing, seems to indicate that the cooling is better if the "cool" side can dissipate more heat. The heatpipes can pull pretty well, but the other side needs to release the heat better. Longer fins, black anodizing, and spreading the heat out should be useful.

[ruprag]

Well since heatpipes work on temperature differentials then it makes sense to have as high a difference between the end points => better cooling for the "cool" side.

Did you try hlowing a fan over it to see if there was a big temperature difference ? (it is a cheap way of checking if the bottleneck is there)

[fmah]

Yeah, I blew a fan over the large heatsink. The temp went down 5°C. My conclusion is that 2 pipes seems to be sufficient for Athlon XP chips.

[cogman]

Another thing, (refering to the shuttle heat sink design) I think that you could increase cooling if you place a heat sink on top of heatpipes over the cpu. So it would be Heatsink, Heatpipe, Cpu. I really would like to see this work out to run at higher speeds. I guess that 2 heat pipes may be sufficiant, but why not overkill

[fmah]

More pipes may be in the works, but it still appears to be limited by the cooling on the other side (the large heatsink). The cooling difference between 1 and 2 pipes was 3°C. I would suspect the temperature decrease diminishes with additional pipes (i.e. a 3rd pipe would likely result in less than a 3°C decrease, probably 1-2°C), unless they can spread the heat over the larger heatsink better.

I want to remove the heat from the case area, because a temperature controlled power supply fan will speed up if the heat is dumped into the case area. That would negate the benefits of pulling the heat out through the side. Also, a passive power supply would possibly overheat if more of the heat rises into it from the CPU. Anyway, it's not too difficult to mount a heatsink to the top of the block with thermal epoxy or something, if someone wanted to.

[JustinCase]

So, has anyone here been able to buy heatpipes anywhere? I've been looking for a place that sells them in europe, but couldn't find anything.
Any ideas?

[prodeous]

Another thing you might try is to seperate the heatpipes on the heatsink side. This way it would give more cooling area per each heatpipe. Currently in your setup, you are sharing the same copper blocks and region of the heatsink for both heatpipes. Just a thought.