fmah's design was first shown in these forum pages, but now it is a full-fledged article with all the details. It's in the Cooling section but could also go into Systems. A very cool design.

Very very nice job. I really want one if they go up for sale.

That article represents the future of silencing high performance computers. Hands down.


How's that for a grandiouse statement?


Given enough time I think it's "read" count will pass the Breadbox (and my Desk ) to be the most read article on SPCR.

As a mechanical engineer friend of mine (I'm aero) used to say, with tongue only partially in cheek, "that's the kind of article real engineers like".

Many thanks.

This mechanical engineer says two thumbs up

Yeah, great job, now we just need aluminum case manufacturers to catch on and start making their cases into passive heatsink assemblies..

Thanks everyone for the comments.

I did some quick calculations and determined I could get maybe up to 10W more heat dissipation out of the large heatsink by anodizing it. If I get around to it I will try that and check the results.

I'm working on a silent/cool hard drive box next, and then the rest of the setup.

Not sure if this is something that would help as I have no idea how these work but here goes. Would the large heatsink work better if it was all copper as well?

If all other things are equal, yes. Copper transfers heat more efficiently. But it would be pretty expensive & difficult to get, unlike large aluminum HS extrusions, which are easily & widely available.

A better way to improve the cooling engine might be to use a larger number of heatpipes or heatpipes of precisely optimized diameter / length for most efficient heat transfer between CPU & HS.

If you don't know how heatpipes work, do explore that link in the article to the Thermacore site -- heatpipes are already in the mainstream, and if Rusty's prediction is to be believed, this article might just help to push them into the middle of the mainstream for CPU cooling by majors.

What if the heatsink was on the top of the case? Like in a desktop setup? The heat would flow better upwards, right?

Copper is better, but it turns green after awhile... (b/c of oxygen) I dunno if spraying it with an antioxidizer would decrease some of its conductivity or not. May still be better than aluminum even w/spray. May be more expensive though. Also, brushed aluminum looks really nice.

EDIT: Didn't see previous post, didn't see that you'd already answered the question.

How much better do you suppose a 6 heatpipe cooler would work?

Great article.

The higher figure (71.6W) is probably the more correct; the 2400+ Athlon XP has a slightly different core, IIRC.

As for the a copper heatsink, I think I would first experiment with laminating a copper sheet on the back of it (thermal epoxy or just thermal goo and squeeze). The important is to get as much of the heatsink as hot as possible.

But passive radiators are complex, it's all about getting the air to move...
So I wouldn't bet on that it would help much. Anyway, 44°C should be cool enough for most needs, so why bother?

I would go for this design too, but I think I'll stick with (passive or semipassive) watercooling for now... At least until I get my own metal workshop.

great article indeed!
things I wonder about:
- painting the aluminum heatsink black or even covering it in a very thin (sprayed) layer of black paint should result in a higher heat transfer. even making it rough (with sandpaper maybe) should have a positive effect on the amount of the heat radiated off the heatsink.
- what is the heat transfer rate of heatpipes compared to solid copper pipe with the same diameter and length? I wonder if substituting the heatpipes with a solid copper element shaped correctly could mean better heat transfer? It's a design for this particular motherboard anyway, so there is no way you can use the cpu heatsink in any other place without bending the heatpipes or anything similar anyway. I'm just curious..
-since the copper parts already are used to cool the processor a bit as well, I'd suggest to roughen their external surface (greater effective surface for the convection heat transfer) and painting them black (for the radiation transfer component). Though these steps might not do much good if the setup was to be used in a closed computer case.

Anyway - great design and execution!

Something new to learn!
McF: would you mind educating me as to why painting it black would help? I would think that would only lead to less heat transfer, as there would be another barrier to it, but it's not something I know a lot about.

Great article, fmah. I was thinking of doing something similar, and it's good to know it works!

Thanks
Semm

[edited to add kudos]

I'm pretty sure one can improve heat transfer by adding more heatpipe. I believe two more would work will for what is out there now. Also Fmah was right about anodizing the Al heatsink. I read an article a while back that anodizing Al improves heat radiation.

black surfaces have a higher radiation coefficient (alfa I believe) then shiny/reflective serfaces. the perfect "black" radiation body has a coefficient of 1 (it will absord 100% of energy radiated onto it), while a polished (mirrorlike) metal surface can have alfa of 0,1. black surfaces absorb heat (transfered as radiation energy) faster also, but that's not a problem here, since the heatsink will always be hotter then the surrounding.
As for the paint layer acting as insulator - thats correct, however, there is also a slight increase in effective surface (the thickness of the paint makes the heatsink slightly bigger) and the effect of increased radiation of heat to the surroundings, and those two combined can have an overall positive effect - so more heat is lost then with a shiny pure aluminum heatsink.
This is why for example thin copper wires transferring high current (Ampere, not high voltage) if "insulated" with a layer of plastic, are less hot then bare wires! There is however an optimum here between increase in surface area, and increase of heat transfer resistance (because of the insulation). Same goes for the insulation of pipes with hot water in cold surroundings - if too thick layer of insulation is placed, more heat can be lost actually

Of course it is important to place the pc in a surroundings that has no surfaces hotter then the heatsink - because those radiate heat as well, and this heat will be absorbed by the heatsink of the ps. So dont place the pc in front of the radiator, or even in straight line "of sight" of a radiator. better to have a barrier between those two. This barrier better not be reflective either (polished metal for example) or the heat radiated from the heatsink will be reflected back So an insulator is best - like a furniture from wood or plastic . Also keep the pc away from the window if the sun shines (same reasons here) as window glass is transparent for infrared energy.

Trip - about placing the heatsink on top of the case - it would actually be worse I believe - I'd have to check the transfer rates/resistances for horizontal surfaces for natural convection. It has to do with the fact that the air has to change flow direction - hot air rising above the heatsink, and cold air is pulled from sides:
something like this:

^
|
|
\----- <-

I hope you get my point, Im no master with ASCI graphics

while with the vertical setup like in this design here, the air heats up along the way and rises along the heatsink fins naturally without any obstruction and without any energy needed for flow direction change.
I'd have to check the books for precize ezplanation and numbers though.

McF- Interesting, thanks. I knew black absorbed radiation nicely, didn't know it radiated better too. But I should have: I walk on the stripes in an asphalt parking lot in summer And considering they're in contact with the asphalt, there isn't going to be a temp difference, just a thermal conductivity one.

Thanks again,
Semm

Not sure, but there is probably not room to easily place 6 for the socket 462, maybe 4. For the Zalman design, the Pentium 4 has mounting holes that are futher apart from the socket than the AMD. So there is space for 6 pipes, but they are basically next to each other.

It seems that a lot of the result depends on the area over which the heat is spread over the large heatsink. In fact, it might be interesting to see what the result would be if the cool end were in a large water bath.

Earlier in the year I had posted about using solid copper for the conducting element, but that would be too heavy, expensive, and large.

Thats a neat setup. Reeks of quality.

fmah, McF: just curious about this anodizing or painting black would this work for copper in your opinion. Or do you have a better suggestion for preventing oxidation on copper w/o negating the heat transfer properties.

Thinking on this heatpipe. fmah said the block on the cpu was very hot. Would it be ideal if it wasnt? (not knocking your design its impeccable. just theorizing) In other words the transfer efficiency would be better if it was warm and the sink was hot?

Tha last sentence: Impossible to achieve, I am afraid, for obvious reasons. If you did achieve it, the world's top scientists would come knocking at your door seeking your benediction.

The other comment by McF about whether this would work better with a solid piece of metal, the short answer is no. If it was that simple, that type of design would be everywhere already. Heatpipes are much more efficient in transferring heat from one place to another. Evaporation away from the heat, condensation to bring the cooled liquid back, cycling back and forth in the tube continuously, that's basically how they work. Check the Thermacore page linked in the article! http://www.thermacore.com/hpt_how.htm

Here's an illustrative graph from Thermacore showing heat transfer performance between several high heat flux heat pipes as well as an equivalent size copper block. The test articles have outer dimensions of 1" wide, 5" long and 0.2" thick:
http://www.thermacore.com/heat_flux.htm

Thanks Mike, I suspected as much, but its good to know for sure I know how the heatpipes work, just wasnt sure if they're better or worse then pure copper.

Bean - do you have any info as to the heat transfer properties of Copper oxide? if it's not much different, I wouldnt worry about it. same goes for the aluminum heatsink.
And since the heatsink is the main source of heat transfer here, the copper parts can be protected from corrosion without significant side effects, right?
if painting I'd check the properties of the paint, and how good it can withstand high temperatures like these though.

As for putting the other ends of the heatpipes in a tank filled with water - water has a lower heat transfer rate then most metals I believe, but if you'd let it evaporate (open tank) the temps might improve. Not sure if I'd like that in my room though.... it might become the new "steam pc"

I think that this was discussed in an earlier thread, but what about adding some heat sinks on top of the blocks? As they are hot, so extra HSs could drop the temps bit. Or would the lower delta T cause the heat transfer to be less efficient (to the big HS, that is)? Thus nulling the effect of added HSs?

Cheers,

Jan

The heat transfered to the large sink would be lessened and overall temperature would be reduced.

McF, from what I've heard radiation accounts for very little in the removal of heat when convection is also used, so black paint would not be likely to help. It would probably insulate. Black anodization however does not hurt convection, and so it may be a viable option.

I think this is where optimization of the system by trained engineers who have experience with heatpipes becomes valuable. It's clear that the worst bottleneck can be at a number of points -- either the evaporation or condensation copper blocks, the heatpipes or the heatsink. An optimized system would keep the heat transfer losses ~equal for all of these points for a given heat source and physical system. From a cost as well as an efficiency PoV, maximizing the heat transfer in one portion of the system without also doing so in the rest doesn't make good sense. At this point, I don't have much of a handle on how one would go about doing such optimized design. It is also easy to see why there are no "generic" "plug-n-play" CPU heatpipe cooling solutions out there. At least not yet.

I'm no ace in chemistry, but isnt andization juast a way to apply a thin layer of protective (black for example) paint through submerging the metal part in some liquid (containing the paint) and using the metal part as anode? Or is it not paint - if not - what is it?

Mike - youre right about the lack of standard available heatpipe systems and the optimalization story. what also is a problem here is the ATX standard - there isnt any room to put a heatsink big enough to cool the cpu naturally (no forced convection), and with the lack of standard on CPU socket placement relative to any part of the ATX case, its virtually impossible to design a system that can adapt to all different socket placements, AND will fit into any ATX case.
With all air cooling HS/Fan solutions, there is no need to worry about those things, all they are concerned about is the weight limit (though the manufacturers seem to forget this a lot lately) and the maximum height of the HS/Fan combo.
The actual air flow is of course dependant on the placement of other fans and parts in the case, and since that is different for almost every pc, the results vary a lot. So if someone tests the heatsink on a mobo in some huge case, or in a small one, or outside of the case, they all get different results, and then they compute the c/w ratios and compare hs/fan combos..
So to overcome the differences, manufactirers just put a fan that can pump as much air through the heatsink as possible, and hope for the best. hardly optimized either if you ask me

Just to add a bit, this article was linked on front page at [H]ardOCP this afternoon, 11th edition for Tuesday.

That will help make the article go ballistic -- like Russ predicted. The # of reads jumped by several thousand to ~5000 in the last couple hours. I can tell the server is starting to really cook... Let's hope it's up to this test. We haven't had a big hit in a while. Looks like it's getting at least 1 new read per second right now.

Anodizing is a process to create a aluminum oxide layer on the surface to help prevent degredation of the surface, at the same time you can add dye to color the aluminum (black, red, blue, etc...). Here is a science project at some high school that explains stuff. Making the surface a black anodize will help the radiative properties by increasing the emissivity of the surface. Higher emissivity means greater heat will transfer off of the surface.

http://www.howe.k12.ok.us/~jimaskew/chmlab28.htm

I will probably test a couple more things.
1) turn rig sideways to put heatsink and fins facing up
2) Remove 1 pipe and test with just 1 pipe in the system (perhaps it makes only a little difference, who knows)

Then I will try to get the thing anodized and do more testing.

I don't think I'll try to put fins or a heatsink on the copper block. Not yet anyways. I want to move the heat outside of the motherboard area.

Never mind, ^ he beat me to it.