Do-it-yourself CPU Water Cooler

Foss writes "This article on EIMod.com shows a (very) cheap and effective way of getting that usually-expensive water cooling system that many of us have thought about. There are some pretty pictures too :)"

Re:Has anyone just stuck the board in the fridge?

[SGDarkKnight]

yeah, a few people have... these were the first to come to mind [totl.net] though.

Personal Experience

[quantax]

As someone who has built his own homemade watercooler, this setup is very jury-rigged and definately not recommended for any sort of serious long term cooling. The copper-tube inlets need to be replaced w/ brass fittings to avoid leakage, and the rubber bands should be replaced with cheap and much more effective band clamps. This waterblock design is not effective water-distrobution wise either as the water is not forced to flow through out the entire design. This would let the side furthest the inlets get hot due to poor water flow. An open chamber is only good for small waterblocks. Lastly, using thread to attach a waterblock to the cpu is ghetto as hell. Either quickly engineer a heatsink clamp yourself, or just look up a guide on the net for this, its pretty simple. Watercooling is one of those things that can be done many different ways, but this particular method is a little more 'amature' than is recommended to put on any piece of equipment that you value. If you are going to take the time to make your own watercooling, also take the time to make sure its engineered right.

Watercooling works great

[xtal]

I run a watercooled machine as my primary work box. It's great, and the noise savings were incredible. No more whirrrrrr. Fits snugly into a standard mid tower case.

I have a page up with all the details of contsruction [nyx.net] for you who are interested. I've been running it for a few months, 24/7, and there have been no problems whatsoever. I took a few additional precautions, but the system as been moved around several times without any difficulties whatsoever and I highly recommend it to others who are interested.

Re:Other liquids

[pkesel]

You'll need a much more expensive pump to push mineral oil, or about anything but water. The extra viscosity will likely cause the pump to overheat. And the seals and such may decay with other materials.

seems like a pretty complex way to make awaterbloc

[Hadlock]

my friend had me mill one for him for a science experiment. actually , ended up about 5 of them. he bought 2x2x1" aluminum blocks. i milled two holes through (one end to the other), side by side. he tapped them, and attached plumbing devices to the newly threaded areas. The other method was to bore four holes - two holes one one side that went 80% through, and 2 holes on the adjacent side that also went 80% through. tapped, and attached appropriate connectors. no leaks to patch. since the tops were flat, he also took the old heat sink + fan, removed the fan, and used that to dissipate evem more heat.

i never got the results back, but if anyone's interested, i can get the data to you, along with pics and more details.

with link now

[fabiolrs]

http://www.crazypc.com/articles/watercool.htm [crazypc.com]

Re:String and rubber bands?

[Grab]

Yeah, pretty grim.

For the benefit of that guy, if he's reading...

1) Most of your water will go straight in and straight out again without doing much cooling. A heatsink should force the water to go all the way round the heatsink - try adding some baffles inside the box to improve that.

2) Have you never heard of hose clips? 50c each and 100% reliable at clamping off flexible hoses without leakage.

3) Get some heatsink compound between the heatsink and the processor. Without it, a fair chunk of the heat is never even getting to the heatsink, so the whole point of having a nice efficient heatsink is wasted.

Grab.

Re:Watercooling works great

[xtal]

Heya.. the sediment only appeared when I did the first test run and drain, and it was obviously copper burrs (although there was some black stuff that wasn't identifyable - this could have been water wetter percipitate). I was concerned about the larger burrs because they look like they could have damaged the magnetic impeller it uses.

However, the box is on 24/7 since then and hasn't had so much as a hiccup :). I can highly recommend it as a upper-end mod for those who want silent running.

Thanks for the compliments on the page, too :-).

Re:What cools the pump, an underclocked '286?

[markmoss]

No, what he needs (and no doubt he is aware of it), is a better pump. Also proper tubing fittings, hose clamps, baffles inside the waterblock to direct the flow so all of it gets cooled, ears on the block to attach spring-clip heatsink holders, and so on. But the cheap 10-minute pump is good enough for checking out whether he's on the right track -- and he claims his budget was 77 pence, which I think is about $1 US. Aquarium pumps run continuously and aren't too expensive, but on that budget you use what you already have...

I do hope that he used thermal grease -- the article doesn't say one way or the other, but even a perfect heat sink can't cool well if the heat has trouble getting _to_ it.

One thing that did kind of bother me: "Although it seems that copper would be best suited for making a water block, I'm not entirely convinced without physical proof." It scares me that anyone who has to ask would be doing this! Copper is indeed the best material, unless you are on a NASA cost plus 10% contract, then use gold and increase your profit. ;-) Copper is resistant to corrosion and has the second highest heat transfer rate (by volume) of any material available in bulk. Gold is better and aluminum is worse on both counts. When aluminum is used instead of copper, they are trading off a little effectiveness for considerable savings in cost and weight; if you are using water, weight better not be an issue, and for a do-it-yourself project you'd spend a lot more trying to solder aluminum than copper will cost...

This story actually teaches a lot...

[jhiv]

This story actually teaches a lot about how not to build a water-cooling system for a CPU.

Some specific observations:

1. The low delta T (temperature difference) for the water going through the system is a sure indication of low efficiency. The most likely culprit is poor contact between the block and the CPU. Thin sheeting was used, and ripples are clearly visible in the pictures. The block is probably only touching in a few places and there is no mention of using thermal grease. A stiffer bottom plate was clearly called for.
2. As mentioned elsewhere, hose clamps should have been used. (String? Let's not go there...)
3. There is no radiator to dump the heat back to the environment. The heat transfer from the surface of the tank is probably not sufficient to keep the temperatures low. The radiator should be after the pump, to dump the heat from the pump also.

Perhaps the story should have been posted with the "laugh, it's funny" icon.

BTW, I prefer Indium foil as the thermal gasket between the CPU and heat sink, not thermal grease. Unfortunately, Indium is usually as expensive as gold.