Slashdot Mirror
Water Cooled Power Supply
lmd writes "Digital-Explosion has an article with step-by-step instructions on how to cool a power supply with water (yes, water) instead of fans/heatsinks to make it quieter. Please read the warning and disclaimer (and buy insurance if you don't have any) if you decide to try this at home."
I'm holding out for a liquid sodium cooled computer, just like valves on decent cars.
Well I guess a site called Digital Explosion is really the best suited to report this.
Here's a thought I had, but probably will never get around to building.
Lots of people go to the expense and effort of building/buying radiators or using large tanks of water as the heatsink for their water-based CPU cooler systems.
Last year, I started measuring the temperature of the water in my toilet tank. After a flush, it drops to 5-6 degrees Celsius. Between flushes, it gradually reaches room temperature, of course, but this is still no worse than a radiator or bucket. In practice, however, it never actually gets above about 10C (while room temperature is about 20C).
In other words, it's a supply of cold water which you were going to simply flush away.
Place a small bucket inside the toilet tank. Put a submersible pump in there, run the water to the CPU coolers, bring the water back and drain it over the bucket in the tank.
Everytime you flush the 6 beers you went through while flaming me for my Linux isn't ready for the desktop article, you can rest assured that the water which cools your CPU is being replaced with fresh, cold water. No mold, no mildew.
The purpose of putting the pump in the bucket is so that there's always a supply of water for the pump, even during the flush. And the purpose of draining the return line over the bucket is so that if your toilet tank doesn't refill for some reason, you'll still keep your bucket full of water and buy some time for hardware monitors to shut the system down if it's getting too warm.
I don't know how hot the water in the toilet will get, but think about this:
Of course, the only thing I'd worry about is the quality of the submersible pump. After all, if water leaked into the pump, then the water in the toilet could come into contact with one side of the AC line... the other side of which is grounded to your fusebox. If you happened to touch another grounded object while urinating (concrete floor, sink faucet, etc), then enough current could find that your stream of urine and urethral tissues are a more attractive ground path than the plastic sewer pipe. I think I'd invest in an isolation transformer (search ebay) to reduce the risk of highly ...unpleasant... damage.
I think if one were pumping water through tubes soldered to the heatsinks of their power supply, the risks would be compounded, conceivably by a failure on the primary side of the power supply: I think I'd make a point of running the computer on an isolation transformer as well.
Ahh... the joys of being an eccentric genius.
Fire and Meat. Yummy.
Here's another site that discusses water cooling your system.
--sexygal
Very popular slashdot journal for adul
9 minutes later, and it's Slashdotted already.
I guess those water cooled things *still* don't hold up.
Some web server somewhere has probably just evaporated in a cloud of steam.
WARNING : All power supplies have capacitors in them. These components can hold an electrical charge for days even weeks. We do not recommend that anyone opens up their Power Supply unless they do know what they are doing and are willing to take responsibility for their actions.
:D While that little sticker's intact, you can go and whine at your supplier if/when it goes bang. In any case, I'm not afraid so lets get stuck in :
:) Look carefully at the bottom of the board and then double-check the top surface. You should be able to work out which joints on the base correspond to the components on the other side. All the components have three legs which makes life a little easier. I've found quite often that the heatsinks have an additional soldered connection or two to help keep them attatched to the board. Here's a couple of pics of the components and the base of the board :
:D
:/ Having soldered the above into the PSU, I quickly re-assembled the thing and here's the results (photos taken just before I replaced the outer casing) :
:D Wow, this things so goddamn quiet - I love it! There's now only the two panaflows at the front running and they're at 5v each so you can barely hear them. With the disks encased in foam, even when there's hard disk activity, my machine's still damn quiet!
DISCLAIMER : The author of this article and the owner of this page are not responsible or liable for any damage caused to any equipment or persons. In attempting what is detailed below you are taking full responsility for your actions.
A Brief Introduction
When I went about water-cooling my first PSU, I was learning as I went along so now I've had the experience, I'm in a better position to do a decent job. In this article I'll go through, step-by-step, showing you how to water-cool your PSU from scratch! I started off with a nice little QTec 550W PSU :
If you've read the first article I wrote on water-cooling your PSU (which was aimed more at inspiring people than being a step-by-step guide), you'll know that my basic plan is quite simple. Basically, it involves replacing the fans / heatsinks with plates of copper. Each plate has a copper pipe soldered to it which is where the water runs to remove the heat. We'll get to that later - for now lets look at taking this thing apart.
Disassembly
If you have any doubts, this is the time to think again. As with just about every mod on this site, the first thing to do is void your warranty
Four tiny screws later and the top should be loose. If you're doing this to a different PSU, you may find there's a fifth screw near the base - there was one on my old AOpen PSU. Here it is, guts exposed :
Now the whole point of water-cooling the thing was to make it quieter so lets go ahead and get rid of those pesky fans :
I found that Q-Tec had been very helpful and given each fan a little connector that can easily be removed. Here's what you should have after removing the noisy beasts :
Noisy though the fans are, your PSU isn't going to work for very long without some kind of cooling. At this point you should be able to see the two heatsinks we're going to replace. Attatched to the sinks, you'll see rows of components - these are the really hot bits in your PSU and it's these that we'll be cooling. Now if we're going to replace the heatsinks with our water-cooled plates we need access to said components. There's two steps to this. First we need to remove the four screws that hold the main board of the PSU :
Next we need to remove the little board that attatches to the connector where you plug your PSU into the wall. If you don't do this, you'll have to bend the back of the PSU to get the board out!
Okay, all being well, you should have a fully disassembled unit :
The next step is a little more tricky. We need to get those heatsinks off those components but unfortunately, you won't be able to get to the screws that hold them on. So what do we do? Un-solder them of course
Right, lets get the first heatsink off :
With a little more de-soldering, here's the second one removed as well :
Removing the components and attatching them to the new water-cooled plates is a doddle. Just remember two points when doing this :
1) Whatever you do, DON'T FORGET what order the components went in - it could be disastrous if you got them mixed up!
2) Be careful when re-attatching, not to leave out the Mica shims (the grey pads). These stop you getting mains voltages going through the heatsink or water-block so they're pretty goddamn important!
Here you can see them attatched to the water-block I made :
If you're wondering how to get the holes on your block in just the right place, do what I did and use the heatsink you took off earlier as a template
Re-assembling the beast
The next step is to re-attatch the components to the board :
Now I run an XP in my machine and I have no intention of moving to Intel so the P4 connector's just taking up space in my machine. The same is true of the old ATX connector so I got rid of 'em :
Next it's time to implement a very handy bit of kit which makes water-cooling a little safer and easier. When you turn on your machine you don't want to have to remember to turn your pump on - if you forget, your liable to burn your chip! So what can you do to get around this? The answers simple - a 12v relay. Basically, when the computer starts, the 12v line coming out of the PC goes from 0 to 12v which closes the relay, starting the pump. I also find it useful to have an overide switch so you can pulse the pump on and off (to get rid of any trapped air in the system). Here's a quick diagram of the way my circuit works :
Apologies for my poor photochop skillz
And finally, here you can see it installed as I wait for the system to bleed :
Time for some tea and biccies! Well, I tentatively flicked the switch and as I cringed, waiting for a loud bang followed by fireworks, my machine quietly booted
That is just what I want, an H20 source pumping thur a 120v electrical device.
That is one hell of a failure mode.
People, this is not something to play around with. The disclaimers and warning on the site fall far short. It's one thing to cut yourself while monkeying around with a case mod; it's a whole 'nother country when you get a nasty zap, or worse.
Computer virus zaps oil giant
The first thing that came into mind when i saw "Water Cooled Power Supply", was "Darwin Award"...
It'll probably end up there somewhere in the coming months, now that this has been on slashdot.
Makes me wonder if desktops still have huge transformers at 50Hz instead of the modern switching type. We do live in the 2000s, the space age once dreamed of, you know. I fancy getting a mini-itx system some day, but only if I could use a laptop style, totally quiet PSU. I mean, PSUs are supposed to convert energy, not dissipate it, or what?
Then again, fans are not that bad compared to the sound from IBM hard drives...
Escher was the first MC and Giger invented the HR department.
As I recall, back in the day, the early Cray Super Computers was water-cooled
--- have you healed your church website?
CGI-limits reached, please try again later!
:-)
That site needs a mirror, and I need karma, here is a mirror. Be nice to it
...because a project involving the real danger of electrocution and fire hazard will certainly separate the real engineers from the "network/information systems" engineers!
Speaking from the perspective of an Electrical Engineering student, I must say this is totally insane. First of all, taking apart a power supply is bad enough if you're not careful. Second, pumping a conductive fluid through a high-voltage power converter is even more insane. Pumping water onto a processor to keep it cool is one thing, because currents and voltages on the motherboard are relatively low but a power supply has potentially lethal current and voltages. The article makes no mention of using a non-conductive fluid, nor does it make any mention of how to even correctly discharge capacitors while working with them. Anybody who tries this is seriously risking his or her life.
--
Adobe's anti-counterfeiting softw
Well I guess a site called Digital Explosion is really the best suited to report this.
Yeah, I have a couple of problems with the way this was carried out. Conceptually, I would love a completely water-cooled computer since I'm tired of the noise. But this is pretty dangerous.
Why remove the existing heat sinks? Rather than removing them from components and risking forgetting a mica insulator or doing other damage, why not simply take advantage of them as an easy surface to which to attach cooling tubes. Most power supplies I've opened, I could solder copper tubing to the heatsinks fairly easily.
The other thing is that the mass of the heatsinks would provide a little thermal inertia to buy you some time in the event of a bubble or other failure.
I've also got concerns about the overall safety of this. Even without mica insulators or any other outward signs, a heatsink may be running at some potential other than ground. Pure water isn't very conductive, but all the same, your cooling water is likely to be grounded - and should be grounded. Pumping water through a tube attached to a component or heatsink will bring the water to that potential; using a piece of plastic tubing to insulate one metal tube from another is NOT safe.
What you need to do is have electrically insulating but thermally conductive means to couple the heat to the tubing. Mica insulators and thermal transfer grease are a good start.
I think I'd solder some copper tubing to some copper sheetmetal, and then I'd coat the flat surface with heat transfer grease, add a sheet of mica and more transfer grease, and then screw it to a heatsink inside the power supply. I'd use off-the-shelf electronics hardware to screw the two pieces together but maintain their electrical isolation: even Radio Shack sells the stuff.
Make sure that the water is grounded, and then run the power supply from a Ground-Fault Interruptor (GFI) receptacle like you'd find in a bathroom. This way, a water leak in the power supply should turn off the power at the outlet and reduce the risk of a bigger problem.
Fire and Meat. Yummy.
*BrrrZZZaaaAAAAAaaaaaaaP*
It'll be really cool after it shorts.
OK, so decreasing the temperature of your PSU with water is... uhhmmm, cool.
But, I want to see some one come up with a ups enhanced power supply. At least in my experience most power supply problems have to be the little brown outs and black outs during storms. They generally last between momentary to 5 seconds. A PSU with 30 seconds of reserve energy at its rating (example 400W*30sec.) would be something worth having.
-- The morphemes of your disquisition are ascertainable, but they have eschewed an ambit of transpicuous exposition.
...is because all of the discrete components need cooling, not just the ones that have heat sinks. It might be as much as a year before he smells the delicate aroma of cooked dielectric when a capacitor overheats and explodes.
Karma
PC Power Supplies tend to be very inefficient, where efficiency is defined as
Power Output
% Efficiency = ----------------
Power Consumed
I wonder how much power we could save as a nation if we had higher efficiency power supplies in our PC.
Why in the world not use some liquid that DOESN'T conduct electricity?
pr0n - keeping monitor glass spotless since 1981.
It'll be really cool after it shorts
And after the fire goes out...
I pledge allegiance to the flag...
of the Corporate States of America...
I'm not too sure liquid sodium is the best choice to cool your computer, since sodium melts at 208 degrees F (98 C). Besides, when you first boot up the computer, you'd have to have special heaters installed just to melt the sodium and get it moving! But, you know, in the end, I know you were being facetious. Nice job.
I want to run mercury through my cooling system. I've got a couple of pounds of it, and it would certainly absorb heat more readily than water.
But just one drop of mercury inside your computer and it's finished.
Maybe could use gallium with small heaters?
Fire and Meat. Yummy.
http://xerithane.nerdfarm.org/watercool_psu.html
Or:
Here
Dacels Jewelers can't be trusted.
As I understand it, after installation and patching any initially detected leaks (which you should do a thorough test for immediately after installation), a water cooled system is very unlikely to develop any further leaks anyways unless you did a shoddy assembly/patch job.
Of course, this is jist what I've gleaned from my own experience
File under 'M' for 'Manic ranting'
Wonderfully funny and all that, but don't you think twice in one week is a little overkill?
Ok, I can't read the site because it is slashdotted, and yeah, you shouldn't trust these guys anyway, because of what they are doing.
But nonetheless, your logic is shit. I know HV electricians, pipe workers, welders, heavy plant operators, who don't even know what the internet is, but spend their lives doing stuff more dangerous that you can comprehend.
It constantly annoys me that geeks think that they know better than everyone else, just because they know the exact ins and outs of computers and networking. Yes, they are important... but there are far more important things in life.
You cannot be a electronic engineer if you don't understand thermal effects with regards to electronics. It's one of the most important performance limitations in a PC. It's important in most electronics, if not all, apart from muppet circuits you build on a 250 in one electronics lab.
Once you actually get a silent PC (I made one of mine silent by sticking it in a closet in the room adjoining my media room, and just running long cables for everything), then you have the joy of discovering that everything else around you makes noise, too.
My neighbor's pipes are loud. So is my refridgerator. Even my TV and DVD player make more noise than I would've thought.
If you're going to be so fanatical as to water cool a power supply, all I can say is, good luck with all the new stuff you'll eventually find annoying.
-- I wanna decide who lives and who dies - Crow T. Robot, MST3K
--as long as someone is modding their case, etc, whynot just remove the power supply to someplace else? Why does it have to be inside the case all the time? Resistance drops on dc wiring ain't all that bad within a reasonable distance, like inside a room to over yonder in the closet. And wires are cheap, and connectors can be cobbed. Put the power supply someplace away from where you are, add a larger low rpm quit fan, not have to worry about fitting in more stuff inside the case or concentrating the ambient noise, and having the power supply outside the case will have it run cooler just from not being so confined and next to the other heat sources like the drives and cpu and vid card.
I guess you could use it as a coffee warmer to disguise it as well if it was still inside the room. heh.
If we are doing this for sound reasons (pun intended), then you should remove all fans from your computer and submerge the whole computer in hydrofluoroether . Well maybe you want to keep your floppy and CD drives out of it. And it boils at 61 degrees Celsius, so you can visually see if it is getting too hot. TechTV has a story about building a case and sticking your motherboard in it, but no reason why you couldn't do the same for the whole computer.
It would look cool in a fish tank with some fake fish and plants. Plus you could use your aquarium thermometer...
Ok, now, at least in countries and universities where a degree counts for something, to do an electronic engineering degree, you need a physics A-level or equivalent. You certainly need a science anyway. So you would know that water has a very high specific heat capacity.
If you know about thermal effects in electronics, it is fairly easy to work out that water is very effective at moving heat. I certainly learnt this in my first year of my degree.
--I've got 12vdc run already, albeit it wasn't a big deal, we live in an RV. It had some 12 volt stuff already, but I added a lot more when we added solar pv to the power mix. My goal is to eventually have zero ac wiring "needs" anyplace, cut a big chunk out of the electric middleman of make dc, invert to ac, convert back to dc for the computer, which is the last actual needed ac appliance we have. Our lights are dc fluorescent, I can run one of the laptops dc cuz I got the gizmo, my radios are all dc, little fridge is propane or dc, vacuum is dc, last major ac appliance I use all the time is this desktop and monitor, which are serious juice hogs anyway. Your idea is kinda neat if you can get a large enough computer power supply that you can pull from for all your boxes in different rooms.
But ya, in a house, having the option of dc wiring is just another + to go along with running ethernet and coax. For low amperage, you actually could use an ac circuit that exists,pull it and put it in a sub panel, but you'd have to be careful with plugging stuff in (duh), and you'd still have to have dc voltage converters to get the 5 or 3.3 or whatever you need. To run all them you'd need a separate hot for each voltage, but they could share a ground I guess.
1) High Thermal Conductivity
2) Very high electrical Resistance (insulators)
3) Fluid at ~0-200 degrees C
If there are no suitable fluids, perhaps merely a powdered solid would be workable
4) Low chemical reactivity - not poisonous or corrosive.
I envision a change in packaging where the
1) silicon wafer is mounted on a stand-off inside a thin composite, mostly electrically insulated 'tube' (see 2), so that a fluid as above could entirely bathe the chip
2) the connections of the chip connect to discrete contact points (possibly in three dimensions) which are conductors through the tube to the outside, which are the 'pins' of the IC.
3) Fluid is constantly pumped through the package, going to a (variously sized) 'vat' of fluid: once the heat is away from the pinpoint source, it is very much easier to cool.
-J
I'm the (crazy?) guy that wrote that article and I feel it's only fitting I answer some of your reservations :)
First of all, the article included a disclaimer indicating the dangers of capacitors inside PSUs and hence the need for caution. That said, I've never opened a PSU and magaed to find any voltage left in them (carefully tested with a multimeter).
The reason I used water is that it is the best combination of price vs performance. Admittedly if there was a leak, my machine would be in serious trouble. The way I've constructed the thing, it's no more likely than the fan dying to be honest!
The PSU does actually still get a little air flow since my case has two 120mm fans at the front running at 5v. They're blowing across the radiator which cools all the stuff that's water-cooled.
With regards to the worries about "live heatsinks" I've tested every PSU I've water-cooled and in each and every one, there hasn't been a single "live" heatsink.
Hope that clear things up a bit for the more sceptical of you ;)
Take a look at the later-model G3 iMacs and the G4 Cube; they have no fans. The Cube uses an external power supply, not inside the case, and it radiates heat like any adapter brick. The iMac has the power supply inside, and as hot air rises out the top it draws cool air in the bottom.
Obviously the iMac only works that way because it's also engineered not to produce very much heat; I wouldn't expect the same to be possible with your standard Athlon system. Even Apple's newer desktop systems don't run cool enough not to need fans. Still, it seems to me that some of the same concepts could be applied.
One of the reasons for having a fan on the power supply is to act as a case fan as well - it draws warm air out of the rest of the system, and cools the power supply on the way out. It seems to me you wouldn't want to just remove the power supply fan, unless you had another way of cooling the rest of the system.
Hmm, I was going somewhere with this. Oh well.
$x='S24;r)>63/* h@<5+oZ)32"5cz';$me='phroggy'x$];
$x=~y+ -xz+\0-Tx+;print$_^chop$me for split'',$x;
Actually...
Switching power supplies are generally designed with only enough ESR in the output caps to make the thing stable. Depending on the switch frequency, there can be lots of ripple current flowing through them. A general rule is that there will be 0.4 times the output current flowing through the output caps as ripple current. If you have a 10-amp supply, count on 4 amps of ripple current. Now, if you're thinking of your ohms law, you wonder how much power is dissipated because of that. The answer is: none that is due to the capacitance. All the capacitor does is change the phase of the power relative to the voltage (power factor). However, capacitors have an ESR (equivalent series resistance) that will dissipate some real power from the ripple current (and put a ripple voltage on the output). Typical ESR numbers for caps used in switchmode supplies range from 5 to 15 to 50 milliohm. So your 4 amp supply could be dissipating enough power to heat up your caps...
Caps are constantly a problem in switching supplies, because a bad one CAN explode violently, especially if it's Tantalum or Aluminum Electrolytic. There are some new Niobium oxides that have a much lower ignition energy than Tantalum, but their ESR characteristics are somewhat less ideal for switchmode power supply applications.
Another thing to note about switchmode power supplies none of the caps should have a high enough voltage to hurt anybody. Usually, the input transformer steps down the line voltage to a level suitable for input capacitors, generally something at least 20% less than 35V (the popular rating for input caps).
Even better would be a complete case and power supply, so you could route some water to the drive carriers, so the sound insulation doesn't make them fry. You'd have to do some custom plumbing to hook in the CPU cooler, but the rest of the system could be completely assembled with the case and PS.
this is what is known as a BAD IDEA. heating up your toilet water would allow the aroma of the remants of last night's spicy-cheetos-and-beer meal to waft with unprecedented stinkiness.
but you are true to your word: you are indeed an eccentric genius. for it is an eccentric genius who would rather have a faster/quieter/more kickazz computer than less smelly puup.
as for me, i'd like to keep my bathroom smelling as fresh and clean as the day a cool spring rain. perhaps someone could develop a system to cool toilet water? maybe use fluorinert, as they used in crays, to freeze the puup on contact?
One problem would be that a computer (PC type) is not a static load on the power supply. I read some Intel design notes on supplying power to the CPU that imposed some very nasty requirements on the CPU voltage regulators. The power drawn by the CPU can vary over a large range and change very quickly. That varying load is going to cause rapid changes in the power drawn from the main power supply. Excessively long wires between the PC and the power supply could cause all sorts of problems with voltage regulation.
Mea navis aericumbens anguillis abundat
the stuff we use at work has to have a minimum resistivity of 18MOhm/cm; I wat till it stabilizes at 18.2 usually, put that in a clean, mostly-plastc system and you'll have no worries. Still, ground it all just to be sure. (i work with 3-10kV H20-colled powersupplies, granted they're in the neighborhood of .1mA, so...)
Facts do not cease to exist because they are ignored. - Aldous Huxley