Slashdot Mirror
Aquarium Full of Oil For PC Cooling
JaredOfEuropa writes "Forget fancy watercooled CPUs or complicated heat pipes. Annoyed with the noise of the forced-air cooling in his computer, this guy simply dumped his entire motherboard in an aquarium filled with mineral oil. (coral cache). No modifications were necessary; he even left the fans running to keep the oil moving about. The only thing not submersed in oil is the hard disk."
The reason that you can't submerge electrical parts in water is the simple fact that water is electronically conductive given the presence of electrolytes (which is so hard to guarantee against in practice, that you can rest assured that water is bad.) Many oils, however, are not electrically conductive. Therefore, as long as the medium used is electrically inert and does not chemically react with anything you're submerging in it, you can consider it a viable medium for immersion cooling assuming that heat transfer properties are favorable.
This has been done before. Interestingly, some projects have looked to it for outdoor computer use (stationary) due to the water-repellant properties of many oils.
I saw a few years ago (probably here on /.) where someone put a MB into a tank filled with glycerin, then put an air conditioner cooler grid into the tank with it. A pump curculated the glycerin over the cooling grid and around the MB. I thought that was pretty extreme. I guess the main point is that you don't want something corrosive or conductive, and you do want something with a sufficiently high specific heat to take the heat away without cooking the board.
There is nothing so silly as other peoples traditions, and nothing so sacred as our own.
I'm an elevator Engineer. This reminds me of a very old residential elevator controller I saw recently that was installed in 1917 and still had all the original equipment in good working order. The controller was in a cast iron tub with all the relays mounted to the lid and suspended in transformer oil. There was a hoist in the ceiling to enable lifting the lid for access to the relays. It would cost a fortune to build something like that today, but it certainly was durable.
"Do the Right Thing. It will gratify some people and astound the rest." - Mark Twain
Even with no heat loss to the air, I've got 10 minutes per degree C raise, at 20 gallons of mineral oil and 200 watts of power.
It's probably moot because that's a ton of cooling surface area.
Yah- I remember an experimental F-5 radar that emitted 40Kilo Watts (take that, Pentiummmm!). It was about the cubic of a full tower ATX case. Aluminum case, machined heat sinks on the inside, the outside was mounted to a cold plate that was chilled with turbine bleed air. The R/T was mounted inside the case in a three dimensional kind of array of solid state and passive components. Fluorinert filled the cavity. Screwed the lid on, and it went to work. Heat flowed pretty well (thermal sensors built in to the circuitry at various spots) so straight convection was used (no fan, etc.). Worked great !
I'm surprised that the PSU and all the cables (like speaker/CAT5) work at all, I feel so uneducated.
Actually, I'd be more worried about the high-speed circuits in the machine. Oil does not conduct electricity, but that doesn't mean its electronically equivalent to air.
Oil has a dielectric constant of between 2 and 3 (depends on the oil) and that will affect the capacitance on and between the traces of the circuit-board. The signals will run a little slower on the board and have a bit more cross-talk. Its probably not a big deal -- the materials in the circuit board have a bigger effect -- but it could slow the signals enough to reduce reliability in a marginal design.
Two wrongs don't make a right, but three lefts do.
He doesn't mean the insulation. Mineral oil won't eat through that (usually). He means the wrapping of the cables. Inside that, the conductors are indevidually insulated with an oil-resistant material (pvc, often). Transformer coils are insulated with a different kind of oil resistant plastic, and are often submerged in oil.
Or better yet, volumetric heat capacity...
The Good Eats episode on How To Thaw A Turkey did a great job of explaining that in layman's terms:
Cool running water melts a ice cube duck faster than a pot of water removed from the stove just as it started to boil.
"I don't know, therefore Aliens" Wafflebox1
And you could add 2 giant fans to blow air across the fins to keep it even cooler!
That is actually a good idea that really would make cooling more efficient. Larger fans can move as much or more volumes of air at slower RPMs than smaller fans. Lower RPMs means less wear on bearings and quieter operation.
IIRC that is the strategy used in the new BTX form factor cases--the heat sync on the CPU is really big with a lot of fins and a big fan that draws air through those fins and over the motherboard (to cool the chipset). Current ATX setups are most often laid out poorly for cooling, and you end up seeing high-end systems with 3 or more fans in the case. It is the need for multiple small fans that makes these PCs noisy, not the fact that they require fans at all.
I still think it would be great to see the return of the days when chips and power supplies ran cool enough to allow for practical convection cooling. My fanless Atari ST was blissfully quiet--even the comparable IBM ATs of the day that only had a single fan in the power supply were horribly loud next to it.
As this is really a re-hash of an old story, I'd be more interested in what happens to these rigs over the long haul. How long does the computer last in oil? How often does he have to change the oil? How does he cool the oil? How long before the mobo and cards are somehow affected by the oil?
Answering *these* questions would make for a much more interesting article than just "Hey, dude, I put my mobo in oil! I'm l33t!"
So what about the connections in the PCI slots and such? Can't mineral oil get between the card and slot's contacts?
Do you have to keep the cards perfectly still to keep a strong contact?
Can you swap cards in the tank?
Can you adjust a connection (USB) while the machine is on?
I looked in the forums and nobody mentions this in english. Just stuff like "bloody brilliant"..
Oil isn't even necessarily the best material to do this -- it's probably just the cheapest and easiest for a hobbiest to get hold of.
Liquid Fluorocarbon does an excellent job. The Ontario Science Centre used to have a great display of an operating television completely submurged in a small vat of the stuff. And fluorocarbon is effectively a plastic itself, and thus is harmless to plastics (unlike many oils).
Yaz.
is that every last connection has to be SOLDERED DOWN in order for this thing to work for more than half an hour.
The problem the guy ran into six years ago was that the mineral oil seeped in between all of the connections and disrupted the flow of electrons; PCI cards, AGP card, CPU, IDE, power... everything. A stock motherboard simply won't cut it, you have to have a custom board with everything hard-wired to it to survive the submersion.
This story is a dupe because it doesn't solve the basic problem.