Actually, yes, I think ALL the electricity that goes into it becomes heat. That's true of any used energy - eventually it all just becomes heat.

The puget systems people (http://www.pugetsystems.com/submerged.php) reported 12 hours to stabilize (it is asymptotic, though, never actually reaching final temperature, but creeping up to it). The final CPU temp was 88C (hot!). But then they added a radiator and it came down to 45C.

Drownmypc.. wow! You've done the figures.

When I get home I'll be applying those formulas to my aquarium size, and see what it predicts for the PC I'm planning on installing there (low powered).

Then, when I get the aquarium (going to have to ring them and ask them where it is!) I will compare the real results against what is predicted by the formula.

If you can come up with the differential equation for predicting temperature over time, +10 points My maths is too weak to attempt that.

So, I ran the numbers through with my tank dimensions (17" x 10" x 14"), assuming that only the sides (not the top and bottom) will be exposed. I assumed an ambient temperature of 24 degrees C.



This is promising, given that my system runs off a 60W picoPSU, and on my cheap power meter it rarely draws more than 50W peak, and typically around 30W. If the numbers are correct, it looks like I can expect my system to stabilize at around 10 degrees C above ambient. Even on the hottest summer day with the air conditioning off, our house is pretty cool. It might reach 35 degrees C indoors.

I can't wait to get the aquarium and test this out.

Excellent! It will be nice to have experimental measurements to back up the calculatations. I won't trust the math until it is verified.


Good grief! My wrist watch uses more energy than that (slight exaggeration).

BTW - I ran your tank dimensions and ambient air temp through my spreadsheet, and my calcs agree with your calcs 100%, so I think we're cranking the numbers correctly (that's encouraging).

BTW - I tried making my own heatpipe yesterday. Didn't work.

Well, I took the plunge and drowned my PC in mineral oil.

It's nothing special to look at. Ain't fancy or big or anything.

But it's pretty quiet.

The PSU and HDD are not in the oil, and they make some noise, but it's a lot quieter than before!

Still seems to work. No problems yet - course, it's only been about 4 hours now.

Congrats!! I 've wondering what happened to this project.

So if you kept out the HDD and PSU, how much quieter is the optical drive now? Aren't the CD/DVD's kinda hard to get the oil off afterwards?



sorry couldn't resist.

Nice work mate!

I'm still yet to do mine. Still trying to get my head around some "mechanical" issues.

I'll be interested to see the photos, and how your results compare with the theoretical results you calculated.

why didn't you put PSU in?

Hi guys, I'm actually the creator of the aquarium computer at Puget Systems (details at http://www.pugetsystems.com/submerged.php). Great discussion here! Vaseline is a really interesting idea!

What do you think of the implementation from a silencing perspective? Too over-the-top? Or is over-the-top the part that makes you want to do it?

Hey Jon. Nice work on your PC and the video. As you can see it has inspired others. It's definitely over the top, and that is definitely part of the appeal to me Did you find silentpcreview due to people following links from here?

Maybe somebody in this thread has some construction advice for me?

http://www.silentpcreview.com/forums/vi ... 08dbf97ea9

To Jon: You guys did a GREAT job. I decided to do this because of YOUR video. It's definitely over the top, and that's why I wanted to do it. It's also extremely effective at silencing the high-speed mobo and GPU fans.

To xan: smart ass!

To wim: I have seen posts/pages showing submerged PSUs, but in this case I didn't want to add the PSU's heat output to the oil. I have a Zalman 850W PSU that has a 140mm fan. It's very efficient and very quiet (but not silent) in open air.

To bruce: I'm having mechanical issues, too. It's not my strong suit. I just sat the mobo and gcard in the oil without any mechanical support -- not a good final solution. I'm still thinking about how best to support the cards, and how to physically organize the whole thing.

I have a double-hull setup: a 5.5 gallon aquarium placed within a 10 gallon aquarium. Initially, I did this because I didn't completely trust the cheap 5.5 gallon aquarium, and I wanted a backup container. More on this later.

The length (front-to-back) of the motherboard is about a half inch shorter than the depth of the aquarium, which is perfect. I attached the mobo to a sheet of plexiglass (1/4" thick, I think) which is about 6 inches longer than the mobo. I put a GPU into a PCIE slot, and dropped the whole thing into the tank. That extra 6 inches of plexiglass gives me something to hold onto when I pull the mobo out for servicing (which I haven't done since filling the tank - I'm not looking forward to that).

The GPU card is supported only by the PCIE connector. I know this is bad. I need to get a proper frame for the mobo and GPU.

I left all the fans on the mobo and GPU. They spin just fine in the oil -- just slower. I'm sure there is a cooling benefit as they churn the oil and mix the hot stuff around the tank. There's two fans on the mobo, one on the CPU, and one on the GPU.

The inner tank (5.5 gallons) sits within the outer tank (10 gallons), lifted two inches by sections of 2" diameter PVC pipe. Initially this double-hull was for safety (ever try cleaning up 5 gallons of spilled mineral oil? No? Me neither, and I aim to keep it that way!). But now I'm thinking the outer tank can help with the cooling.

Currently, the outer tank holds only air. This is bad. It's acting as an insulator, because the air isn't moving around much, and is slowing down transmission of heat energy to the room. I think my next step is to fill it with water.

That will make 5 gallons of mineral oil and 5 gallons of water. Since water has 2.5 times the heat capacity of mineral oil, I'll be more than tripling the thermal mass of the system, which means it will take over 3x the heat energy to achieve a given increase in temperature.

Furthermore, instead of having a layer of immobile air surrounding the inner tank, it will be water, which conducts heat much better. The insulating effect will be gone, and the whole shebang should be able to shed heat (albeit passively) much more effectively.

As for active cooling -- I may experiment with ice packs or water chillers for the outer shell. I'm just playing around with this, after all. I won't pretend any of this is practical.

I'll see if I can get photos of this - I realize it is hard visualize what I'm describing.

Anyway, this system is a midrange CPU and midrange GPU sitting on a high end mobo. Front side bus is pretty fast with 1.3GHz@1.2v. Honestly, I don't know if this system pulls much juice or not. But even when it is just idling, the oil gets HOT! Looks like the oil has taken about 8 hours to get close to 120F, with the system thermometers reporting 118F for the CPU, 127F for the system, and 144F for the GPU. The system is just idling, showing only the desktops. No graphics, and CPU meter shows 0%.

I would consider this a dismal failure, but I still have that outer tank holding (very warm) air, acting as an insulator.

I'm going to run a few torture tests on the system, record the stable temps, and then see what happens when the outer tank is filled with water. I really do expect that to help, although I don't know by how much.

-stefan

I resemble that remark. And at least im not getting called a dumbass - like usual!
I just had this nagging image of an opti drive tray poping up out of the oil.

--> On topic.
Id be real curious to see how temps differed in 10 gallons of oil. I think 5.5 gal isn't enough mass to cool it properly.

Maybe cooling fins sticking out of the oil would help in such a small case. In fact the most practical application of this idea i can think of would be to use a finned aluminum tank in stead of plexi/glass. the oil is the heat pipe for the whole mobo and components and the aluminum finned tank is the heat sink. Of course you loose the coolness factor....

as insurance maybe pick up a plastic kiddie pool too put the whole thing untill the kinks get ironed out. Cleaning up 5 gallons of spilled oil is probably not as hard as cleaning up 5 gallons of oil mixed with 5 gallons of water.

One other thing to consider is the capillary effect. Oil will travel up wires inside their insulation. This means it will eventually get into the harddisk, monitor, keyboard and mouse (if not cordless), unless the part of the wires and cables under oil are completely oil-proof. It can be stopped by baring the wire completely in one spot and then soldering it, from what I have heard, but I've never actually tested it myself.

I started planning an experiment myself, involving submerging a PSU in oil and then watercooling the oil, but I stopped because I couldn't get the special oil I needed, at least not in affordable quantities. I am talking about Luminol TRi, which is a special inhibited transformer oil that is engineered to not eat away electronics. They only sell it in big drums though. :p

One thing I considered doing, but never checked, was spraying the PSU with a teflon spray first, to insulate capacitors and such from the oil. It should allow for a good heat transfer and insulate against electricity, but I never got around to really looking into it.

Edit: I should add, like has been mentioned already, the system should be completely sealed so the oil doesn't evaporate into the air. Bad for the health.

Using more oil means it will take longer to reach equilibrium. Greater surface area of the oil will lead to faster heat dissipation, as well. There can be no doubt that more oil/bigger container is better, but the real issue is finding out how much is good enough and that is going to depend on what components are used.

I filled the outer tank with water last night. Now the idle temp of the oil is 100F instead of 120F. The ambient air temp is 70F.

I'm having doubts about whether these tanks will dissipate enough heat to cool a powerful system.

Celoth: interesting point about the capillary effect. At first, I didn't see it happening, so I doubted. But now when I touch a cable which is well outside of the oil, I can see just the tiniest glint of oil on my fingertip. So, I think you are right -- but it is very very slow. That is worrisome - I don't want it gumming up my HDDs.

Also, about the health. I've read a few hazardous materials sheets on mineral oil, and none have mentioned hazardous vapors. Can anyone pont me to an authoritative source that describes this hazard?

To Xan: you mentioned a finned aluminum tank. I was thinking EXACTLY the same thing! I wish I knew how to weld.

Once again, I was looking for info about hazards of mineral oil - because of Celoth's remark about vapors.

Here are data sheets I found:

http://www.sciencelab.com/msds.php?msdsId=9927364

http://www.arb.ca.gov/db/solvents/solve ... al_Oil.htm

http://www.finalube.com/reference_mater ... al_Oil.htm

http://www.jtbaker.com/msds/englishhtml/M7700.htm


I'm not a chemist, but I think the term "mineral oil" is not extremely precise. It covers a range of molecules in the alkane family (see http://en.wikipedia.org/wiki/Mineral_oil and http://en.wikipedia.org/wiki/Alkane). Someone (Spare Tire?) in this thread mentioned vaseline - it is an alkane, too, just with longer chains than mineral oil.

Data sheets on "mineral oil" may actually refer to various different compounds within the alkane family. That might account for the apparent disagreements among them (for instance, flashpoints cited at 135C, 168C, 215C).

There seems to be disagreement about vaporization of mineral oil. One data sheet claims insignificant evaporation rate at ambient temps (24C). But, one of the links above show nonzero vapor pressures at higher temperatures. I don't know how vapor pressure translates into evaporation rates.

Anyway, given Celoth's warning, I definitely want to find a way to seal the tank.

I had a science teacher who tried to weld alumnium and he said it explodes (!) as it was oxidized by ambiant air before it melted to any workable point. In the industry, i believe they do it under innert atmosphere.

Just met a gal who welds - I will ask her about the ins and outs of it over coffee Hoping she'll teach me to weld some day!

I know aluminum can be welded - my bike is aluminum and its welded... Just a matter of how.

I would imagine any case made from metal would out perform glass - Shoot buy a 55 gallon metal tub of mineral oil and just plop the mobo in - figure out a grommet in the lid to stop evap and over come the capillary action...

I can't believe I've missed this thread this long! I built my "Oil Box" PC over 5 years ago, although it certainly doesn't seem like it was that long ago. I wish now that I would have documented it better, but I never really expected it to work, so I didn't bother. There's a limited discussion of it in this thread though. It also got a passing mention in the Wall Street Journal article written about us crazy SPCR'ers, but I don't recall the link to that.

My hardware was a K6-2 350, overclocked to a screaming 500Mhz. One of the more useful discoveries I stumbled upon was that 80mm fans would spin just fine in the oil, and made a significant impact on the temperatures. Just like with air cooling, forced circulation is much more efficient than natural convection is. And in the oil the fans are silent.

Except for the HDD's, there was no negative effects on any of the hardware. In fact the NIC from that system is still in use around the house here today.

Rusty,

Wow. You were ahead of your time! Do you still have an oil box running?

You mentioned converting your desk. I followed the link, but the page is gone! Waaah!

I read through the thread you mentioned. One of my favorite remarks is:


Anyway, I'm full of admiration -- if you have time, I'd be interested in knowing about:

How portable was the oil box? Was it hard to move? I think I'll have to drain my machine before I can relocate it, because 10g of water/oil is freakin heavy! Plus, the fluids slosh around if I move or tilt the rig.
Did you take any special measures to cool the oil itself? Such as radiator, case fins, etc?
How expensive was the turbine oil?
Did you investigate health impact of oil fumes?

I just had an epiphany! (Uh, gesundheit?)

My feeling is that my passively cooled oil+water double hull doesn't throw off heat fast enough to handle a heavy load. After all, the oil is at 100F when the CPU is idling. I'm guessing that if I put a BIG graphics card (or two) in there and crank it up, it would stabilize at an uncomfortably high temperature.

So I've been considering various ideas such as:

1) running the water (or perhaps the oil itself) through a radiator/chiller/heat exchanger.

2) dumping ice/dry ice/frozen gelpacks into the water shell for temporary cooling (just for fun - I know that's totally impractical).

3) replacing the glass outer tank with an aluminum finned tank (I actually plan to do that).

4) running ambient (room temp) air through copper tubes submerged in the water/oil.

Let's think about #4 for a moment.

Heat transfer is all about surface area and temperature differential.

I have a 20ft coil of copper tubing I picked up at a plumbing supply store. I could submerge that in the oil (or the water) and force air through it. Exiting air is simply blown into the room away from the system. The copper tubing has a 1/4" inner diameter, and is 20ft long, so the surface area where hot metal (100F) meets air (70F) is 2*pi*1/8inch*20ft = 188 sq inches. That's only 1.3 sq feet, which isn't much compared to a proper radiator (or even one face of the aquarium, for that matter).

Now, I recall that the Puget Systems people added a bubbler and reported a slight temperature reduction. That makes perfect sense, because they were creating a water/air interface. The air in each bubble will gain some heat from the hot water, and when the bubble reaches the surface, the warm air is released into the room and carried away from the system.

So, how much heat energy will the air from a bubbler absorb? To even begin to estimate this, you'd need to know the aggregate surface area of the stream of bubbles, which I don't know how to measure.

But I do know that small bubbles rise through fluid more slowly than big bubbles do. So, small bubbles would have more time to absorb heat energy. So, when selecting a bubbler, you'd want one that has high volume and produces little tiny bubbles.

Rusty & Jon -- did you experiment with bubblers for cooling?

Nope. Disassembled it when I graduated college and moved. A couple of years ago my former roommate took the Oil Box idea and applied it to just the PSU in his watercooled system. There's an article about that little experiment over at Overclockers.com. That turned out rather well, and he used it for almost a year. The capillary action through the PSU cables was a problem though...he would slowly accumulate a little puddle of oil under the motherboard.



Oh it's still here, it's just buried way, way on the back shelf: Desk PC




The short answer: not very portable at all. My system was in a plexiglass box that I built, roughly a 12" cube. I sized it so that I could stand the mATX mobo up against one side, with the backplane connectors just poking out the top of the oil. That prevented capillary draw up through the network and keyboard cables. And there's more than just the wieght to worry about, there's also the stress on the aquarium. Those things are basically held together by caulking in the corners - a little flexing while full of liquid and you could pop a huge leak.



Nope. My system was such low power that it probably didn't need much in the way of cooling help. I used an aquarium thermometer to monitor the oil temperature, and after a day or two it leveled at something like 120°, if I remember right.

"Turbine Oil" is exactly the same as Mineral Oil. Different companies call it different things depending upon its intended use. I bought it from an industrial lubricant supply house, in a 5-gallon bucket, for about $40 (but with the price of oil now compared to 5 years ago, it's bound to be more today)

And being a crazy college kid the idea that this could hurt me never crossed my mind. Hey, they sell this stuff as "Baby Oil", how bad could it be, right?

Rusty,

About the Desk PC: WOW, Rusty, that's a really nice job.


Yeah, I can't move mine, either. It's not a huge issue, but it is inconvenient. I have the oil filling the aquarium up to about a 1/2" from the top, which I have decided is too close. The oil could slosh over the side if I shift the aquarium around. I think maybe it is best to arrange it so that the oil is a few inches below the top, to make small adjustments easier. There was also the idea of using Vaseline instead of mineral oil, which intrigues me, although I'm skeptical that it would cool sufficiently until completely melted.


I was worried about this too. It's why I went with the double hull (a 5.5g glass aquarim inside a 10g glass aquarium), in case the inner aquarium should spring a leak. Otherwise, I have a similar arrangement as you, with the backplane poking out the top of the oil and the keyboard, mouse, and other cables coming straight up for a few inches. The power cables are not so nicely arranged, however.



That's exactly where mine leveled out when the PC was idling, and when I had the outer hull empty of water (air filled, which was ineffective). But I want a system that is both quiet and powerful. Water in the outer hull helped (100F idle temp for oil tank), but I think more cooling will be needed.

Exactly! I worry a little - I keep looking out for health and safety information. I haven't found anything that makes me worry about vapors, but I still wonder. I'd like to have this thing running 24/7 in my basement, and even a little bit of evaporation could accumulate.

Tonight, I tried out an aquarium bubbler (perhaps actually called aerators, I think).

The first thing I noticed is that the "whisper quiet" air pump is redonkulously loud. It sounds like a jack hammer! I don't know how aquarium enthusiasts tolerate it. But, I was dumb about it. I just went to a pet shop and selected a pump that listed "quiet" as its primary design feature. I'll have to surf the aquarium web sites for a truly quiet air pump. (but only after I determine that bubbles actually help with cooling - I'll tolerate the noise while I experiment).

One thing I learned about the air pump: it is quieter under load. That is, it is loudest when running without any tubes plugged into it. Sort of like a car without a muffler. The air pump delivers air pressure in rapid bursts, and without something to dampen the output, it sounds very loud. But even after tubing up, the pump itself makes a lot of mechanical noise.

The second issue I ran into is foaming. I mentioned in a previous post that, theoretically, little bubbles would cool better than big bubbles. This is because

1. per unit air volume, little bubbles have more air/water boundary surface area than big bubbles do, which allows more heat energy to transfer from water to air
2. little bubbles float to the top more slowly than big bubbles

One of the products is an air stone (a porous rock that you attach the air tube to which emits bubbles) which makes tiny bubbles (they said micro bubbles -- no doubt someday to be obsoleted by nano bubbles). I submerged that baby and turned on the pump. Yup, those bubbles were pretty small. And, immediately the bubbles formed a foam at the top of the oil. And the foam built up like the head on a mug of beer. I had to turn off the pump, lest the foam overflow the inner tank and spill into my outer (water) tank.

Not good. For one thing, a layer of foam would act as a thermal insulator, possibly negating the cooling benefit I sought by percolating air through the oil in the first place.

The other issue, of course, is foam spilling over and making a mess.

The Puget Systems people (Jon Bach @ http://www.pugetsystems.com/submerged.php ) reported that they had installed an aerator and occasionally it would start to foam up (they think, because they'd sometimes find oil beside the tank). They didn't know why it sometimes did and sometimes didn't - but speculated it might be humidity. Personally, I have no idea what might be causing it. Jon, do you have any more recent experiences to report on this? Also, your website mentions that you'll be revisiting this idea this spring -- how's that going?

Anyway, my bubble options seem to be

1. Use bigger bubbles. They form less foam, perhaps because they pop more easily. But that means less cooling (theoretically).
2. Turn the airflow down (again, less cooling)
3. Don't fill the oil tank up so close to the top. Allow a couple inches of foam to build up - just don't let it spill over the sides of the tank.
4. Research some kind of additive that makes the oil less foamy. I have no idea if such a thing exists. I still have to be mindful to keep the mixture nonconductive.
5. Use bubbles in the water layer, not the oil. Water might not bubble up so badly.
6. Don't use bubbles.

I hate to give up on the idea of bubbles for cooling, because it seems theoretically sound. And, if bubbles can be functional, that makes them extra cool to look at (pun intended).

Stefan, three thoughts:

1) I'm impressed you're doing this.

2) One reason heat exchangers do not usually involve bubblers is because of fluid contamination. Crap from the air can get in the oil, and oil can get in the air. I don't know if evaporation is a health concern, but if it is, this will increase the rate of evaporation. The copper tube you have solves these problems, which is why they are much more common in heat exchangers.

3) LET'S SEE PICTURES!

Hi, Avalanche,

1) thank you!

2) Very good point about fluid contamination and increased evaporation. I vaguely knew about / guess at those issues, and I wondered whether we could do even better when not constrained by those problems.

3) I have pictures! But I'm a phpBB novice. I'm trying to figure out how to store an image someplace on the net and get a permanent URL for it.

Wow. Bubbles don't seem to work very well. I made that tank crazy with bubbles, and the temp didn't budge.

I did an experiment, and found no measurable difference in cooling between having no bubbles and having lots of bubbles for two hours. The Puget Systems folks reported that their bubbler made only a couple degrees difference. So, until I do something more careful and quantified, my results support theirs in saying "bubbles don't help much".

Frankly, I'm baffled. Theoretically, bubbles should carry some heat away. Maybe I'm wrong about the theory, or maybe I didn't run the experiment long enough, or maybe bubbles really don't generate much surface area. I dunno.

you're right about that last one; most bubbles are roughly spherical; the sphere is the geometric shape which has the least surface area to enclose a given volume, so most of the air inside the bubble is not in contact with the heated oil. also bear in mind air has quite poor heat capacity and the transit time of the bubbles through the oil is probably quite short.

All very good points. Especially about the heat capacity and transit time.