i think you can quite easily get an air cooled machine using high quality, big, low-rpm fans down to a quieter and smoother noise signature than doing anything with an aquarium pump will get .. but perhaps you are mucking around with the oil computer idea for novelty value rather than practical reasons..?

It is for the pleasure of tinkering with something off the beaten path. Definitely not for practical reasons.

Suppose you have a bubbler at the bottom of a tank. The bubbles are 5mm diameter (largish, I guess) and have a 1/2 second transit time from bottom to tank surface. The pump moves 30cc of air per second. I think these are all reasonable figures.

Here's the question: this bubbler has cooling effect equivalent to what sized radiator?

Consider a single bubble. At 5mm diameter (r=2.5mm), it has volume V = 4/3 * pi * r^3 = 0.06545 cc, and surface area A = 4*pi*r^2 = 78.54 sq mm. At 30cc/sec, the bubbler makes 458.4 of these per second. Wow.

Now, freeze time and look at the bubbler in action. Because the transit time is 1/2 second, there should be about 229.2 bubbles in the water in any given instant. That is a total water/air boundary of 229.2 * 78.54 sq mm = 18 001.3 sq mm. That's a little less than 1/5 of a square foot.

Granted, bubbles are being created and destroyed continuously, but there is an aggregate boundary of .194 sq ft being maintained all the time, across which heat energy may move from warm water to cooler air. So, to a first order approximation, I think this bubbler is like having a radiator with 0.194 sq ft of surface area.

Considering that my 10"x12"x20" tank has 5 sq ft of surface total area on the sides, plus 1.4 sq ft on top, this bubbler has a heat transfer boundary that is about 3% of the area of the tank's.

What if we fiddle with the bubbler to improve the effective area of the heat transfer boundary? The "micro bubbles" from one of my air stones look to be about 1.5mm in diameter, and they rise more slowly (about 1 second, I think). Each bubble is 0.001767cc in volume, and at 30cc/sec, the bubbler makes 16978 of them. Each has 7.07 sq mm surface area, for a total heat transfer boundary of 120034 sq mm (about 1.3 sq ft, which is 4x what the 5mm bubbles had).

So, theoretically, it appears that bubble size could have a huge impact on the bubbler's heat transfer.

Of course, all this is just an approximation. What are the secondary effects? Here are a few I can think of:

1. The bubbler's boundary goes directly from water to air, with nothing in between -- whereas the tank's sides has glass to retard heat transfer -- and a real radiator has metal. This helps the bubbler do even better per unit area than a tank or radiator or submerged air tube. Could be significant, but I really don't know.
2. If the fluid can evaporate (undesirable in oil, maybe desirable in water), then evaporation carries away additional energy. This favors the bubbler. This too could be significant (I think 1 pint water requires 300kWH of energy to evaporate).
3. As jaganath pointed point, air has a lousy heat capacity. Could an individual bubble approach water temperature before it reaches the surface? If so, heat transfer is retarded as the air warms (because the temp differential is diminishing). Even radiators have this problem -- as warm fluid travels through the pipes of a radiator, the heat transfer rate per unit area diminishes as the fluid cools and approaches room temp.
4. The bubbles start out slightly compressed, thanks to water pressure, which makes them smaller at the start (PV = NRT). Is that effect significant? Atmospheric pressure is 14.7 psi. The additional pressure under 12 inches of water is 1.3 psi, so the water contributes about a 8.8% pressure increase, which is an 8.8 % volume decrease, which leads to a 5.7% surface area decrease. So the effect is significant, but not overwhelming.

So, smaller bubbles increases air/water boundary per unit air volume moved by the pump.

A thicker fluid would slow the bubbles, lengthening the transit time and effectively increasing the aggregate air/water boundary.

A deeper aquarium would lengthen the transit time, increasing the aggregate air/water boundary (but also make secondary effect #4 worse).

I think there are interesting possibilities here.

(if anyone cares to check my math, I'd be grateful.)

Avalanche - here's a photo,



See a larger image here for a closer look.

This is my oil-submerged PC. It's really kind of a lab setup, where I can experiment with different cooling techniques. This is a double-hull arrangement where the mobo+GPU sits in a mineral-oil-filled inner tank (a 5.5gallon aquarium), all of which sits in a water-filled outer tank (10 gallon glass aquarium).

I have the inner tank raised 2 inches, so that its underside can shed heat, too. The inner tank is resting on a base of L-girders which in turn sit on two segments of 2" PVC pipe.

The PSU and hard drive are sitting on a gray box to the right of the tanks. I didn't want to submerge them because (1) they're already pretty quiet and (2) they don't need a lot of cooling. I think PSU placement is actually a tricky issue if you aren't submerging it, because the main power cables are fairly short, and they connect to the front end (i.e. deep end) of the mobo.

The mobo is attached to a plexiglass sheet for physical support. The plexiglass extends about 10 inches above the tank (almost invisible) - the video and sound cables are draped over it. Clipped to the plexiglass is a blue/white twisted pair of wire with the power switch. The reset switch and other front panel header components are not installed (I just didn't bother).

Originally, the outer tank was empty (as pictured here), and served only as backup in case the inner tank broke (I worried that the mineral oil might dissolve the glue/caulking of the aquarium -- it hasn't, yet). But, that outer layer of stagnant air was an insulating layer. After a few days I filled the outer tank with water to assist in cooling, and the oil temp dropped by 20F (not pictured).

At the rear left corner of the inner tank, you see a gray cable ending in an aluminum rod (dangling into the oil at an angle - you see it between the two white PCI slots). That's a temperature probe connected to a data logger (a LabJack U3) so I can record temperatures as I experiment with different cooling solutions. In the background, at the upper left of the photo, you can see the laptop I was using for logging the temperatures with this probe. I don't rely on the oil computer for its own temp measurements because I wanted to be able to measure oil cool-down rates after the power is off.

I have a power meter (Kill-A-Watt from P3) which the PSU plugs into, so I can estimate how much power is going INTO the tank. At idle, the PSU draws about 115 watts. The oil heats to 100F and the water reaches 90F. It is a wimpy system -- while running torture tests, it draws 145 watts, maximum. But, I have room for expansion: The mobo has two PCIE slots, and can accommodate two double-wide GPUs in SLI mode (and the PSU has the extra power cables for them). The PSU is rated for 850W and is 80+% efficient, and someday I intend to put more power hungry components in the oil and see how well they get cooled.

Earlier in this thread I gave tables on how quickly a hot aquarium would shed heat. Was theory in the right ballpark? Now that the system is plugged into a power meter, I can do a rough check.

The power meter measures 115W at idle. The PSU is rated at 80+% efficiency. So that makes 92+W being delivered to the system. Subtract a few watts for the HDD and the PSU's fan, and the remainder goes into the mobo+gpu. I don't know what the HDD pulls when it isn't doing anything special. Maybe about 12W?

That would leave about 80+W for the mobo and gpu.

Now, given the temps we measure, what does theory predict for heat loss rate?

The outer tank is filled with water, and measures 10x20x12 (12 inches in height). The sides and top have total area of .6 sq meters. The air in my basement is 68F (=20C).

I ran my spreadsheet with these dimensions and ambient temp, and got this table:



When the system is idle and reaches stable temps, the water (outer tank) is at nearly 90F (=32C). My calcs predict that it should be shedding about 85W of heat energy from the aquarium.

This might be a slight underestimate, because my calcs treat the top face of the tank as if it were at water temp (90F). Actually, much of the top face is the oil tank, which is at around 100F. On the other hand, I have that gray box next to the tank (on which sit the PSU and HDD), which may hinder some heat transfer on that side, so maybe my approximations balance out.

Anyway, theory and experiment are in the same ballpark. Shoot, I would have been pleased if they were within 2x of each other.

I wish I knew the real power draw of the mobo+gpu. I don't like having to deduce it from estimates of (wall power * PSU eff - HDD power).

-stefan

Hey Stefan,

Keep up the great work! I was happy to see that your results are pretty close to what you predicted with the spreadsheet.

I don't know if I asked you before, but where did you buy your oil and how much did it cost? I got mine from a horse supplies store.

Also, where did you get the data logger and temperature probe? Did you have these already or purchase them for this experiment?

Bruce.

Hmm, following up on previous post: so theory seems to be in the right ballpark, and my spreadsheet might be almost accurate.

Now, suppose my mobo+GPU were pulling 500W. My spreadsheet predicts the water temp would have to be at 80C (=176F) in order to shed 500W of heat energy from a 20x10x12 tank

Geez, and that's just the outer tank.

Passive cooling will definitely not cut it.

i remember seeing this same idea years ago when rusty tried it and thinking at the time it was interesting, innovative and a really bad idea.. if you think about the total surface area of the tank, which is just a rectangular prism, theres not all that much area to dissipate this heat by natural convection.. i haven't checked the numbers but its probably not much more surface area than say a scythe ninja or thermalright heatsink has exposed to the air. and if it really gets up to >50C (!?) you have a tank full of hot oil there which could give you a nice burn if you have an accident

Hi, Bruce!

I was pretty surprised by that, really. I mean, so much of it is estimation and guesswork. I didn't think they'd be even close.

I got the mineral oil from an agricultural supply store here called "Southern States" (http://www.southernstates.com/). It cost me about $10 per gallon. I dont think they have an online store.

The data logger is called the "U3", made by LabJack (http://www.labjack.com/). I bought it specifically for this and other related experiments. For instance, I have a bunch of heatpipes and was using the temp probe to gauge their heat transfer capability (unfinished experiment). The temp probe is sold separately. But it has a lot of channels! And it is potentially much more than a data logger -- you can control it from Visual Basic, C++, other languages, and it can do output too, and so control things like fans or relays or whatever, with the right accessories. It seems quite versatile - but I'm not an expert user.

Hi, wim.

Yeah, I have a reserator2 sitting next to me (I tinkered with water cooling a while back).

Lessee here...it's about 12 inches long, 11 inches tall = 1.83 sq ft without the fins. But there are 50 fins, each 11 inches long and 1 inch wide (at about 1/2 inch spacing). That's another 3.82 sq ft.

Just the main body has 5.65 sq ft, plus there's other parts with nooks and crannies. Overall, this thing has more surface area than my big ol' tank.

When I learn how to weld, I'm gonna make an aluminum tank with fins. And a spoiler.

It seems like the perfect application for a radiator. You have warm oil, and you need to cool it. The oil is surrounded by glass and air, both terrible conductors. I bet a Reserator or an auto radiator would make a huge difference. I wonder what kinds of submersible pumps work in oil...

Really cool project! Thanks for posting the pic.


For measuring power draw without hdd - download a F@H linux live cd distro and load it to ram with no hdd atached. after it loads the opti drive is unused. or boot to a similar thumbdrive distro -just a thought...

I would still would love to see what would happen to the temps if you slid in a piece of metal, about the same size as the plexi the mobo's on, into the tank. maybe heavy gauge corrugated roofing?
I would think it would really help a lot ...especially if the psu exhaust fan was aimed at the exposed end of the metal.

(I'll try it myself as soon as the kids kill the remaining fish...or i can learn to weld a custom metal tank):wink:

Good luck!

I've thought about doing that. Well, I happen to have a bit of steel plate. I'll try that out.



Interesting idea! Can Windows be made bootable without HDD? For instance, can you install XP on a thumb drive and boot from there?

Booting xp on a thumb drive is not for faint of heart. Linux is much easier in this regard. I would download a linux distro and just let it run from a cd.
The notfred F@H cd-iso will boot right to F@H core.
http://reilly.homeip.net/folding/

Just be sure to enter the SPCR team # 31574 and your username before downloading the iso, so we get credit for finished work units.

for cooling I was thinking that a piece of heavy wire mesh might work too.

Okay, I tried it. I had a sheet of steel, about 1/16" thick, maybe 16x24. I stuck it into the tank long-ways, so it was about 12 inches submerged and 12 inches sticking out.

The result: about 1 degree F drop.

The oil was running 125F in ambient of 75F. So, 50F above ambient. The steel was hot to the touch just above the oil. But about 2 inches further up, cool to the touch.

What does this mean?

It means the steel loses heat to the air FASTER than the heat can travel upward through the steel. Consequently, only a couple inches of of the steel is radiating heat away -- the remainder is wasted area.

The limiting factor is how fast heat energy can travel along a fairly thin cross-section of metal (1/16 x 12 = .75 square inches of metal).

Steel is not a great conductor of heat energy. Here's a link: http://www.engineersedge.com/properties_of_metals.htm. It shows that steel is not very good (10 - 73), pure aluminum is much much better (220), and the best are copper (386) and sliver (418). The units are W/mC. I guess this says that a sheet of pure copper would carry about 5x as much heat away as the same-size sheet of 1% carbon steel.

Consider that by comparison, when I put a small fan on the whole setup, I got an 8 deg F drop. Forced air rules.

-stefan

My double-hulled oil system has been submerged since early February - about 5 months now.

I use it intermittently for games and photoediting. It is off most of the time, but each week it averages maybe 5 powerons and about 5 hours of on time.

It still works fine - no loss of performance, no functional problems. Occasionally a label comes loose and floats around in the oil until I fish it out.

I keep having to refill the outer hull with water. It evaporates. I haven't lost any oil. I have seen no evidence of the oil-wicking phenomenon reported by others. The oil is still crystal clear, but the water is getting cloudy, and some of the metal framework that supports the inner hull within the outer hull has corroded.

My problem is a hot graphic card to cooldown silently,I don't like watercooling so ... in order to feel secure ,Ithink I would rather use one water c. under oil:oil would keep water away from the electronic devices?By the way you know water and oil don't mix together A big leak would go through oil?

Thanks for the update drownmypc.. and big kudos for doing this!

My aquarium is still sitting in my living room, empty, and my PC is sitting idle underneath.

why dont you buy some fish for your empty fishtank?It will ease your mind,let you think of new ideas....
Drownmypc,Rusty075...+1,I am looking for somebody ,involved in watercooling ,ready to dip it into ????a non acid ,basic liquid that would keep water away from electronics.About capillarity, wire we use on psu are getting warm then back to cold.Because, they are stuck into plastic tube :every use of psu make the metal warmer+bigger(due to heat),so it work as a seringue,or push pull .That is the way oil is sucked into plastic tube.
I am still looking for the ?oil liquid coolant that would not attack plastic,sealing,don't,can't work because of heat:thing a getting bigger then reducing make them suck oil...into their (oilthght) protection.

Hey guys, so i'm currently just starting out on this project! and i'm trying to get my hands on some mineral oil in the UK. I've got this website :

http://www.icis.com/Search/ProductNumbe ... (8042+47+5).htm

which gives me a list of suppliers for the UK, however whenever they ring them they ask what grade mineral oil i want! I dont have a clue, any help?

Also a lot say they deal in bulk so won't sell me the 50 litres I want, any other outlets, i've heard horse laxitive?

Paul

I drowned my PC in mineral oil 9 months ago. It has worked fine, although it hasn't seen heavy usage. The oil temp gets up to about 120F after about 12 hours of ontime, where it stabilizes (my basement is typically at 65F-70F).

The oil tank (inner tank) was always pretty clean looking, except for the occasional label that came off the PCB or hapless bug that fell in.

However, the water (outer) tank in my double-hull arrangement got really nasty.

First, I believe algae has grown in my tank. Covered everything. Nasty, slimy stuff. Fortunately, it wipes up easily once the tank is emptied.

Second, mineral scale appeared on the glass walls of the aquarium. Probably caused by high concentrations of minerals in the tap water I used. Frequently, I had to refill the water (outer) tank to replace the evaporative losses. But the minerals in our tap water don't evaporate, so I guess the minerals just got more and more concentrated over time, and eventually collected on the sides of the tanks. It took a lot of hard scrubbing with a brillo pad and tile cleaner to get that stuff off.

Third, the metal L-beams had gotten really crusty. Some white stuff covered them. Possibly that is the mineral deposits that collected on the glas, or possibly it is galvanic corrosion. The L-beams was made of one kind of metal, and the nuts&bolts used to attach them to the PVC pipe was a different kind of metal (and they looked pretty clean).

So, I have the follow adjustments to make :

(1) no metal in the water hull, to avoid galvanic activity.

(2) use distilled water only, to avoid mineral accumulation.

(3) find a way to prevent evaporative losses of water, because I don't want to have to keep buying jugs of distilled water. Somehow seal the top?

(4) Find a way to prevent algae from forming in the water. A simple additive should do the trick. Or maybe I can just use bleach or vinegar or something. Dunno.

Hi, blundell.

I'm guessing they wanted to know whether you needed "food grade", "medical grade", or "industrial grade", etc. I don't know precisely what grades there are, but I have a feeling that's what they were getting at. I should think that any grade mineral oil would be fine. Your submerged PCB isn't going to be terribly sensitive to the grade of mineral oil.

The mineral oil I bought was from "Southern States" which is an agricultural supply store here in the southern US. www.animedproducts.com. It is intended as a laxative for farm animals (horses, cows, etc). Says on the label, "Mineral oil is a gentle and effective intestinal lubricant. Mineral Oil aids in normal intestinal evacuation and is not absorbed, is non-fattening and will not increase body temperature. Use one pint for each 120lbs of body weight."

Yeah, okay, they say it is safe but there is no way I'm ever drinking the stuff. It's a petroleum product! And if you should happen to aspirate it (get it into your lungs) there is no known method for removing it, and the body is unable to process it, so it stays there forever. Aspirating mineral oil is bad.

Anyway, I suggest you look for agricultural or veterinary suppliers in the UK. Try using Kelly Search to find something: http://chemicals.kellysearch.co.uk/sear ... l+oil&gcc= or search for "veterinary supply", etc.

-stefan

Good to hear its working out!

How hot does the water get? Maybe you just need some high temp loving sucker fish.

I gotta get off my duff, and try this soon. I am wondering how a semi-oil-submerged ninja, or two, might help dissipate heat.

Heh! I doubt any fish can tolerate these temps. The oil temp stabilizes at about 120F after approximately 12 hours of power-on time. Ambient air temp is about 70F, so that's a 50F temp delta.

I think this temp is safe for my machine, but when the water is that hot, it evaporates quickly. It's annoying to have to refill the outer hull. So, I'm looking for a solution to that problem.

[Edit: continually updating the table with experimental results]

Well, last night I started an experiment on evaporative water loss -- because, as I mentioned recently, I'm annoyed that I keep having to add water to my outer tank.

I filled two identical glasses with water. Then I added a little bit of mineral oil to one glass. The oil floats on top of the water, making a barrier between the water and the air, about 8mm thick.

Then I weighed both glasses with a postal scale and put them both in front of a fan (a stiff breeze to accelerate evaporation).

Glass A had just water. After a little over nine hours, the water level has visibly dropped in the glass. I measure 23g of evaporative water loss.

Glass B had water+mineral oil barrier. After many hours, no visible drop in water level, and no visible thinning of the oil layer. Sometimes I measure 1 g of weight loss, sometimes not. Since the postal scale I'm using has a 1 gram resolution, I can't be sure that we really have lost a gram of mass. It could just be an artifact of the postal scale's accuracy.

I'll be filling out the table below as the experiment runs.




[Edit -- strange that the evaporation rate would accelerate like this]

One thing is certain -- the layer of mineral oil on top of the water definitely impedes the evaporation of the water.

The questions I still have are:

(1) exactly how fast is the protected water evaporating, if at all?

(2) is any of the oil itself evaporating?

(3) if I dope the water with a bit of vinegar (for algae prevention), how will that affect evaporation? And might the vinegar interfere with the oil's protection of the water? I also wonder whether the vinegar's acidity could weaken the silicone caulking that seals the aquariums.

[Edit: after 4 days, we don't seem to have lost more than a gram of water from the oil-covered glass.]
-stefan

The water evap rate is limited by the solubility of the water in the oil: its near 0.

The mineral oil has a very very low evaporation rate at room temperature, I doubt its measurable outside a lab.

Vinegar will not prevent algae, and it will smell nasty. I recommend a few changes to your setup:

1)remove the metal supports from the outer chamber, as you have seen, they corrode in water. You can support teh inner tank on a plastic support or use a bed of aquarium gravel to distribute the Load.

2)Fill the outer tank with a mix of bleach and water, and pour a mineral oil cap over it (use about 1 inch of mineral oil)

That will prevent algae and the oil will prevent evaporation. Just don't spill it.

Better still would be to get a aquarium pump and a radiator and do away with the outer tank entirely.

I looked at putting one of these together a few years ago, for a slightly different purpose: with proper cooling, you could bring such a system down to temps below 0C, with no worries about condensation. With memory, processor, mobo, video card all operating at 0C there is a significant overclocking potential.

I was stopped by the problem of wicking, I never found a good solution and the mess was abysmal. Seeing the successes here I'm tempted to give it another go...thanks for the inspiration!

Hi, Gambrinus.

I was impatient and reassembled my system yesterday, before I could receive any suggestions from knowledgeable SPCR readers. :-(

I replaced the metal spacers with plastic ones -- so, that should take care of the corrosion. It was stupid of me to try using metal in the first place -- I should have known better than that!

I googled around on the web, and found reports that vinegar might eliminate the algae. I'm sorry to hear you report otherwise. I refilled the water tank (5 gallons) and added 1 cup vinegar, and then poured on the mineral oil cap (about a pint, gives me a 15mm layer of oil). I think maybe the oil is containing the vinegar smell, because I smell nothing. Also, I thought maybe the oil would prevent the algae from receiving the air it needs.

If I see algae form, I'll redo the water tank with bleach instead. Thanks for the suggestion!

About the oil cap -- yes, the insolubility of water in oil is exactly what I was counting on. I didn't know how completely insoluble water and oil are, though. My experiment is revealing that it pretty much completely stops evaporation, for all practical purposes.

About using a radiator -- I considered that, but I favor silence over cooling, and my (limited) experience is that water pumps are still audible. Also, I don't know how well the pumps would move the oil (it is less dense, yet more viscous). And finally, I prefer the simplicity of the static water tank -- no tubes, no electricity, no moving parts to fail. If I do a build from scratch, though, I might play with something more complex.

Interesting idea about using cold oil to supercool without condensation. I like it! You'll have to use a different type of oil, though, because mineral oil thickens at lower temps. Just for kicks, I threw a pint bottle of mineral oil in the freezer, and it looked and felt just like vaseline until it warmed up again. I wonder if it can be mixed with something (affordable and non-toxic) to make it thinner at low temps?

About wicking: I have seen others report issues, and for some reason I'm not having a problem with it. I find a very light film of oil on the power supply cables leading to the mobo, but no pooling as reported by others. I do sometimes wonder if oil is accumulating inside the PSU -- I fear I may someday inspect it and dump a bunch of oil onto my shoes. :-)

The keyboard, mouse, audio, and usb cables appear to be oil-free, possibly because (1) the connectors along the top edge of the mobo are actually sitting ABOVE (not immersed in) the oil, and (2) the cables leading from those connectors lead upward about 8 inches which forces the oil to defy gravity a rather long distance. These two factors may possibly slow down the wicking so much that I cannot perceive any happening. Se my post of (Wed Mar 26, 2008 6:28 pm) to get a picture of that.

Did you use mineral oil in your previous attempts, or some other type of oil?

Edit: Scratched original post because I didn't do my homework.

So isn't it possible to fully passively cool by using a larger aquarium and about 40 liters of mineral oil ? That sounds expensive but doable. You could probably do some tricks to either increase surface area (passive radiator?) or you could change bottom, back, and top sides to metal, making you need less mineral oil.

I think so - but maybe a small aquarium with oil will be sufficient. But I haven't even tried that yet, so it is unproven.

My original motivation for using a double-hull design was to prevent bad oil spills. i was afraid the oil tank might crack or something, so for my first oil system I put the 5-gallon oil aquarium inside an empty 10-gallon aquarium as a fail-safe. That ran way too hot, so I filled the 10-gallon tank with 5 gallons of water, which improved things a great deal: it increased the surface area of our interface to the air, and it increased the thermal mass of the whole system.

A small (5 gallon) tank with only oil will have a relatively small surface area and a relatively small thermal mass. So it seems like it would heat up quickly (small thermal mass) and have trouble transferring heat energy to the air.

But, I'm beginning to believe that the double hull concept is a total bust, because (1) I haven't seen an oil leak after 9 months of operation, so maybe it isn't necessary as a safety precaution, and (2) I suspect the 5-gallon oil tank would more readily transfer its heat to room temperature air than it would to rather warm water.


Another poster gave a convincing argument why the material of the tank is unimportant, provided it is thin enough. I'll try to locate it and give a reference.

Even if the material doesn't matter, which I'll believe in an instant, can't you get a bigger surface area / volume ratio by finning the metal plates? It sounds a lot easier to get a finned metal plate than a finned glass plate.