I feel bad contributing to this thread but I just wanted to comment on a couple things.  Computers in fridges don't work.  First, as has been mentioned, they aren't designed to run constantly.  Second, the compressors in fridges aren't all that powerful, so you it won't be keeping your rig very cool even before the compressor fails.  A/C units or dehumidifiers will work better.

Sorry about taking so long, work got really hectic, and I haven't had the energy.

This will be light on the math since I couldn't find my textbooks.  This is also why i haven't made a diagram; I've been really busy and making one from scratch with no references would probably be less than totally effective.

Think of a radiator as a piece of finned tube (baseboard radiation).  I'm not saying its exactly the same, but similar enough for this purpose.  So, just a piece of pipe with fins on it.  Consider the fluid flowing through the pipe; as the fluid proceeds down the pipe it will lose energy, decreasing in temperature.  Heat transfer is a function of temperature differential.  As the temperature of the fluid decreases, so does the heat transfer, assuming a constant ambient temperature.  Therefore the most effective length of pipe is the very first part.  If you have a really long pipe, the fluid will approach ambient temperature.  Really slow flow would also allow your fluid to reach ambient temperature, but that would require a similarly low flow in your block, allowing the component in question to get much hotter.

Higher flow will increase the effectiveness of the finned tube (or radiator, or block) by increasing the distance the fluid has to travel to cool down.  However, this will serve to decrease your outlet temperature relative to your inlet temperature.  This seems wrong, but isn't.  If you change anything in the system you change everything.  You also don't really care what the temperature difference across the radiator is.  What you care about is the energy transfer, which is constant.  What you care about is the leaving water temperature.  Even though you have reduced the delta T across the radiator, you will have lowered the average temperature of your system, so the temperature at the inlet of your radiator will have decreased as well.

Finally, this is something of an ideal situation.  Having a huge pump that is trying to force way to much fluid through a relatively small tube will have a huge pressure loss. This will effectively cause your pump to heat your fluid.  I'm not saying that you want to boost your flow as much as possible no matter what, I'm saying that your fluid is constantly moving through the system so it doesn't need to take time to cool down.  Your radiator is transferring the most energy when it has the hottest fluid in it.

I've heard it goes rancid and smells bad eventually though.  People definitely do it though.

Correct conclusion, incorrect reasoning.

The idea that the fluid needs time to cool down is common, and is a result of thinking of the heat transfer fluid in individual parts rather than a closed system.  Heat transfer is dependent on your delta T.  Your delta T will decrease across the length of your radiator as a function of fluid velocity.  If your fluid was moving infinitely fast, your radiator would be transferring the maximum amount of heat since it wouldn't have a temperature gradient.  The same would hold true across your water block.

However, I will believe that people have tested much larger pumps then they need and found that they are running higher temperatures with the higher flow.  Really high flow will have really high pressure loss through tubing, radiators, blocks etc...  Your pump has a certain flow rate for a given head loss.  The higher your head the lower your flow.  Think of head loss as friction, which basically increases your fluid temperature.  So the higher your flow the more energy your pump is heating the fluid with.

If this isn't making sense I'll try to put together a diagram or something tonight after I get off work, maybe with some of the math if I'm feeling ambitious and dig out my old textbooks.

I agree, though I don't know about better.  Oil cooling does make it easier to use sub-ambient cooling methods without worrying about condensation.  My gaming rig is hooked up to a home-built chiller.  My idle temperatures can run as low as 18C in the summer.

Only go for oil cooling if its something you're interested in.  Its got lots of disadvantages.  Although I'd probably say the same about liquid cooling.

BTU is a measurement of heat/energy
Watt is a measurement of power (energy/time)
A BTU is similar to a joule, a BTU/hour is similar to a watt.

When an AC unit says 5000 BTU it means that the unit will remove 5000BTU/hour from a space, though this depends on the temperature difference.  It is not a measurement of how much power the unit is drawing.  That would depend on the specific unit you are looking at.

I tried using an older cpu with no heatsink, and got hot fast (I shut the computer off as it climbed past 90C).  You need something to increase the heat-transfer surface area.  With the same cpu I have had good success using a low-profile passive heat sink.

I haven't had any problems with fans in the oil.  My guess is that the increased resistance is offset by the increased lubrication and cooling provided by the oil.  The video card fans seem to be pretty effective at moving oil through the heat sink.  The squirrel cage fans they typically have aren't much different in concept to an impeller pump.

If you're mining with even one decent card you'll want to use a radiator regardless.  The heat dissipation through a flat surface (container wall) just isn't that effective.  There's a reason every piece of computer hardware uses fins.  If you use a glass aquarium than you will dissipate some heat.  If you are using an acrylic aquarium or plastic tub then you will lose less heat through the walls.  Having a pump attached will help circulate the oil which is also a good thing; relying on natural convection for your fluid circulation will raise the temperatures.  If you're using multiple cards than you will certainly need additional cooling.

Both my mining and gaming rig are in oil, but I would recommend against it unless you have some idea what you are getting into, and you enjoy that sort of thing.  Puget systems does have a good kit and explanation, which makes it much easier.  I tried their kit some time ago, and decided that the aquarium they were using was a bit too fragile.  However, the oil is not cheap (~$20/gallon), is very messy, almost certainly voids warranties, and still requires additional cooling equipment.  Also, your hardware is difficult to get to, and your computer becomes damn near unmovable.

If this is the kind of thing you enjoy doing, then by all means, do it.  But don't get into it thinking its going to be cheap and easy.

I've been playing with oil immersion for a couple years, and its definitely not cheap.  I bought all my oil from a local feed store.  They stock the stuff as a horse (and other livestock) laxative.  The cheapest I've found it was $20/gallon.  Plus you will need pumps, radiators, heat exchangers, and fans.  Also, as pointed out by someone else, Oil makes a huge mess.  My hardware is basically inaccessible unless i have a full afternoon to play around with it.  Finally, plunging your equipment into mineral oil is a pretty good way to void the warranty, so if it breaks its your problem.

My gaming rig right now has a pump and a heat exchanger, connected to chilled water with a water chiller in the garage.  I can keep everything under 60C even while mining.  My mining rig is in a separate tank with just a pump and a radiator+fan.

Distilled water doesn't conduct electricity.  However, distilled water will not stay distilled for very long, especially if you dunk stuff with metal pieces into it.  I saw a forum post once where someone tried it with an old computer to see what happened.  If I'm remembering correctly it only took a few minutes before the computer shorted out.

I keep the cooling for each tank separate.  If they were connected it would be somewhat easy for the pumps to be slightly unbalanced and overfill one tank while emptying another.  You could install float switches and additional pumps to keep this from happening, but that would increase the cost and complexity.

The reason I do it?  I can keep my cards 50-55C while mining, lower for gaming. It is almost completely silent in my living room.  This kind of thing isn't really practical for mining rigs, unless you already have the equipment and oil, and you enjoy playing with it.  Oil immersion is a hobby.

Its been posted before, but its very hard to make accurate comparisons between supercomputers and bitcoin mining.  They really aren't doing the same things.

I would be somewhat interested in this as I have several Nvidia cards which aren't really useful for mining.  I would also like to see more information about it.

Watch the prices... I'm hoping for a drop, then buy.

I've been mining for about almost 2 weeks.  So far I'm doing about 110Mhash/s each on my two gtx470's.  I have a 6870 on the way to slap into an older rig, and I'll see how that works.

Personally, I'm running a mineral oil computer with an attached chiller, and I'm maintaining about 55C on my GTX 470's (its a gaming rig, only started on bitcoins recently) while mining.  I got into computer cooling as a hobby several years ago, and have played with liquid cooling, mineral oil, and mechanical chilling.  From what I understand, every 10C over 50C you roughly halve the lifetime of the component in question, though I don't know what the normal expected life of a well cooled card is.

If you're maintaining 90F on a card while mining, I'd be astonished.  I don't know how you'd manage that without some serious phase-change equipment.

My impression from the forums is that you would do better with an ATI gpu rather than NVIDIA.  Also, if you are planning on running multiple gpu's you will probably want a larger power supply.