Chilled oil cooling with mineral oil

[tom89194]

Hi, I've spent a lot of time looking at using mineral oil submersion and was wondering about any personal experiences with oil submersion as well as random (constructive criticism is what is desired) thoughts.

I am planning to submerge my components (minus drives with mobile parts) in a mineral oil bath (5-8 gallons) and pump oil from the top 3 inches of the tank through an enclosure surrounding the evaporator on a window AC unit. The return line will eject the cooled oil (T=0-10C) on or near the CPU heatsink which is a large Zalman (CNPS10X).

-I hope to keep the mean temp of the bath at ambient to begin with (condensation concerns)

-I know oil is messy (swapping 85 gal of oil out of 3-4 tractors (that's 85 gal each) every spring will teach anyone this lesson)

-I have plans to seal the output ports on the back of the MB and cards (silicone)so that my devices don't wick oil through their cords (I refuse to go wireless)

-AC unit is supposed to be capable of moving 1.1 kW of heat and testing of a single pump in the room temp oil with 10ft of 3/4 ID vinyl suggests that the pump will be capable of moving enough oil to transfer 1.1 kW with less than 15C drop on the oil temp after considering that the colder oil will reduce flow rate and increase head losses in the return.

-Only thing I am not sure about is that Tom's Hardware claimed that sealing the cpu from the oil is necessary since it causes capacitance changes between the pins but I theorize that they had a small dielectric breakdown due to the small distance between pins and the fact that they used vegetable oil which has extraneous compounds that I'm certain could cause a small degree of dielectric breakdown. (I do understand the change in capacitance theory, I made water level sensors based on the fact that water replacing air between the plates changes capacitance. They had a nice linear response but got scrambled by my homemade AC solenoid valve in the final assembly :{)

-I would like to know if any of you have done both mineral and veg. oil and if my dielectric breakdown (due to veg. oil's extraneous compounds) theory is plausible. Also maybe those extraneous compounds had some heavy dipoles that resisted when electrical currents changed in the pins (not a capacitance change but an inductance change, crackpot theory???)

-I would have added peltiers to this project but the fact that they can compound the heat rejection by large amounts, i.e. to move 100W at a decent dT you will most likely need to create at least 50W through the peltier and now the hot side must reject 150W (as well as the high current requirements of a peltier). I do understand that higher dTs require larger power additions through the peltier.

I may add more later but this should be most of the thoughts for now.....

 

[Diablo]

A couple of thoughts:

One, this will write off any future use for these components, do this only if you can afford to replace everything.

Ensure the PSU is not in the oil, as the increased voltages could risk a spark, and you really don't want 8 gallons of fire.

Vegetable oil will have all sorts of compounds in, as it is mostly triglycerol based, whereas mineral oil tends to be long heavy hydrocarbons. This is only relevant if the fatty acid head (containing a fair number of polar O atoms) will create a dipole moment and thus conduct. I reckon their use of vegetable oil probably set up a small induced dipole moment when the processor's pins had electricity in them. You would have to check exactly what the oil is made up of, but if it is purely hydrocarbons you should be safe. If it has any lubrication properties, check what these are (graphite will carry charge).

The AC unit will work with phase change, which is pretty efficient for sub ambient cooling, avoid peltiers, as they will need a lot of electricity for the dt, and mean the AC unit then has to work against a higher ambient.

Looking at the TH article, I reckon sealing the socket is a good way to go, as if the oil does get an induced conduction, you are never going to get it clean.

I've not ever tried submersion, but I do have some experience with induced dipoles, and other liquid and solution chemistry and physics. The most important thing is to get the board clean, enusring that any liposoluble components (grease etc.) will not have a charged group (pretty unlikely as liposoluble products are uncharged).

Oh and mineral oil is apparently very corrosive to plastics (capacitors as well

)

 

[tom89194]

Hah, I found something fun:

LN2 and Dice guys often use a silicon dielectric grease to fill the cpu socket to prevent incursion of rouge condensation (Backup protection to the thermal insulation of the socket area to prevent condensation in the first place)

permeability of some materials i looked at today...

silicone e=3.2-4.7

edible oils e=3-3.2

mineral oil e=2.3

air e=1 (not EXACtly 1 but close enough to vacuum for physics class)

Higher e means that pins may have higher capacitance between them.... minuscule changes since capacitance changes linearly with e and the surface areas are very very small, even though distance is also small.

I hypothesize that one may have to seal the CPU socket with vegetable oil but not with mineral oil due to the compounds present in veg. oils (like fatty acids) allowing leak currents when the distance between conductors is very small.

Also I found several published articles on using conductance to measure different types of fats in edible oils.

 

[tom89194]

In response to Diablo

where do you get your corrosion of plastics theory? I have found no real reports pointing toward mineral oil corroding plastics

I do admit that regular capacitors being susceptible to incursion by mineral oil (and thus causing them to burst due to internal pressure or loss of structure caused by chem reaction) sounds plausible, but I've seen many posts using baby oil and the like, which have extraneous compounds that may have been the cause for the cap failures (those extra compounds breached the cap and reacted with the cap materials)

The fact that you posted this leads me to believe that you like regurgitating what you have read without supporting evidence and that you failed to read my entire post....

"The AC unit will work with phase change, which is pretty efficient for sub ambient cooling, avoid peltiers, as they will need a lot of electricity for the dt, and mean the AC unit then has to work against a higher ambient."

-The AC wont work against a higher ambient (unless the room is sealed well enough for 1.1kW to raise the temp of the air significantly) It will, however, need to move much more heat because of the pelts (which I said I wasn't gonna use)

-The AC will not be directly cooling any parts of the computer, it will only cool the circulated oil to sub-ambient temps.

Please read more than half the post....

 

[siravarice]

I remember reading something about not wanting to put the hard drives in the oil because it gets inside and changed the spinning speed of the disk, which slows it down and causes lots of errors, thats all I can add really.

 

[alienware]

I don't see how that's possible unless the oil degrades the seals on the drive.

Hard drives are air tight. Literally like a duck's backend. They're put together in a sterile enviroment as one micron of dust on the platter would ruin the disk.

So if air can't get in I can't see how oil would...

It's kind of like a turbo charger in theory. The back end of the turbo charger is full of oil for lubrication and cooling and the top end is dry with a spinning fan. If that leaked you would end up spraying oil into your engine's intake.

Mind you let's be honest, you don't overclock a hard drive and cooling is pretty simple on them. So there's no reason to submerge them.

/science mode off.

 

[Diablo]

Couple of things:

One, Hard drives aren't air tight, they have little filters in them (google hard drive oil submersion) but rely on air pressure inside and out to keep the head off the disks. An increase of pressure will cause the head to drop a few microns and hit the disk.

Two, Bittech has an article on a guy who did just this (front page), so I spotted the bit about the tubing and capacitors suffering mineral oil damage.

Three, I did read your post, thus noticed that when you said peltiers, these are inneficient, add heat to a system (this case the system being your room), and the AC has to work a little bit harder as ambient is higher. Not much higher admitedly, but 50w will heat a room up (on average a plate glass window 2m^2 will lose 25watts of heat if the delta T between the inside and outside is 30C), thus your room will get warmer. There is also the matter of then cooling the pelt, which often needs a liquid coolant (see Scan's dream PC for example). Also, 1.1KW is what some people call a 1 bar heater, so it will heat your room.

The AC will be cooling oil, which is heated by your computer, unless you have developed a way of beating the 0th law of thermodynamics, this means that the AC is cooling the PC. If you want subambient cooling you will need to put more work into cooling the oil, thus the external system will have heat deposited into it. (s > dq/T, or s= dqrev/T, and this is not reversible).

If you think I didn't do a good enough job answering your questions for free, you can send a cheque in the post and I will endevour to answer your question more fully, rather than say, enjoying hanging out on a forum in my spare times as an activity I enjoy.

 

[tom89194]

Looked at chem resistance for acrylic today...

Mineral oil=resistant

vegetable oil=limited resistance

Maybe that's why some caps went kaboom.... vegetable oil strikes again, or baby oil... maybe (alright read Diablo's more recent post, and the caps failed because a rubber seal at the bottom of the caps was made with a material that absorbs oils and expands, I'm sure nothing was corroded or dissolved)

also, this is from Puget Systems about one year after submerging their pc:

"There is no sign of weakening of rubber seals or PCB. We have found that prolonged exposure to mineral oil does not eat away at any components. However, you will notice in the pictures that the voltage module for the LED light has fallen down. That module was stuck in place with nothing more than a sticker -- it took 9 months for it to come down! We're amazed it stayed up that long, but definitely recommend you do not rely on stickers or tape to fasten anything. Zip ties will be more solid and long lasting."

One of the PCI cards in their system had electrolytic caps..... I'm sure if they had a card blow they would have posted but they also have heavy disclaimers to keep people with burnt computers from suing them (even the ones that bought their custom oil pc system. So maybe they would happen to "overlook" an expansion card with blown caps and "assume" that the failure was "normal" since they want to capitalize on the oil cooling business... I do not know if they would really do this though.

Their oil was a little murky after 1 year but that was probably due to the air bubbling adding dust?? though it could be dissolved materials, the fact that it ran somewhere near 70-80C most of the time should be enough to account that mineral oil is non-reactive with pc components?? (high temps usually accelerate oil penetration and chem reactions)

Also for you people who think cleaning the mineral oil off would be near impossible there is a compound found in spray cans that is used to clean electronics (it also works well against mineral oil) it's called hexane.... I do understand that getting the oil out of the minuscule cracks and crevices will be difficult even with hexane. I also understand that mineral oil WILL dissolve many (if not all) sticker adhesives but I'm sure they cannot be classed with solid plastics in chem reactivity.

side note: It's amazing how soft daily SAE50 oil from a CAT machine will make your hands, even if your calluses are epic. No lotion matches the effect....

 

[Diablo]

I only quoted the bit from bittech because it seemed related. I'm guessing the damage to some plastics may well be due to the pipes bing treated with something to prevent them from becoming stiff, and this is dissolved in the mineral oil. Vegetable oil is a lot different from regular mineral due to the triglyceride based heads.

Hexane will probably work fine, just be aware if you need to clean anything, make sure there is no electricity and no spark possible. It goes up more easily than gasoline (and with heavy mineral oil I'm guessing it would be pretty similar to napalm for burning skin). Having knocked over a beaker of hexane in the labs with a naked flame nearby, it was pretty impressive.

 

[tom89194]

don't worry I've been through ether explosions before, I've caused controlled ones for work and had an uncontrolled one because of a stupid coworker. I've been an amateur firewalker... was real lucky it happened in the winter (carhart coat with a hood). I understand solvents vs open flame....

Well thanks for informing me of the likelihood of the vegetable oil causing the cpu problems... and why sealing would have helped them.

Well maybe here is why the bittech guy's caps failed, they were caps with epdm or natural rubber seals.

Why do I think this...?

epdm or natural rubber will expand mightily when placed in mineral oil

for mineral oil: PVC is rated "B = Good -- Minor Effect, slight corrosion or discoloration." courtesy of the Cole-Parmer Technical library (I vote discoloration)

I'm gonna guess that since guidelines, for the use of Shell Diala AX say specifically, "NO PVC" that the additives in the Shell cause PVC destruction. The guy on bittech got some free cooling oil that cost him the piping on his cooling system. Rigid PVC is mineral oil resistant and should not have shrunk.

Also when something smells sorta like bug spray you are talking about something fishy in the reactive department in my experience.

 

[Diablo]

As you say I guess that the caps were rubber sealed. Were his all solid caps (as I suspect those are at lower risk).

Its wierd because generally mineral oil is designed to be pretty unreactive and act as a (shock and horror) lubricant. Maybe shoot for some oil that has the minimum of other stuff in. I'd also be tempted to see if the brand of pipe affects its oil resistance, to see whether its the oil additive or the pipe being sprayed with something which comes off. Perhaps Tygon (which I use in labs) would work better than generic "watercooling" tube.

 

[tom89194]

Mineral oil isn't designed, It just happens to be made of some hydrocarbon chains left over from the main purpose of oil distillation, fuels and the more common lubricants. (I think)

Anyway, it is used as a laxative and to lubricate lots of machinery used in food preparation. Since it can be consumed there is little worry about small amounts leaking into a large batch of food.

 

[SiN]

I think mineral oil isn't worth all of the hassle, to be honest.

- I mean the oil gets hot after a while, so you can't just leave the system on continuously.

- Takes up a large amount of space for a tank.

- Hard to clean up.

- Better overclocks can be achieved on DICE/LN2.

There are many other reasons, but those are the only ones I can think of off the top of my head.

 

[tom89194]

I think mineral oil isn't worth all of the hassle, to be honest.

- I mean the oil gets hot after a while, so you can't just leave the system on continuously.

- Takes up a large amount of space for a tank.

- Hard to clean up.

- Better overclocks can be achieved on DICE/LN2.

There are many other reasons, but those are the only ones I can think of off the top of my head.

SiN, read the post, the oil will be cooled by an ac system. I am not looking for maxing out an overclock for a short period of time. I want to have permanent stability with high overclocks and that means liquid cooling. The ac should almost certainly be able to keep the oil in the main tank at room temp.

I decided on this because I could (hopefully) have the effect of putting 4-7 water blocks with an enormous radiator (I guess this ac could be used to cool a water based coolant, though those additives that would prevent a freeze would decrease the water's effectiveness). I also want the possibility of sub ambient ability that will keep condensation away (rust, shorts, ect., I know condensation is ~PURE H2O which does not conduct, that is it doesn't conduct until the water picks up ions, which will happen in a few minutes then kablooie)

The bottom line is that I wish for a cold permanent cooling system with GREAT stability and my only real worry with the mineral oil is that some caps (at least maybe some used to be this way) may be sealed at the bottom with a rubber that will expand enormously by absorbing mineral oil. The guy that found this out built an oil cooling system, but he didn't use mineral oil, he used an industrial coolant that was mineral oil with fire suppressants, the additives may have caused the caps to die but I know that the additives are what caused his pvc pipes to shrink and crack.

 

[SiN]

SiN, read the post, the oil will be cooled by an ac system. I am not looking for maxing out an overclock for a short period of time. I want to have permanent stability with high overclocks and that means liquid cooling. The ac should almost certainly be able to keep the oil in the main tank at room temp.

I decided on this because I could (hopefully) have the effect of putting 4-7 water blocks with an enormous radiator (I guess this ac could be used to cool a water based coolant, though those additives that would prevent a freeze would decrease the water's effectiveness). I also want the possibility of sub ambient ability that will keep condensation away (rust, shorts, ect., I know condensation is ~PURE H2O which does not conduct, that is it doesn't conduct until the water picks up ions, which will happen in a few minutes then kablooie)

The bottom line is that I wish for a cold permanent cooling system with GREAT stability and my only real worry with the mineral oil is that some caps (at least maybe some used to be this way) may be sealed at the bottom with a rubber that will expand enormously by absorbing mineral oil. The guy that found this out built an oil cooling system, but he didn't use mineral oil, he used an industrial coolant that was mineral oil with fire suppressants, the additives may have caused the caps to die but I know that the additives are what caused his pvc pipes to shrink and crack.

Sorry mate, I was quite tired yesterday.

I think it would be better to achieve your overclocks with some DICE anyways, seeing as it can be removed like a CPU cooler. Then all you would really need is a DICE pot and and some DICE.

That's just my 2 cents though.

 

[Diablo]

DICE still needs topping up every few minutes, and then there's the issue of getting DICE. Or LN2, they are very cheap, but phoning up BOC and asking won't get you far. I reckon a decent water setup, with the AC cooling water would be easier, cheaper and more reliable, but that's my opinion. Ethan-1,2,-diol will not affect the cooling potential of water by more than a couple of percent, whilst a 10% solution will drop the water freezing temperature by 1.86K/molality (mol per KG of water), giving you a heat capacity of at minimum 3.8KJ/K/Kg.

Beside the latent heat of freezing water is pretty high, and if it is flowing the formation of a seed ice crystal becomes so entropically unfavourable it doesn't happen.

 

[tom89194]

Remember that a radiator must have a dT between its metal surface and the mean outlet water temp for a radiator area less than infinity in order to transfer measurable amounts of heat. Therefore it is easy to freeze an antifreeze on the surface of a condenser (temp near cooling surface will be below mean outlet temp at some point along the surface) if you have a small heat transfer area and if you are trying to cool the antifreeze down to its minimum temp rating. (not really an applicable argument but I felt like saying it)

Interesting info:

Lytron recommends using a 30/70 glycol-water mixture with its recirculating chillers whenever the coolant temperature set point is below 10°C (48°F).



This 30/70 means the antifreeze is protected until about -15C

The main reason for choosing the oil over water is condensation, an oil bath could prevent condensation on a cold antifreeze system. but I don't want to replace the heat sinks with water blocks and I want to cool all of the power regulation and supply components.

I consider condensation unacceptable, as well as the work for the normal measures used to control it (insulation) so an oil bath would be necessary.

(I want a system that is sustainable for long sim runs (days) eating up most of all six cores on a high OC, that means GOOD ALL around component cooling)

If I can move a sufficient quantity of oil with the pumps then the AC can move up to 1.2 kW of heat off of the system and it looks like i could do that with only one $25 pump (I may have to chain another $25 pump in series to take care of head losses though, it looks like it can only do 1.2 psi at the pump when the oil temp on the impeller is 16C)

Also Im pretty sure that any water suspended in the oil well enough to get pumped to the AC might freeze on the evaporator. dT between evaporator and oil may have to be over 5C, though the oil may prevent ice formation and water should settle on the bottom of the main tank.

 

[tom89194]

Well the system is finally up an running, runs nicely around 17C and I get to OC my 1090T fairly high, I hope to get past 4GHz before I give up on tweaking settings.

I'm seeing a massive difference between the core temp on the cpu and the motherboard's cpu temp (It is supposed to be under the cpu socket i believe)

While running prime95 torture test at 18C oil running 3.8GHz 1.45V core I get 34C core/29C MB cpu temp. I tried this setting under good air cooling and the cpu temp ran near 60C (the oil uses the same cpu heatsink but oil is forced through it by a small pump rather than a fan)

 

[CRAZYCAP]

Sound interesting Tom. I've seen vids on the tube but would never do it. Any pics mate??

 

[tinytomlogan]

Pics with your user name on a bit of paper in or it never happened