LIQUID SUBMERSION - a few helpful learned facts.

[4voc]

I completely submerged a Asus P5N32SLI w/Pent D Processor, Dual 7800gt Vid Cards, 4MB Corsair XMS Ram in Light Pure Silicone Oil. The oil was recirculated through a fan cooled radiator.
Did it work? Yes and No.
The good: Electrically Perfect Operation, Immediate Booting to windows, Great Overall Cooling. Looked Bitch'n.
The bad: Regardless of the circulation and Overall Fantastic Cooling,,, I discovered a small issue never talked about in the COOKING OIL, AND VARIOUS OTHER Submerged computer projects. This small item basically prevents liquid submersion from being viable with all of the experiments/projects I have seen.


Any (ANY!!!!!) practical liquid INCLUDING DIONIZED WATER. Silicones, PTFE, Glycols, etc, through capillary action, surface tension effects, etc. will always flow into the tiny spaces, pins, etc. in the board and system components and will NOT circulate away rapidly enough, regardless of flow rates, and will create tiny spots of INSULATED, NOT COOLED, areas that will cause localized overheating on your system components. This took awhile to determine why my processor would overheat when the surface of it and surrounding areas were 70F but the internal temps would ramp to 200F.

Solutions:
1. Supercooling- Unless you want to use Liquid Nitrogen or an equivalent to supercool the liquid to about 0F throughout the bath, you will not get the heat to dissipitate away from these areas faster then air can. If interested I can go through the physics calculations with someone who is VERY interested. The waste of energy/electricity to run a strong compressor or using compressed N2 or other defeats the purposes of going with liquid submersion to begin with.
2. Sealing all of the tiny spaces into the boards connection pins, and all of the associated system components. Good luck, you must use a dielectric, themally conductive material, and even if fully successful, you know have locked in the air that would normally circulate out of those spaces as it heats and expands. Sealed air is still an insulator. BAD IDEA.
3 Submersion of only the back side of the board. Sealing all of the small spaces through the motherboard with a thermally conductive, Dielectric sealant. This was my original plan and it works better overall then straight air cooling and water blocks. Unfortunately, it is time consuming and costs a few bucks to custom and thoroughly seal the back side of the board from the front compared to the amount of cooling you gain. I'ld really only recommend this to someone who really has the time and money and really is obsessed with jacking that extra few mhz from their system.

FYI: I am a chemist & chemical engineer. I have my own fab. shop and hav e lots of spare rare materials to work with.
I'm writing this general note for those who are looking to take on this fun project to let them know of an inherent flaw (I learned) in Liquid Submersion. I wrote this quickly so please don't bash for spelling errors, etc. I just thought I would contribute to those who are nice enough to contribute themselves.

 

[wun911]

I have heard of pplz doin this with double distilled H2O, and a fan to circulate the water around the submersed MB.

Also I have heard of pplz cooling the liquid with water cooling. ie the water is pumped out of your box and though a rad and back into the box again. The pump's inlet was placed over the CPU and the CPU did not have a HSF it was just exposed bare.

 

[yuseaname]

I have seen some instances where people have used a liquid developed by 3M, called Flourinert. It stays liquid at very cold temperatures, is not electrically conductive, and best of all, is very very expensive (~$500USD per gallon, don't flame me if i wrong 😱 ). There was a site i saw that did a review using it as a coolant. There was a chamber of liquid nitrogen, which acted as a heat exchanger cause a pipe carrying the flourinert was run through it. The liquid was deposited into another chamber, which was filled half full of flourinert and the motherboard/ cpu and stuff was fully submerged. I will post again if i find that site.

 

[tool_462]

Sorry if I missed it in the original post but, did everything come out undamaged? I remember reading a few times when the solution gains a few ions and starts conducting, ruining CPUs/Mobos and so on...

 

[440bx]

This may be a dumb idea but if I were you I'd give it a shot.

I imagine that you submerged the motherboard horizontally flat which I can see causing the problem you mentioned.

Just to convince myself that it doesn't work, I'd try submerging the motherboard vertically thus gaining the help of gravity to move the liquid in those tiny areas that normal flow doesn't affect. Of course this will necessitate a differently shaped container or a greater amount of liquid.

It's a long shot but I don't give up easily.

HTH.

 

[4voc]

4VOC answers and comments:

Flourinert and Transformer Quality Pure Silicone (both used for submerging industrial electrical components) are similar in viscosity, temperature behavior, etc. The silicone is however a bit cheaper and is not likely to generate any HF (hydroflouric acid) if it gets overheated. (ie by getting into a space it cannot escape from, an electrical failure, arc, etc.)

The board and all components all came out completely fine and perfectectly functional. Silicone (when pure) is COMPLETELY inert (inert means it cannot and does not react with any of the pastics, ceramics, metals, etc.) The only thing silicone can do is absorb very slightly into any silicone o-rings (sometimes on liquid pumps) but,, it does not destroy or attack them, it just will slightly swell them a bit (the unreactive chain lengths of the silicone oil and the O-RINGS are drastically different and therefore do not chemically effect each other). This actually can be a good thing here because it helped the seals. A bit of DI (deionized water) and light citrus degreaser soak followed by a rinse with 70% isopropyl and air dry cleaned off the silicone residue pretty easily and thoroughly.

I did Recycle the Liquid using two pumps and two radiators (For recirculating and thorough cooling overkill) Keeping the liquid cool was not a problem at all nor was keeping the outside of the components cool. The problem was you can't circulate the tiny spaces behind the pins of the CPU and other like components and THAT"S where there is a problem with submersion and that's why it works great for TRANSFORMERS and LARGE electrical devices but really that practical for microchips. Flourinert, Light Silicone, DI Water are all two viscous too move thoroughly in the TINY spaces around the cpu, etc. That's where the heat is retained and that is where you get the problem.

I've read numerous posts on the various liquid cooled systems.
-DI water is just too short term and will , ALWAYS will pick up dust, particles etc. and no longer be pure after a short period and will cause issues, Unless you make it in a class 10 chip room, its a waste of time.
-Flourinert and Silicone act the same and are Almost interchangeable and cause the micro localized hot spots that I have discussed. Cooling with liquid nitrogen is not practical except for a quick fun but VERY short experiment.
-Cooking oil is just silly as it breaks down, allows for bacterial growth, smells, and will slowely react with some of the bonding materials, etc. Additionally, it does not prevent the same problem of localized heat spots behind the chips.

Another thing I did not mention, I even put a SEPARATE WATER BLOCK with isolated coolant flow to the CPU and resubmersed the board in circulated and cooled silicone. This had little or no change. I did several other quick experiments and concluded that with any liquid access to behind the chips it will build up heat in these tiny spots which will effect the chips performance negatively.

 

[mitch074]

a simple solution would be to look for a mobo that has drilled holes right under the CPU (inside the socket). If my memory serves me well, there used to be some like these at one time. Not anymore?

 

[4voc]

A Few Answers:
1. Refrigerator: Yes you can do this but, without a pump, and either metal walls or a radiator, you will just have the same problem (albeit less). The liquid will not move away quick enough from the processor and chips.

NOTE: The goal here was not to SEE if it could be Done but to See if it could be done easily, cheaply, in a clean setup that does not require MORE energy but LESS energy to run. Unfortunately, a refrigerator or compressor (which is what's in a refrigerator) uses too much electicity and defeats the purpose. The theory is sound and tested to be good here. The DOD, the EVERY ELECTRICAL TRANSFORMER YOU SEE and many other companies use this technology all the time. The problem is to do it cost effectively and practically with the way PC chips are made leads to this problem with tiny localized insulation spots behind the chips.

2. There was no and would be no moisture issues as long as you have a reservoir or case setup that allows any condensation to accumulate away from the process.

3. I tried this setup horizontally, vertically ATX Style and ATX 90deg. No difference in either of these setups. All had great overall cooling but tiny hotspots formed behind the chips. This all supports my conclusion of the problem.

4. The Socket 775 has not openings behind the chip and is litterally saturated with tiny pins. There would be no way to realistically vent this spot. Therefore, there would be no realistic way to get flow back there unless Intel redesigned the socket or I custom fabbed a socket setup which would.... umm take alot of time and would be very likely to open up a another set of issues. Venting the board itself would not help much as redesigning an entire board to remove these tiny spaces would kinda defeat the purpose of this project. If someone wanted to grant me a couple mil. the patent rights and a circuits mfg shop I think I could do this easy enough.


4. I hope I have some pics. Camera is with one of my work crews so I'll try to check. I was originally planning on writing an article for my friends online magazine on the finished product with a how to guide and clocking Specs. As I did not forsee this issue, I never finished the fancy dressed up product and I stopped documenting the test rigs. I already have taken out the board and put it in an office computer while I wait for my new gaming board and core2 chip to arrive.
Don't really like the ASUS P5N32 board as it is junk and the company released about 20 versions of it under the same name and false advertised and covered it's problems. IE: Their Website Stated it was CORE2 Compatable up until a month ago well after they released the SE version (unnamed originally "SE") without telling customers their was a difference and without accepting returns or upgrades to all that bought the original versions thinking they could run Core2. The PentD series is already outdated by the Core2 and uses WAY TOO MUCH Power. If you try to run a Pent D and with SLI, Forget about running stable with a 500-550W power supply. It WILL run, but you will have voltage issues under stress.

 

[GrimReaperGuy]

Forgive me if I'm wrong here, but working from memory, isn't Silicone Oil a terribly poor thermal conductor?

 

[MadHacker]

I beleive water is also a terrible thermal conductor...
but water can abosrb heat... to get water cooling to work best you have to get each water molicue to hit the heated side of a water block(making the storm jet impingment so good).

but not sure about Silicone oil.

Please correct me if i'm wrong...

 

[praeses]

Someone I know of tried a similar thing in the past, but although with a fairly disposable machine. His solution (although undoubtedly not the first to do it) was similar to prepping for a peltier. Basically he siliconed the backside of his north/south bridge, cpu, (v/g)pu, agp/pci slots and ram, and shoved some dieelectric grease down the cpu socket pin holes. Anyways, it worked, but being a much older machine (p3-800) it is probably much more leniant than your machine. Oh, he used thin over the counter engine oil.

 

[4voc]

The thermal conductivity of DI water is roughly 3-6x better then Light Silicone oil, Vegetable/cooking Oils, or Ethylene Glycol (coolant in water block systems and cars), PTFE ... 3x6 roughly based upon mixture (ie w/water). The viscosity (in laymans terms) is the measurement of how well it flows. I used a very low viscosity liquid so it would flow like water.

Now 3-6x worse then water seems bad, but that is why you use active forced circulation. Ie. your water block systems use a pump and 300gph flow per chipset. (roughly) This more then makes up for this inadequacy.
Additionally, I tried DI water, vs. pure silicone, vs. ethylene glycol/water mix through a seperate water block with and without silicone submersion.

Without Any Submersion: All the liquids more then adequately cooled the Processor steadily at about 35-37deg. C.

With Submersion: No water block: 90deg-100plus C and rising: Shutoff system.

With Submersion & water block: 90-100plus C and rising: Shutoff System.

With All the Submersion scenarios, The overall temperature of the Surface of the Chips was fantastic and the heat was pulled away and cooled by the radiator. BUT... In all cases of Submersion the internal Cores of the chips would build up slowely getting hotter and hotter while the exposed surface would stay very cool. As it was easy to see that the heat was being pulled off the surface of the chips (where a water block or cooling block would sit) The only conclusion was that the liquid was not able to circulate out from behind the chips (which in retrospect makes sense as these spaces are very very tiny and will and would always pull in fluid through capillary action/surface tension and not willingly release these fluids). Air However will circulate in these areas and in fact expand out and release from these areas at an increased rate as the area heats up.

 

[Nitro350Z]

I'm sorry if this has been mentioned, but maybe try pointing a few powerheads/aquarium pumps towards components that overheat and maybe you can force some of the liquid to circulate around these insulated spots? or maybe having a tiny space between the motherboard and socket with a pump pointed towards it to push the fluid inbetween the pins?(if that would work at all, it might have been easier with older sockets like s478)

Just my $0.02

 

[Enrubi]

There are basically two things to think about with heat: specific heat capacity (the amount of energy needed to raise the temperature 1 degree) and heat conductivity....

Metals (in general) have great heat conductivity but almost no heat capacity,
water on the other hand has great (really really high) heat capacity but bad heat conductivity.

This means that water can absorb alot of heat (but it also stays warm) and if you also circulate the water you can cool it off somewhere else.

From the top of my head, I think that water has one of the highest heat capacities of any liquid (seems ammonia is a little better according to wikipedia 😉 ).

For heat conductivity, diamond wins the race but an order of magnitude... I've always wanted to grow a diamond heatsink.

 

[sirheck]

Any (ANY!!!!!) practical liquid INCLUDING DIONIZED WATER. Silicones, PTFE, Glycols, etc, through capillary action, surface tension effects, etc. will always flow into the tiny spaces, pins, etc. in the board and system components and will NOT circulate away rapidly enough, regardless of flow rates, and will create tiny spots of INSULATED, NOT COOLED, areas that will cause localized overheating on your system components. This took awhile to determine why my processor would overheat when the surface of it and surrounding areas were 70F but the internal temps would ramp to 200F

so here you are talking air bubbles right?
if so i could see how the that would cuase problems.

there would be, maybe thousonds of them.

 

[4voc]

Sirheck, Not air bubbles. Go take a straw, put it in a glass of water and watch the water move up the straw above the water level, the less the diameter of the straw, the more it will move up. I'm trying to explain this the easiest way I can. I only tought chemistry one semester.
The spaces fill with the liquid after a bit of run but are just TOO small for the liquid to move out of quickly enough. Go pop out your Socket 775 Chip and look how tight the spaces are between the pins behind the chip. This is also the case on the vid chips, ram, etc. Liquid goes in,,, liquid does not want to come out. The air bubbles (if any) disappear quick when the chips heat up, the air expands, bubbles away and the liquid fills in.

Nitro350Z, I did put the inlets of the pumps pointed right at the chips, I even tried not submerging, just spraying the chips. I had several radiators and 4 different pumps and countless configurations i laid out and swapped about to try to solve this,,, I spent a few 5am sessions at the shop obsessively trying to prove that this could be done easily. Oh Enrubi, and they do grow some pretty big synthetic diamonds now.... but... they're still pretty dam expensive.

Older Chips with less pins and that do not run so hot (like the Pent D series) would definately be easier. Putting dielectric grease is possible but, really wouldn't help very well unless you used a Potable dielectric /themally conductive plastic or ceramic (which are available now but this again would be extremely time consuming and add to the cost, plus with the imperciseness of this kind of molding in conjunction with the sensitive areas you are working, I really really would not attempt this with nice stuff unless you have a circuit board mfg facility.

 

[Enrubi]

Any (ANY!!!!!) practical liquid INCLUDING DIONIZED WATER. Silicones, PTFE, Glycols, etc, through capillary action, surface tension effects, etc. will always flow into the tiny spaces, pins, etc. in the board and system components and will NOT circulate away rapidly enough, regardless of flow rates, and will create tiny spots of INSULATED, NOT COOLED, areas that will cause localized overheating on your system components. This took awhile to determine why my processor would overheat when the surface of it and surrounding areas were 70F but the internal temps would ramp to 200F

so here you are talking air bubbles right?
if so i could see how the that would cuase problems.

there would be, maybe thousonds of them.

I think it's more about the fact that the speed with which a liquid moves in say a tube depends on the distance from the walls. The speed is the highest in the centre and very very low along the walls because of surface tension etc.

A high viscosity liquid (like an oil) flowing in a complex geometry will most probably form "dead zones" where there's next to no flow at all.

Air can be considered as a low viscosity fluid btw.


Growing synthetic diamond isnt that hard: 1 oven, 1-1.5% methane gas, the rest argon, perhaps a few tiny seed diamonds, time 😉

 

[sirheck]

Sirheck, Not air bubbles. Go take a straw, put it in a glass of water and watch the water move up the straw above the water level, the less the diameter of the straw, the more it will move up. I'm trying to explain this the easiest way I can. I only tought chemistry one semester.
The spaces fill with the liquid after a bit of run but are just TOO small for the liquid to move out of quickly enough. Go pop out your Socket 775 Chip and look how tight the spaces are between the pins behind the chip. This is also the case on the vid chips, ram, etc. Liquid goes in,,, liquid does not want to come out. The air bubbles (if any) disappear quick when the chips heat up, the air expands, bubbles away and the liquid fills in.

ok cool just wondering how you got all air out of the system.
i was thinking when air is heated it expands?

 

[sirheck]

aahh yes similar to bernoulli,s law. ok

 

[cb62fcni]

Fluoroinert, man it's been a while since I've heard that! We used it for cooling solid-state power frequency converters on my ship, the 60-400Hz converters that make power for our "delicate" systems. FC-72 to be exact. FC-72 has roughly the viscosity of water, but it evaporates immediately upon exposure to air so you'd have to have an air-tight seal or it'd be gone in a day. Another big problem with it is that at ~200C it will form hydrogen fluoride, so hot spots could be a big problem. Also, it's several hundred bucks a gallon (we get it for around 300-400 it changes). Interesting idea though, I've often thought of using it in a closed-loop water-cooling setup, but it really has a way of escaping closed-loop systems, and it's too damn expensive and I'm not about to defraud the government cuz it's not that much greater at heat transfer than H20.

 

[cb62fcni]

Oh, and a funny fact, if you drink FC-72 you piss it out almost instantly. Good times.

 

[Rripperr]

This just goes to show why PCC (Phase Change Cooling) is the only pratical (but expensive) way to achieve super-cooled states. Remember, your not trying to 'put cool in', you have to transfer heat out. Simply 'dunking' something in a cool substance might cool it temporarily, but once the substance you're using reaches a certain state, you usually wind up 'heating up' the object you're trying to cool, no matter how much circulation you might have in a confined space. You need to have a radiator, or some form of heat exchange, to remove the excess heat from the substance. Otherwise the heat is simply recirculated into the system, and cooling drops off at a rate determined by the specific heat of the substance.

For maximum cooling, you need a substance with 3 requirents:

1. Low to Meduim Viscosity of coolant substance to surface area:
It's important to remember that any cooling has to reach ALL areas. If the substance has too high of a viscosity in comparison to the flow path, you will get 'dead zones' of little to no circulation, and therefore little to no cooling. This is dicttated by geometry, as well as the physical properties of both the coolant, and the object being cooled. A flat wide area will have very different needs than a crowded area with lots of irregular shapes in the flow path. The smaller and more irregular the flow path, the lower the viscosity needs to provide adequate heat exchange.

2.Defined circulation pathways:
Any cooling substance, whether liquid or gaseous, needs a well defined 'flow path' to ensure proper circulation and heat transferance in all the nessesary areas. Simply dumping something in a tank with a circulation pump does no good. Ideally, the flow should start at the hottest spot and flow past the lesser heat areas away until it is gathered and sent to the radiator/heat exchange. Once there, it is circulated to release it's heat, then it is re-introduced in to the cooling system.

3.Thermal exchange:
Thermal exchange is governed by the values of a substance in 3 different areas: specific heat (a measure of how many units of heat it takes to raise the substances temp/volume), thermal conductivity (how quickly/easily a substance absorbs or releases heat), and finally, latent heat (or vaporization)which measures the heat gained or released during a change of phase (from gas to liquid, liquid to solid, etc.). Heat tranferance is most efficiently accomplished by the phase change, since larger volumes of heat are absorbsd or released during phase changes. Phase changes can manipulated with the use of pressure. Low pressures lower the vaporization point (so you can boil or condense a substance at a lower temprature), while higher pressures raise the same (to absorb more heat by avoiding the 'boiling point', achieving 'super-saturation').

OK, now I can fully explain how it works;
1. High pressure/low temp. liquid is sent to the cooling area (evaporator). The liquid enters an 'expanded' area, where the expansion allows the liquid to 'boil' off into gas, and absorb lots of heat in the process.
2. The low-pressure/low temp. gas is sent to a compressor, where it is turned into a high pressure/ high-temp. gas.
3. The high pressure/high temp. gas is pumped into a radiator, where it is allowed to slowly expand (as it works its way through) and release its heat (through convection), returning to liquid form.
4. The low-preesure/ low temp. liquid is sent to a Condenser. The condenser re-pressurizes the liquid, which is then sent back to the evaporator.

With out some phase change, you can't get decent cooling in a small closed system.

 

[4voc]

good overview and summary of what we've been talking about... but ...

If you read the original or follow up posts.... the overall submersion theory isn't a bad one and if done right transfers the heat away great...
the coolant was flowed directly to point of heat generation, and pulled away and recirculated to a radiator where it was cooled to ambient temperatures. even additional direct water blocks, pumps and radiators where added with no effect. it was extremely easy to keep the liquid cool, the case cool, the board cool, the surface of the chips cool.

What could not be cooled was the tiny spaces (in many cases less then the width of a hair) that cannot be practically circulated away.

one of my points i'm getting at is... if you DO want to waste money on a condenser unit and DO want to waste another 1000kw of electrical usage to run it so that you can drop your coolant down to any sub-ambient level and so you can jack an extra bit of mhz your STILL BETTER OFF using water blocks instead of submersion.

FYI, doing this would have defeated one of my goals. To make a practical way to create a quiet, low energy, cheaper way to provide above average cooling. Cheaper because a sealed case is cheaper then water blocks for the twin vid cards, CPU, RAM, North Bridge, South Bridge, and the circuits inbetween that begin to get hot when you start to seriously overclock the major components.

submersion still is bitch'n looking though.

 

[john_thor]

Very interesting! 8)

 

[rcs2749]

Maybe I missed it but what happens if it is submerged with the heatsink in place?