Sciencemadness Discussion Board

Suggestions for Non-volatile, non-Conductive Liquid for Cooling Electronics

VSEPR_VOID - 3-2-2018 at 17:34

In the past mineral oil has been used to some success for the cooling of electronics for high end gaming PCs and concurrency mining machines. It has the advantage of conducting heat at a greater rate and having a larger thermal capacity than air. With both fluids being able to cycled through the heat sinks of CPUs and GPUs with normal fans. Mineral oil is relatively cheap, non-conductive, is non-corrosive, non-flammable, and has a both a large thermal capacity and conductivity. Mineral oil presents several problems for the electrical engineer: It is difficult to clean up, viscous (putting stress on the fans) and so there is room for improvement.

A replacement or alternative should meet or exceed the following criteria:
Is non-conductive
Is non-corrosive
Has a competitive price
Has a lower or similar viscosity
Is Non-flammable and preferably non-volatile
Is non-viscous and is able to be moved by standard cooling fans
Is thermally conductive
Has preferably a high thermal capacity

Does anyone here have any experience in cooling electronics with liquids? Does anyone know of a liquid with these, or some of these, properties?
I would like to see what crazy ideas we can come up with.

ninhydric1 - 3-2-2018 at 17:46

Isn't water a viable option?

AJKOER - 3-2-2018 at 17:49

Distilled water prepared under an atmosphere of say nitrogen, removing any dissolved oxygen, which could lead to reactive oxygen species.

Meets many of the criteria. For example, pure water doesn't have enough ions to conduct electricity, is cheap, non-corrosive except for highly anodic metals, low viscosity, non-flammable, non-volatile, non-viscous, and thermally conductive.

[Edited on 4-2-2018 by AJKOER]

VSEPR_VOID - 3-2-2018 at 17:55

Quote: Originally posted by AJKOER  
Distilled water prepared under an atmosphere of say nitrogen, removing any dissolved oxygen.

Meets many of the criteria. For example, pure water doesn't have enough ions to conduct electricity, is cheap, non-corrosive except for highly anodic metals, low viscosity, non-flammable, non-volatile, non-viscous, and thermally conductive.



I thought about that. For closed systems that will not have contact with normal atmospheric gases, and not come into contact with sensitive components it would work. I do not think that gamers and miners would or could use it. Having to seal everything against CO@ would be too difficult.

Sulaiman - 3-2-2018 at 18:21

Why the requirement for the fluid being an electrical insulator ?

wg48 - 3-2-2018 at 19:05

I recall seeing a small portable television operating in a tank of what I recall as a silicone oil. It was a display to demonstrate its inertness for cooling electronics.

Here is a pc in mineral oil:


oilcooling.jpg - 29kB

VSEPR_VOID - 3-2-2018 at 21:08

Quote: Originally posted by Sulaiman  
Why the requirement for the fluid being an electrical insulator ?


The fluid is in contact with all the bare electronics. Imagine how a computer would fair is air suddenly was conductive, it would be a light show followed by the release of some magic black smoke.

Melgar - 3-2-2018 at 23:00

Propylene glycol? While it is flammable, its low volatility makes it pretty difficult to ignite, especially accidentally. The most annoying thing about it is that it's sticky if you accidentally spill it, and doesn't readily evaporate.

Silicone oil is good if propylene glycol isn't good enough.

Perfluorodecane is probably like, the gold standard here.

Σldritch - 3-2-2018 at 23:48

Anything fluorinated that heavily will probably have pretty bad specific heat capacity because atomic heat capacity is fairly constant for the elements so the more light elements the better. But perhaps density makes up for it in some cases. Alkanes are probably the best though, you should be able to sacrifice some non volatility for viscosity.

Of course the obvious answer is a NaK cooled computer. Or mercury if you are at risk of spilling your drink on it.

unionised - 4-2-2018 at 01:35

Quote: Originally posted by Σldritch  


Of course the obvious answer is a NaK cooled computer. Or mercury if you are at risk of spilling your drink on it.

Doesn't really meet the "non conductive" requirement

Twospoons - 4-2-2018 at 13:31

In my opinion, total immersion of a PC for liquid cooling is unnecessary, risky, and possibly counterproductive.
Unnecessary : There are only a few places where extreme cooling is actually beneficial, such as the CPU, GPU and memory. The rest of the PC gains no benefit, considering it has been designed to work well with just air cooling (and often passive air cooling at that)

Risky: There are a number of plastic components in a PC - some of these may degrade with long term contact with oils. Don't really want the CPU socket falling apart, do we? Also the excessive loading on fans designed for the speed/torque curve of pushing air - thats going to occur with any liquid, no matter how runny it is, as the liquid density will be close to 1000x that of air.

Counterproductive: PC motherboards contain many traces delivering high speeds signals with harmonics reaching into the GHz. Adding a liquid dielectric to the top of these carefully designed traces is going to change the impedance, loss characteristics and group-delay characteristics of the signal path. You may have noticed traces with squiggly bits : these are for delay matching between traces. This means a real chance of bit errors creeping in, or at least a greater susceptibility to interference, with a corresponding crash.

Much better to use closed loop cooling with liquid heatsinks on the hot spots - then you can use the best liquid of the lot , which is water.

[Edited on 4-2-2018 by Twospoons]

Dr.Bob - 14-2-2018 at 19:14

Just put the whole thing in L2N or liquid helium. That is how they cool quantum computers. I think that would keep them cool enough.

woelen - 15-2-2018 at 00:24

I agree fully with Twospoons. Modern computers work in the GHz range and this requires careful design of signal paths. These signals paths can be quite long (order of magnitude of 10 cm) and very close to each other (0.1 mm or so) through very thin traces. If you immerse such a trace in a liquid, then expect the capacitive coupling between the traces to be quite different from what it is when the traces are separated by air. I expect total failure of the PC when it is immersed in some liquid. The high-frequency behavior simply changes too much. Only the CPU, GPU and possibly the memory need to be cooled.