Mineral Oil as Thermal Grease

I've never heard of replacing thermal grease. Silicone grease doesn't evaporate. The grease in very old products seems to still be greasy.

How would "dried out" affect thermal conductivity?

A key spec is "bond line thickness", how thin a filled grease will squash down. A liquid will squish to zero. Fillers add minimum thickness.

If the surfaces are very flat, fillers make things worse.

Dow 340 specifies 100 microns, but I measure much less.

Well, there is one place where the Imperial system is superior to both Metric and US measures:

An Imperial Quart (of beer) is 1.137 liters, and the US Quart is 0.946 liters. Rule Britannia!

Joe Gwinn

What is the price difference?

Oil lamps also burn their wicks, and catalyze any number of chemical changes...

Oh, but kerosene IS a mineral oil. It's a mixture, probably with lots of heavier molecular weight stuff (like the 'mineral oil' fraction you note).

Not a full picture; force is only applied at the close-contact region within the bearing, and it forces oil to other regions, also within the bearing structure. There, it's retained because of capillary attraction.

Commercial thermal compounds are various, not 'conventional' recipes. Why would you recommend replacement? There's no such recommendation from computer builders, unless and until one removes the heatsink. At most, there's a fastener torque specification and a requirement to remove old-goo residue before reapplying the exact same goo. "Dries out" probably means becomes non-fluid, so it won't spread on reapplication of the heatsink, but that does not imply any fault in the heatsink properties if undisturbed. You're removing the old stuff so the new application can squash down to suitable thickness, just like the old application.

Try putting kerosine on a baby's bottom and see how he likes it.

Try running your engine without oil.

Real commercial thermal paste also dries out and it must be periodically re-applied.

The lifespan of thermal paste varies depending on the manufacturer and typically ranges from 3 to 5 years.[3]

Thermal past contains fillers that keeps the surfaces separated. This reduces heat transfer.

Mineral oil squishes down to zero thickness, giving excellent heat transfer due to metal-to-metal contact betwen the cpu and the heatsink.

The only purpose for the oil is to fill the voids that prevent metal-to- metal contact and prevent them from being filled with air.

most engine bearings are fluid bearings that rely on a continues supply of pressurized oil to maintain an oil film

rolling element bearing are a lot more forgiving

Bolt the little beggar to a heat sink and he'll like it even less.

I read about Sony using a low melting point metal alloy as a thermal compound in some gaming machine.

Jeroen Belleman

There is a product called Liquid Metal. It contains gallium that is corrosive to aluminum and cannot be used for heatsinks.

Quote:

"If we add liquid metal, you’ll see that, over the period of a night, it becomes chalky and blackened, and embrittles the aluminum coldplate. This is because aluminum is highly soluble in gallium and will form an alloy with the gallium. Typically, aluminum will build an oxide layer around it, which protects the aluminum from exposure to other elements. Gallium penetrates the oxide layer that forms around aluminum, causing the gallium and aluminum to form an alloy."

See photos at

The best approach is to ditch all these substances and rely on pure metal- to-metal contact, with mineral oil to fill the voids.

Galinstan, need copper heatsink or heat spreader since it eats aluminium

you can also get thin indium foil, soft enough to fill voids

You continue to tout mineral oil, even saying it lasts for years, however you have not provided any support for that statement. Do you have any evidence that mineral oil will last?

Yes, I've used indium in fact; the (gallistan?) 'liquid metal' is impractical against aluminum, but heatsinks don't have to be aluminum (a copper slug is sometimes inserted, for instance).

The indium solution works well in cryostats using liquid helium, because the heat conductivity of materials goes to zero close to absolute zero temperature; it's not economic, wouldn't be suitable at room temperature.

I have been researching measuring the cpu heat spreader flatness. I found

File #5 shows the effect of clamping the cpu in the socket. The wings press down on the cpu and distort it.

The measured profile shows a deviation of 0.0466-0.0224 = 0.0242 mm, or

I have been trying to find what effect this would have on thermal compounds, including mineral oil, but have had no luck.

The solution is to simply monitor the cpu temperature. If it is satisfactory under load, then the issue is solved. My previous measurements showed this was the case.

Here are the links:

Intel IHS CPU Heatspreader Flatness

While your measurements of a CPU without a heat sink show the device can be warped by the pressure of socket clamp, that is most likely irrelevant since the pressure of the heat sink will tend to warp the CPU mating surface back to a flat plane or possibly warp the heat sink into a matching shape, or more likely a combination of both resulting in a much smaller gap for heat sink thermal material.

I would also point out that the temperature of the components involved likely impact these measurements as well.

These are the sort of theoretical measurements that result in thermal measurements of CPU temperature being the only way to measure the effectiveness of CPU cooling techniques.

Search for "Our thermal epoxy versus store bought" on youtube. All contributions of "tech ingredients" are amazing.

He has concocted the follow thermal paste (it sounds like alchemy) epoxy dendritic copper graphene diamond powder

Because some components are hard to get, you can buy it from him. (dendritic copper is snow flake like copper that result from electrolitic deposition.)

The half our video demonstrates his paste against commercial products.

He discusses many of the subjects in this thread in detail. So you will learn not use gallium (which sounds nice because it melts and flows) and that indium sheet is a good choice, provide you apply sufficent pressure (5000 N for a typical cpu), which is way more specific than the Sony messages.

It is truely amazing the expertise and the sheer amount of technical demonstrations of this youtube channel. (Other subjects are stiffness of loadspeakers cabinets, noise isolation properties of helium, super magnets)

Groetjes Albert

We had an assembly where we epoxied a small part to a copper heat spreader. Both were very flat.

Diamond-filled epoxy was worse than plain epoxy. The diamond grains increased the gap.

You can buy tiny glass balls to mix into epoxy to increase gap breakdown voltage, but it wrecks thermal conductivity. The ball-powder is a high grade of the stuff they use to make paint retroreflective.

Of course you run the risk of delamination if the bond line is too thin. That's the other reason for the glass beads.

Electrically-conductive epoxy is mostly made that way, by silver-plating the beads first. The really cheap stuff uses graphite-coated glass beads.

Cheers

Phil Hobbs

I have found the solution.

The equation is simple: k = Q * L / (A * (T2-T1)), but the calculation is a bit more difficult. I used Roger Schlafly's Mercury, and got the following results. T2 is the cpu temperature, assuming the heatsink is infinite.

I used the LGA1155/1156 heat spreader dimensions and assumed a power dissipation of 65W. The distance L between the cpu and heat sink is

The results show that anything is better than air, and there is little difference between Liquid Metal and the best thermal compound.

Mineral Oil is the worst of the thermal compounds, but even in this application it provides excellent results, with a temperature rise of only

Even when the power dissipation is increased to 130W, the temperature rise is only 48 - 25 = 23 degrees.

With a cpu temperature of only 48 degrees, who needs thermal paste?

Best Thermal Paste K = 12.5 T2 = 25.125

Worst Thermal Paste k = 0.5 T2 = 28.132

Mineral Oil k = 0.136 T2 = 36.516

Air k = 0.0262 T2 = 84.780

---------------------------------------------------------------------- Mercury Calculations

;CPU Heat Transfer

;Roger Schlafly's Mercury is available at ;

; Where:

; k = thermal conductivity ; Q = heat flow (W) ; L = length or thickness of the material (m) ; A = surface area of material (m^2) ; T2-T1 = temperature gradient (K) ; h = height of heat spreader ; w = width of heat spreader

; Typical Conductivities ; Air, atmosphere : 0.0262 ; Liquid Metal : 73 ; Oil, transformer : 0.136 ; thermal paste : 0.5 to 12.5 ; Water, Fresh : 0.609

;

; k = 0.0262 k = 0.136 ; k = 12.5 ; k = 73 ; k = 0.5