Your experience with thermal compounds silver/ceramic vs generic white

[Miyuki]

Hi folks
I want to ask if you have experience with some fancy thermal compounds
In cases when I don't need insulation, like have ISOPLUS 247 package or on an external ceramic insulator when I want to cool huge powers like 100-200W

With "common" grease, the table value of the thermal resistance case to heatsink is about 0.25 K/W
I know on a bare die to heatsink it can make a huge difference with better compound

[bobbydazzler]

What about thermal compound designed for desktop cpus?  It handles 100's of watts fine(maybe it doesn't age as well?).

[Whales]

I have discovered that my cheap greases dry out after a year or so.  If you want your product to last: might be worth seeing if thermal greases have datasheets specifying lifetime.

Random notes:

* If you can mechanically lap the two surfaces well enough: thermal grease becomes optional or worse of an option than direct contact.
* There are super-thin thermal sheet like products available ("graphene" or something?) that could be worth considering.

[sandalcandal]

Towards desktop CPUs and "pc enthusiasts", the highest performance (non-cryogenic temperature) compound used there was "liquid metal" thermal compounds ("Our special mixture of metals like tin, gallium and indium") which people used to good effect de-capping their CPUs and doing direct-die cooling. Advertised at 75 W/mk https://www.pccasegear.com/products/35854/thermal-grizzly-conductonaut-thermal-compound-1g

These systems did >200W dissipation in a very small package/die. The relevance of such modifications in the pc enthusiast space has decreased with Intel's apparent improvement of their die-to-case TIM improvements using iridium based solder and doing die thinning.


Longevity from short term tests (1 year) looked ok https://www.gamersnexus.net/guides/3359-liquid-metal-aging-one-year-test-how-often-to-replace-liquid-metal
Not sure about industrial applications with >20yr life expectancies.
Gallium based so NOT compatible with aluminium based heatsinks.

Available in small consumer grade packages as above but I presume there are industrial suppliers.

A quick search for Gallium TIM turned up this https://www.tomo-e.co.jp/upload/cProductsJA/25QU109-cProductsJA_content-001.pdf 83 W/m*k

Some of the more standard phase change TIM and greases I've been looking at are less than a 10th of that
Phase change TIM 7.5 W/m*k https://www.laird.com/products/gap-fillers-thermal-interface-materials/phase-change/tpcm-7000
Thermal grease 3.5 W/m*k https://www.laird.com/products/gap-fillers-thermal-interface-materials/thermal-grease/tgrease-2500

Edit: Good set of test results on high performance greases for consumer CPU cooling https://www.guru3d.com/articles_pages/guru3d_thermal_paste_roundup_round2_2021,10.html

[Conrad Hoffman]

If the interface is good, meaning the parts are smooth and flat, and if you don't use too much grease, the thermal conductivity of the grease doesn't matter much. It's all about getting the gap as small as possible. Do the math- it's interesting! ALL greases are pretty terrible compared to metals, even the fancy ones. IMO, you still need a trace of the stuff, unless you can get things gauge block flat. A name brand thermal grease/paste like Dow 340 or Aavid/Thermalloy will do the trick. Aavid Ultrastick is also something to investigate. It's a phase change compound that basically liquifies when it gets warm, giving the thinnest possible interface. Outperforms grease.

[sandalcandal]

* There are super-thin thermal sheet like products available ("graphene" or something?) that could be worth considering.

Test of carbon based thermal pad in consumer CPU application


There are also different carbon film products that are more heat spreaders rather than a TIM e.g. Panasonic Pyrolitic Graphite Sheets (there were some others that made similar products)

Insanely high thermal conductivities (>1000W/m*k) but anisotropically only in the plane of the sheet, thermal conductivity through the sheet is pretty average.

[nctnico]

I'd use a pad. That will give a repeatable performance. I'm usually using the praphite / carbon type pads for CPUs but these are available for various package types.

[bobbydazzler]

Towards desktop CPUs and "pc enthusiasts", the highest performance (non-cryogenic temperature) compound used there was "liquid metal" thermal compounds ("Our special mixture of metals like tin, gallium and indium") which people used to good effect de-capping their CPUs and doing direct-die cooling. Advertised at 75 W/mk https://www.pccasegear.com/products/35854/thermal-grizzly-conductonaut-thermal-compound-1g

These systems did >200W dissipation in a very small package/die. The relevance of such modifications in the pc enthusiast space has decreased with Intel's apparent improvement of their die-to-case TIM improvements using iridium based solder and doing die thinning.


Longevity from short term tests (1 year) looked ok https://www.gamersnexus.net/guides/3359-liquid-metal-aging-one-year-test-how-often-to-replace-liquid-metal
Not sure about industrial applications with >20yr life expectancies.
Gallium based so NOT compatible with aluminium based heatsinks.

Available in small consumer grade packages as above but I presume there are industrial suppliers.

A quick search for Gallium TIM turned up this https://www.tomo-e.co.jp/upload/cProductsJA/25QU109-cProductsJA_content-001.pdf 83 W/m*k

Some of the more standard phase change TIM and greases I've been looking at are less than a 10th of that
Phase change TIM 7.5 W/m*k https://www.laird.com/products/gap-fillers-thermal-interface-materials/phase-change/tpcm-7000
Thermal grease 3.5 W/m*k https://www.laird.com/products/gap-fillers-thermal-interface-materials/thermal-grease/tgrease-2500

Edit: Good set of test results on high performance greases for consumer CPU cooling https://www.guru3d.com/articles_pages/guru3d_thermal_paste_roundup_round2_2021,10.html

Using a copper heatsink and soldering the part to the heatsink would be best.  Strange I never see that unless you count soldering a part like a dpak to a pcb with lots of copper.

[mazurov]

"liquid metal" thermal compounds ("Our special mixture of metals like tin, gallium and indium")

The trade name for this (or very similar) compound is "Galinstan". $8-ish for a small syringe, lasts a long time, works very well on copper or nickel surfaces. In EU, modern household "mercury" body temp thermometers are filled with the same compound since it is less toxic than mercury.

[Miyuki]

Using a copper heatsink and soldering the part to the heatsink would be best.  Strange I never see that unless you count soldering a part like a dpak to a pcb with lots of copper.

Heating a huge heatsink to solder temperature will be a tough job and heatsink fins are better aluminum, so must be some dual-material job or something

[james_s]

Something to watch out for is that some thermal compounds are conductive. I once made the mistake of using some Arctic Silver I had laying around on a TO-3 transistor with a mica wafer. A bit of the compound oozed into the gap around the transistor lead and burned.

I don't think it matters if the compound dries out so long as the device is not disturbed. It is in liquid form so you can spread it on in the first place and it will fill the microscopic voids, once that has happened it shouldn't really matter if it dries out.