18 greases compared – Ronald Dy Geronimo
After 3 months of testing, I present this TIM (thermal interface material) comparison featuring 18 different materials in paste, liquid, “stick”, pad and foil forms.
The heat source used is an Athlon XP “T-Bred B” 1800+, overclocked to 2410 MHz at 1.75V – heat output is unknown. Die surface area is approximately 85mm². The CPU temperature sensor is a customized LM35 Precision Centigrade Sensor, thermally-coupled to the backside of the processor via epoxy and thermal grease:
There are separate LM35 sensors for both ambient air and radiator outlet water temps.
The processor is water-cooled by a copper waterblock that is held by a specially designed clamping mechanism, which delivers force along a single axis towards the center of the processor die:
This greatly reduces waterblock mounting variances. The closure force is approximately 24 pounds (the maximum AMD recommended force) which results in approximately 183 psi on the tiny Athlon XP die. Spring compression was kept constant over the test runs.
- A CPU loading utility, BurnK7, was used to provide constant, maximized heat output. A special Linux installation on a bootable CD-ROM was used to run BurnK7 so as to minimize any OS interference (see HERE)
- Each TIM was burned-in for 1 hour @ 55°C (nominal)
- Each TIM was allowed to “settle” for 24 hours
- The waterblock was lapped in-between runs in order to remove any residual TIM
- Second runs were optional
Results are expressed as temperature of CPU probe minus temperature of water probe (Tcpu – Twater), in degrees Celsius. Lower is better, of course.
|X23-7804E3 (Shin-Etsu MicroSi)|
Thin BLT required, gray paste
|Ceramique (Arctic Silver, Inc)|
Sticky, white paste
|Radio Shack Silicone HS Comp (metal tube)|
Typical white grease, metal tube
|Stars White (Balance Stars)|
1 run only, typical white grease, “soft pack”, came with generic HSF
|X23-7783D-S (Shin-Etsu MicroSi)|
Post-heating required, gray paste
1 run only, typical white non-silicone grease, expired
|Powerfilm Bar (Power Devices – Loctite)|
Black “stick” PCM, crayon-like, WB preheated on application in order to dispense evenly
|Radio Shack Silicone HS Comp (plastic tube)|
1 run only, white grease, less opaque, plastic tube
|T-pcm hp105 (Thermagon)|
Tacky, white PCM
|Stars-700 (Balance Stars)|
1 run only, silver grease
|X23-7798 (Shin-Etsu MicroSi)|
Viscous gray paste
|G-751 (Shin-Etsu MicroSi)|
Viscous gray paste
|T-grease 2500 (Thermagon)|
1 run only, abrasive, viscous
1 run only, black PCM coated on aluminum carrier
|FSF52 (Orcus, Inc)|
1 run only, white PCM, non-tacky
1 run only, blue suspension, very fluid
|Blue II (Nanotherm)|
The best performer overall was the Shin-Etsu MicroSi X23-7804E3, followed by the WHITE greases (Ceramique, Radio Shack metal tube, Stars white). Apparently you can’t go wrong with “generic white goop” as far as performance is concerned, although longevity characteristics are equally important; what may perform well at the onset may deteriorate several months down the road, owing to migration, pump-out and drying issues. Which brings us to the supposedly longer-lasting TIMs, the phase-change materials (PCM).
PCMs are solid at room temperature but soften/liquefy at CPU operating temperatures. The best performing PCM was Power Devices’ Powerfilm, scoring within 3°C of the best grease, performing even better than many of the greases in this roundup. This should lay to rest the notion that all PCMs are very poor performers.
Not all PCMs perform well, though; the FSF52 by Orcus Inc. did poorly, as did Berquist’s Hi-flow 225F-AC due to its aluminum foil carrier.
At the bottom is Nanotherm’s Blue II, the worst performer in this test, retested three times to verify the poor results. Nanotherm’s novel concept PCM+ (new formula) also performed very poorly in this test.