Thermal grease can only go so far – Joe
SUMMARY: Thermal grease can only go so far – even the most theoretically efficient grease might only yield another 1 C compared to what’s now available.
We pay a lot of attention to cooling, and one key element in cooling is thermal grease. As the only interface material between a heatsink (air, water, whatever) and the CPU core, it plays an essential cooling role. We’ve seen some great leaps forward in thermal greases, and current high-quality products on the market do an exceptional job.
However, as exceptional as they may be, there appears to be some confusion among thermal grease users as to what benefits can be expected; specifically, how much does thermal grease contribute to lowering CPU core temps?
Let’s get real, folks. There are no magic bullets out there – among the “quality” greases, the differences are minor; for the “average” user, it’s doubtful that, given the incredible imprecision of on-board temp sensors and temp monitoring gear, coupled with variations due to heatsink mounting pressures, joint flatness and grease application, the differences are measurable.
To give a more balanced picture as to what thermal grease can do, let’s first get technical. The following excerpt from an article in Electronics Cooling called Thermal Interface Materials, by Dr. Miksa deSorgo, Chomerics Division, Parker Hannifin Corporation, describing the key parameters:
“The rate of conductive heat transfer, Q, across the interface [eg, between a heatsink and CPU core] is given by
Q = kA(Tc – Ts) / L
where k is the thermal conductivity of the interface, A is the heat transfer area, L is the interface thickness and Tc and Ts are the device case and heat sink temperatures. The thermal resistance of a joint, Rc-s, is given by
Rc-s = (Tc – Ts) / Q
and on rearrangement: Rc-s = L / kA
Thus, the thermal resistance of the joint is directly proportional to the joint thickness and inversely proportional to the thermal conductivity of the medium making up the joint and to the size of the heat transfer area. Thermal resistance is minimized by making the joint as thin as possible, increasing joint thermal conductivity by eliminating interstitial air and making certain that both surfaces are in intimate contact.” [my emphasis]
Thermal grease is the “k” here. The better a grease can conduct heat, the less its resistance between the heatsink and CPU core. DUH!
The key question here is what can be expected? Is it conceivable that there is a thermal grease out there lurking in the bushes with some secret goop that will knock the socks off the competition? And second, even if there is, what can it do for us??
To put some bounds on this question, I’ve done some looking around at thermal grease specs, some you know and some less well known. The key characteristics in which we are interested, thermal conductivity and thermal resistance, can be measured by:
Thermal Resistance: C-in²-W
Both measures are readily available for commercial greases but noticeably lacking for “retail” greases, with some exceptions. First, let’s take a look at the
Thermal Conductivity of Selected Pure Metals: W/mK
- Silver: 417 W/m K
- Copper: 394 W/m K
- Gold: 291 W/m K
- Aluminum: 217 W/m K
You’ve probably seen this data before, so no surprise here. Now let’s contrast that with the
- Arctic Silver III: >9.0 W/m K
- AOS Thermal Compounds 57000:7.21 W/m K
- Shin-Etsu G751: 4.5 W/m K
- AOS Thermal Compounds HTC-60: 2.51 W/m K
- Thermagon T-grease 412:1.3 W/m K
- Radio Shack Thermal Grease: 0.735 W/m K
Quite a range! Note, however, that even with W/m Ks approaching 10, it’s far below metals (eg., silver: 417 W/m K). Now let’s look at the flip side:
- Arctic Silver III: 0.0024 °C-in²/Watt¹
- AOS Thermal Compounds HTC-60: 0.01 °C-in²/Watt¹
- Thermagon T-grease 412: 0.022 °C-in²/Watt
Less published data for °C-in²/Watt, but enough to give some interesting observations. Now let’s assume a CPU core, such as the Palomino at 129mm sq, for a surface area of 0.20 sq inches. Calculating the C/W:
Arctic Silver III C/W = .0024 / 0.20 = 0.012 C/W
AOS Thermal Compounds HTC-60: 0.01 / 0.20 = 0.05 C/W
Thermagon T412 = .022 / 0.20 = 0.11 C/W
So at 100 watts, using AS III between the heatsink and CPU core would increase CPU temps by 1.2 C, 5 C for HTC-60 and 11 C with the T412.
Remembering that there is an INVERSE relationship between Thermal Conductivity and Thermal Resistivity, it stands to reason that any thermal grease with a W/m K less than Arctic Silver’s will be more resistive and therefore result in higher core temps.
Let’s assume a grease EQUAL TO pure silver.
With silver at 417 W/m K Thermal Conductivity, Thermal Resistivity would essentially be equal to 0. So the “ultimate” thermal grease, compared to what’s now available, will reduce CPU core temps by about 1 C.
I repeat: The “ultimate” thermal grease will reduce CPU core temps by about 1 C.
So, if you see anecdotal evidence, or a “test”, of a thermal grease which gives something like 3 to 7 C better performance compared to the best that’s out there now, I’d be REAL sceptical.
The upshot of all this is that thermal greases available today are about as good as it’s going to get. Incremental improvements are likely, but considering just thermal grease alone, don’t expect any quality grease to cost you more than 1-2 C at 100 watts radiated heat.
Considering other factors affecting the thermal resistance of the joint between the heatsink and CPU core, users will get more benefits in improved cooling performance by lapping heatsink bases, ensuring heatsink mounting pressures are to spec and re-greasing at regular intervals than searching for the “ultimate thermal grease”.