Applying Thermal Grease

4/2
2008

Applying Thermal Grease

Posted by Joe Citarella in Uncategorized

Like it says – Joe

SUMMARY: Anything but the “razor blade” works well.

Applying thermal grease to get a uniform pattern is an important part of getting the most out of CPU cooling. It used to be that the accepted method was to apply a thin coat using a razor blade of the edge of a credit card. Most performance thermal compounds in use today are lightly bulk loaded thin or wet compounds that spread easily, but a compound that has a bulk loading >90% will not spread using the razor blade technique.

Alternatives to the razor blade have included a number of patterns. To ascertain how effective these are, I applied thermal grease (Tuniq TX2) to a pane of glass and then clamped a second pane over it to simulate heatsink clamping to a CPU’s IHS. It was baked for about ten minutes at 150ºC to accelerate spreading. I tried five different spreading schemes:

Razor blade
Straight line
Pea size
Rice gain size
“X”

The following pic shows these patterns:

Grease Patterns

After clamping and baking, the following pic shows how the grease patterns fared:

Grease Pattern After Clamp and Bake

The WORST pattern was the razor blade – blame my application technique, but when you think about it, any ridges or gaps in applying grease may result in coverage gaps – a closer view:

Razor Method Closeup

Of the other techniques, all of them will spread nicely over time. I think, however, the best method is to use any technique that starts with grease in the center of the IHS and spreads from there. Personally I believe the best technique is the “pea” approach – this concentrates grease over the hottest point of the IHS (the center) and spreads out uniformly from there. Any excess will be squeezed out.

CONCLUSIONS

Forget the razor blade and go for the pea.

Email Joe

Tags: Cooling

18 Comments:

Advent

10-27-10 01:14 PM

I've been using whatever grease comes with the heatsink lately, but is Artic Silver 5 still the top-dog when it comes to 3rd party thermal grease for watercooling?

m0r7if3r

10-27-10 01:20 PM

nope!

tons of reviews out there, this one's good, if a bit limited...

http://skinneelabs.com/tim2010part1.html

mx2 is sorta being surpassed now and falling to artic silver 5's place as the old top dog.

beenthere

10-27-10 01:52 PM

If 1-2 C changes your life then there are a few expensive heatsink compounds that may be better than AS 5. From the tests I've seen and my own experience, AS 5, Ceramique, MX-2, etc. are virtually all within a degree C of each other. Proper application is probably more important than which one of the better heatsink compounds you use?

YMMV

SunJean

10-27-10 02:09 PM

Found out that video 1 week ago, if you are interested in the best method to spread the termal grease

I tried it and works very nice the pea method(rated the best due to not bubbles of air)
If there is something wrong with it let me know

http://www.youtube.com/user/yaochoon#p/u/17/ffK7L0Qj13Q

I.M.O.G.

10-27-10 02:20 PM

Yup, thats the only TIM review I'd trust - I haven't seen one other that impresses me as accurate or reliable. The problem with TIM evaluations is that you are dealing with very small differences in temperature due to the TIM, and relatively large variability due to differences in the interface quality and testing procedures.

For a TIM review to be really meaningful, recording and reporting things like this are important in order for the reviewer to understand how much variation there is:
http://skinneelabs.com/tim2010part1.html?page=4

The big TIM roundups mostly get it wrong, because the amount of time required to do it right is prohibitive for actually doing it. That's why Cam only did a limited number of products here.

m0r7if3r

10-27-10 03:23 PM

Actually, the TIM is all done by vapor (eric), but he's just as meticulous and skinnee would never let anything fly under his banner that wasn't the highest quality work around. I like [H]'s tim testing, but they don't really do it anymore...they had logged something like 2500hrs on the bench doing their testing and their temp probe is an actual probe put into a groove on the cpu they had milled, so it's pretty damn accurate...but yea, tim testing can be dubious at best, multiple mounts is a MUST and excluding the outliers is necessary as well, so you have to have enough data that you can lose a few sets of it and still present something reliable and conclusive.

I.M.O.G.

10-27-10 04:08 PM

Heh, ya, they're interchangeable as far as their testing quality. Robots are robots.

skinnee

12-01-10 12:52 AM

Yes, I'm going thread necro... sorry, work has had me buried and I am finally out from under the pile during the day and now back to the pile waiting in the lab.

Anyhow... mort, think about that milled IHS for a minute, how does that affect surface area or TIM spread, and so on? There are many reasons why we do not mill the IHS's on our chips.

Robots, yup, pretty much (no sarcasm, it is the truth). And to make matters worse, we're both in an automation kick trying to remove all the manual steps we possibly can.

m0r7if3r

12-01-10 01:02 AM

I think it'd be workable if you did it right, but I could see either side of the argument and you guys' method is just as, if not more, accurate than [H]'s method...when's vapor's next round of tim testing out, btw?

skinnee

12-01-10 01:26 AM

Yeah, there are pros/cons either way you slice it.

I think the initial batch of 7 TIM's (I think we're up to 50+ in total) will be out around Xmas time, but I'm 99% sure there are changes coming to the current TIM bench. We've settled on 5 mounts at three different mount pressure/contact settings (think bondline thickness), so a total of 15 mounts, each at 10 hours duration. The change I spoke of will be in one of the pressure/contact settings... just didn't turn out like initially planned.

BTW, the amount of awesome data coming from the TIM bench is going to make your head spin. Vapor may have grown a bit tired of me asking questions as we were going through the initial data.

m0r7if3r

12-01-10 01:36 AM

That's mindboggling...you guys' devotion never ceases to amaze me.

SMOKEU

12-01-10 03:46 AM

I don't like any of those methods as none of them spread the paste over the whole chip. I use the pea method then spread out the paste with a plastic card to cover the whole top of the processor.

I.M.O.G.

12-01-10 05:59 AM

Did you read the article above?

Lot's of people do things they haven't really reliably tested... Often times people tend to think something is better, but they don't have the means or process to accurately observe differences and gauge results. More often than not, people prefer the way they do things, regardless of what may have been demonstrated elsewhere to "be better".

beenthere

12-01-10 06:38 AM

Thermal compounds for PC people are like engine oils for car people - a religion. Everyone has an opinion and little technical data to support their beliefs or ad hype.

As far as the video and thermal compound application is concerned:

1. In many of the shots the TIM was applied too thick or not spread completely by the pressure the person applied.

2. TIM is only designed to fill the imperfections between the heatsink and CPU heat spreader and should be as thin as possible and providing 100% contact on the heat spreader

3. Different TIMs spread differently based on viscosity and chemistry

4. All TIMs flow with heat meaning that what you saw in the video changes with heat/pressure and time

5. The air trapped in the first application appeared to be induced by the bending of the plexiglass used to show the spreading action. There was also way too much TIM which is why air could remain between the heatsink and plexiglass

6. While the highest heat is concentrated in the center of the CPU heat spreader any of the methods that do not cover the entire CPU heat spreader with a very thin layer of TIM may be giving up a bit of thermal conduction and thus CPU cooling?

In my instrumented testing of heatsinks and TIMs over many years the best results come from using the thinnest TIM application (almost translucent in appearance), evenly spread across the entire CPU heat spreader. That being said the difference in CPU temp between just enough TIM and too much IME is typically 1-2 C max. A small tube of TIM last me for many CPU applications.

The reason why you want the thinnest application of TIM is because even the best TIM is not 100% thermally conductive. Thus the the thicker the TIM the less efficient the heat transfer compared to 100% contact of the CPU heat spreader and heatsink.

People are free to use whatever TIM and application method makes them happy. I'm just the messenger so use the info. as you please or completely discard it.

Just a nickname

12-02-10 04:56 AM

I was wondering if the double line:

-------------

-------------

What is the best solution?
I always put a small drop on the cpu and the tim never went outside of the surface nor wasn't fully applied on it...

freeagent

12-04-10 07:15 PM

I like IC Diamond 7, after useing AS5 for years.

Ive tried a few diff application techniques, and the pea works for me the best with IC7.

Hey skinnee nice Greyhound, I have one too