Overclock.net banner

Has it been settled, how to apply TIM?

8K views 141 replies 35 participants last post by  Iwamotto Tetsuz 
#1 ·
Last I checked, there were people who spread the paste manually and people who apply a small dot and then let the pressure from the heatsink spread it out.

It was a long time ago, but as I remember it I applied it manually first, spreading it out as evenly and thinly as possible. The temperatures were horrible. Then i used the other method and must have lost at least 15 degrees. The results spoke for themselves. It seems to stand to logic, as well, as it removes opportunity for air-pockets to form and creates a vacuum seal. And for some strange reason when I took the heatsink off I saw that the TIM had spread up to the edges without going over, maybe someone can explain that to me.

So why does anyone defend the other method? Does it depend on the type of TIM being used?
 
#5 ·
We want a print similar to tiny dot / small dot gave in pictures below. Amount of TIM needed depends on cooler base. smooth polished bases us less TIM than direct contact bases like 212 which have voids that need filling between copper and aluminum. Any more TIM / bigger print is wasting TIM and usually results in higher temps.
wink.gif



 
  • Rep+
Reactions: francisw19
#6 ·


http://www.hardocp.com/article/2009/04/07/thermal_paste_shootout_q209/3#.VGSluRxwq7Y

Why isn't the drop always winning and why is the difference so small? Because a properly done spread (properly, as in non resulting in 15C difference), covers more surface and doesn't include guessing. More surface coated= more area for heat exchange, which counterbalances the famous air pocket advantage of the pea... Pea has less air pocket, but also includes guesswork, which may leave areas uncoated, uneven spread or too much paste.

So, at the end, if you 're comfortable with one method, use it. It's better than trying to follow a method that you aren't good at.

(I spread, because i want to see what i am doing, before positioning the heatsink and i use high viscosity TIM, that according to the above table, does better spreaded).

Guessed wrong pea size? Didn't position the heatsink perfectly vertically and the TIM spreaded more towards one side leaving much of the heatsink without paste? With the pea you will only know once you remove the heatsink:



There are times, one wishes his heatsink was made of plexiglass, so to be able to see this without removing heatsink...



At least when you spread, you can only blame yourself for lack of common sense and poor eye vision, but at least you can see the result before your sit the heatsink.

Now let's all fight about this 0.5C! My method is sooooo much better! So much!
biggrin.gif
 
#7 ·
The CPU chip, the heat source, does not touch the entire area of IHS. And the IHS doe snot spread heat any better than a good cooler base. The result is only a square or rectangle in center area of IHS is radiating heat.

Something like this



The more TIM that is applied and the bigger the area it covers, the less pressure there is and the thicker the TIM remains. And as the TIM does not transfer heat as well as copper / metal, the thicker the TIM the less / slower the heat is to leave the chip.

Hope that makes sense.
 
#8 ·
Quote:
Originally Posted by doyll View Post

The CPU chip, the heat source, does not touch the entire area of IHS. And the IHS doe snot spread heat any better than a good cooler base. The result is only a square or rectangle in center area of IHS is radiating heat.

Something like this



The more TIM that is applied and the bigger the area it covers, the less pressure there is and the thicker the TIM remains. And as the TIM does not transfer heat as well as copper / metal, the thicker the TIM the less / slower the heat is to leave the chip.

Hope that makes sense.
I hope i won't sound offensive but, no, it makes no sense to me. If it did, the difference between the pea and spread would be much greater. Heat doesn't transfer in perfect vertical fashion. The heat will transfer first in more surface of IHS, than what the pea method leaves to imagine and then to the cooler. Just like if you heat a pan, the heat will not remain just in the center of a pan where the flame is concentrated, but will heat up the "walls" of the pan too. This because the heat travels best within the same metal, than between two metals or 2 different substances anyway. The result, is why if you touch the walls of a pan, you will burn your hands, despite the walls not being in contact with the kitchen flame. Because heat, travels through the same metal very well. Maybe less heat than in the part directly hit by the flame, but still hot enough to burn you. The result in a CPU, is that the entire IHS will be radiating heat, just more in the center, exactly like a pan. Because, like you say, the TIM, despite being in the center of the flame, isn't as good in thermal transfer as to completely eliminate the horizontal thermal spread in the IHS, because metal to metal, always beats metal to TIM. So, at the end, the theoretical advantage of the pea, that eliminates all air, is in reality shrunk, by the larger area of contact that the spread method has. The more the IHS surface that is coated, the more the heat transfer, because heat transfer isn't limited to the center of the IHS, but to the periphery too.

Or to put it in more "scientific" terms, heat in conduction, spreads through gradient:



Where k is the thermal conductivity of the material,
A is the cross sectional area,
THot is the higher temperature,
TCold is the cooler temperature,
t is the time taken,
d is the thickness of the material.

http://formulas.tutorvista.com/physics/heat-transfer-formula.html

Why does the heat transfer to the rest of the IHS too and then radiate to the CPU cooler?
- The K, thermal conductivity of the IHS metal, will always be higher than the K of TIM (metal to metal is better than metal to TIM and even better than metal to air). This explains why the heat will also move horizontally to the periphery of the IHS.
- THot-TCold (DeltaT) means, the heat will always go from hotter to colder point and the bigger the difference, the quicker the transfer. So it can't just "bypass" the rest of the IHS in order to go to TIM and then to the CPU Cooler. The rest of the IHS is cooler than the center of the IHS too. So heat will move to the periphery of the IHS too... For the same reason that heat can't bypass entirely the pan's external walls and just heat up just the food in it.
- d is inverse proportional to the heat transfer. So, the thicker the TIM, the bigger the d, the worse the heat transfer. Which explains why thick TIM is bad.

I think this explains both why there isn't a spectacular "win" of the pea method and why thin is good. EVERY little corner of IHS, will heat up and partecipate to the heat transfer (increasing "A" - the cross section area), as long as that corner of IHS is in higher temperature than the CPU cooler base above it. Which will always be the case, because metal to metal, beats metal to TIM.

Had it been otherwise, the pea method would give surprisingly better results than the spread method, for eliminating those pesky air pockets that act as insulators (lower K, higher d). But it doesn't, because the lowly IHS periphery, counterbalances for good part this effect, contributing to the heat transfer much more than a non coated surface of the pea method (coated IHS to cooler is much better than non coated IHS to cooler, in the periphery).

If only the TIM had better K (thermal conductivity), than the metal itself, would the pea method shine. But alas, the metal to metal is always more efficient than metal to TIM. So despite the source of the heat being in the middle, the rest of the IHS, offers better K (the same as the entire IHS) than the TIM. And this why the entire IHS will partecipate to the heat radiation towards the CPU cooler. (passing through 1 material vs passing through 2 materials).

It also explains why liquid metal TIMs, perform much better than any synthetic TIM, despite the latter being applied with the "perfect" pea method. (much bettter K, increased A compared to simple pea method, no air).

Because, at the end, coolers aren't made of plexiglass...

EDIT: To put it in 2 words:

- Pea method, eliminates the air, allowing for low d, but has decreased A (areas not covered by TIM, will transfer heat much worse than any air pocket in the spread method).
- Spread, has higher d (air), but larger A.

Result: Instead of a crushing victory for pea, the differences are negligible and possibly within statistical error...

 
#9 ·
And a quick thought that just came to me, thinking again about HardOCP's results, that separates the TIMs according to viscosity. I think, it may have to do, exactly with the air... A high viscosity TIM, when spreaded, is maybe less likely to allow air trapped in it, than a low viscosity TIM (practical example, is it easier to blow air (form air bubble) into water or into liquid cement that is more viscous? Into water of course. In a CPU's case, the force blowing air, is the pressure applied by the heatsink). So high viscosity TIM, could be more prone to have less air bubbles than low viscosity, which could explain why high viscosity TIMs did better than low viscosity in HardOCP's spread test. Basically, the high viscosity TIMs, would suffer less the "air pocket" penalty (d) compared to low viscosity TIMs. Just a hunch though, haven't given it much of a serious thought.
 
#10 ·
Those who complain that they see a horrible performance of spread, are also usually badly educated about "how a spread is supposed to be". I was googling recently for reviews of BeQuiet Dark Rock 3 and found this review (machine translated), where the author was advertizing his spread as "excellent" :



I don't know, by what standards, a professional reviewer would boast for this spread and claim it's very good. It's too uneven, too thick near the edges, too thin near the center, paste around the socket, that's a horrible spread like if he did it using his finger! And readers are "educated" by this reviewer on what a good spread looks like (yes, that photo is BEFORE applying the heatsink).

Bottom line: Follow the method that you are comfortable with. I am not good with guesswork, so i follow spread. One like the guy above, who is clearly out of his mind if he puts this photo on the internet to show a good spread, should do the pea, hoping he is better than spreading...

The best way to spread, is using the small spatulas that come with several pastes, like this:



They allow for adjustments in small areas, where credit cards for example, being too big can fix an area and ruin another. And one must use common sense. For example, Ceramique being viscous and sticky, for the final refinement, only needs a gentle caress with the spatula, otherwise it sticks to it or removes the paste from the IHS. One needs common sense (that the reviewer above lacks for example).

When spreading, one wants to imitate the way factories spread, like this, which is thin enough to allow to see the grooves of the surface:

http://archive.benchmarkreviews.com/images/articles/Best_Thermal_Paste_Application_Methods/Thermalright_Surface_Finish.jpg

Not play "Pablo Picasso on a cpu", like this:

http://www.online-tech-tips.com/wp-content/uploads/2010/06/ThermalPaste1.png

This is a nice spread, but even here, one could possibly do with slighly less:

http://fatoo.co.uk/wp-content/uploads/2012/04/good_thermal_grease_application.jpg

In the spread method, since air pockets are impossible to eliminate, being maniac about complete even surface is less important than the thickness. Makes more sense to go for thin layer, than for perfectly even layer (within reason).
 
#11 ·
First, we cannot use a flame heat source for a comparison. An electric heating element has to be used.
Second, their will be much greater heat in center of pan than on sides.

Getting the heat to spread better is why pans have thicker bottoms with a core of copper or aluminum.

As for heat spreading better in metal, it depends on what metal it is and the finish also has to be considered.. Stainless steal is terrible at transferring heat.

Indeed, the IHS radiates heat .. same as your pan analogy. But the same pan analogy can be used to show how much hotter the center area of pan gets than the sides. I'm sure you have burned something in the center of a frying pan when the edges were not even done cooking.

Another factor that is very important in how well heat moves from IHS to cooler is how much pressure the mount uses.

We can talk theory and show equations until the cows come home. The reality is what gives best cooling in actual use .. and TIM over the entire surface of IHS to heat sink versus just enough to make a print similar to what I drew shows that the small print works as good or better than a fully covered IHS.
thumb.gif
 
  • Rep+
Reactions: Iwamotto Tetsuz
#12 ·
If you spread it you'll get decent results if it's thin. It was my method for years (corner of a plastic bag + finger + spread) but a small dot the size of the hole in the tube of thermal paste is 10 times easier.
 
#13 ·
Quote:
Originally Posted by The Pook View Post

If you spread it you'll get decent results if it's thin. It was my method for years (corner of a plastic bag + finger + spread) but a small dot the size of the hole in the tube of thermal paste is 10 times easier.
LOL
thumb.gif

When mounting lots of coolers one definitely needs the most efficient way of cleaning and installing. When using the same kind of TIM on installs, I just wiping the IHS and cooler base clean of old TIM and putting a dot on IHS works as good or better than spending lots of time getting them super clean and tring to spread a even thin layer of TIM. The "experts" call it "timming" the surfaces .. I just call it efficient use of energy and time.
biggrin.gif
 
#14 ·
Used to go with a "tiny dot" then with 2011 I moved to a "small dot" lol. Seems to work well and is easy.
 
#15 ·
Nothing works better for me than tiny dot, no matter what socket it is. I use Socket 2011-3 and I first tried a small dot with a 5-6mm diameter. What I ended up with was bad temperatures, a lot of TIM on the socket and motherboard and a corroded IHS. Yes! You read that right, corroded!

You want a minimal amount of thermal paste because it's just there to make sure there are no air bubbles between the IHS and the heatsink. It's better to have too little than too much, especially if it's a conductive TIM.
 
#16 ·
Quote:
Originally Posted by doyll View Post

First, we cannot use a flame heat source for a comparison. An electric heating element has to be used.
Second, their will be much greater heat in center of pan than on sides.

Getting the heat to spread better is why pans have thicker bottoms with a core of copper or aluminum.

As for heat spreading better in metal, it depends on what metal it is and the finish also has to be considered.. Stainless steal is terrible at transferring heat.

Indeed, the IHS radiates heat .. same as your pan analogy. But the same pan analogy can be used to show how much hotter the center area of pan gets than the sides. I'm sure you have burned something in the center of a frying pan when the edges were not even done cooking.

Another factor that is very important in how well heat moves from IHS to cooler is how much pressure the mount uses.

We can talk theory and show equations until the cows come home. The reality is what gives best cooling in actual use .. and TIM over the entire surface of IHS to heat sink versus just enough to make a print similar to what I drew shows that the small print works as good or better than a fully covered IHS.
thumb.gif
I am not trying to convince that spread is better nor do i wish to offend other people's opinions. I just thought to provide some more tests with various TIMs performed by professional reviewers and give also a theoretical explanation, based on the formula for heat conductivity, about how these results can be explained. Like i said, to me the difference in temperature is minimal, possibly inside the statistical error, so at the end, one should choose the method that he feels he has mastered well. Using a method that you screw up, is worse than anything else. Myself, because i use CPU coolers that latch on the stock AMD bracket and when i "land" them i can't see well what i am doing, i tend to have more bad applications with pea. So i stick with spread. On the other hand, the reviewer i showed in the previous page, should abbandon spreading and concentrating to pea, because clearly he can't spread. At the end, the difference between the methods when done by expert hands (as in HardOCP), is trivial, so it's like hair splitting. I just can't stand seeing the comments of "pea is 15C lower than spread"... That's all.
 
#17 ·
Quote:
Originally Posted by HiTechPixel View Post

What I ended up with was bad temperatures, a lot of TIM on the socket and motherboard and a corroded IHS. Yes! You read that right, corroded!
I think you're confusing oxidation and corrosion. Unless your IHS was ruined afterward, it was oxidation.
 
#19 ·
Your TIM ate metal?

I think you're using the wrong material.
biggrin.gif
 
#21 ·
You people are arguing about a 2c difference between all methods (from best to worst). FFS, this debate has been happening for decades and there are still people who gives 2 nuts about which method is better? It doesn't bloody matter. 2 degrees celcius man! Your CPU isn't going to explode either way.
 
#22 ·
Quote:
FFS, this debate has been happening for decades and there are still people who gives 2 nuts about which method is better? It doesn't bloody matter. 2 degrees celcius man! Your CPU isn't going to explode either way.
Well said!

This is how I applied mine:

TIM application (Cryorig TIM)

The chip under the cover is rectangular so it made sense to me to have a line of TIM on top of the chip than a dot. Yes, I probably applied a little too much but my temps are excellent and for CPU idle or web browsing etc, I can turn the cooler fans off and run it passive. 5 to 10 degrees above ambient. Max CPU temp with Aida64 FPU is well below TJ max. So I'm happy. Which TIM did I use? The one that came with the cooler (Cryorig R1 Universal)

In the end what TIM you use, and how you apply it doesn't really matter as long as you're sensible. Read up on it all, follow what guidelines you think make most sense to you and don't worry so much. A couple of degrees here or there is of little consequence. IMO. There's far more important things in Life to worry about. . . .

Ali.
 
#23 ·
Quote:
Originally Posted by benbenkr View Post

You people are arguing about a 2c difference between all methods (from best to worst). FFS, this debate has been happening for decades and there are still people who gives 2 nuts about which method is better? It doesn't bloody matter. 2 degrees celcius man! Your CPU isn't going to explode either way.
Why then are you getting into it here and now?
biggrin.gif
 
#24 ·
Quote:
Originally Posted by Undervolter View Post



http://www.hardocp.com/article/2009/04/07/thermal_paste_shootout_q209/3#.VGSluRxwq7Y

Why isn't the drop always winning and why is the difference so small? Because a properly done spread (properly, as in non res

Now let's all fight about this 0.5C! My method is sooooo much better! So much!
biggrin.gif
This sounds pretty much ture to my experince, since i trided diffrent paste high conductive vs my normal silicone heat sink grease and what ever, no matter what method i tried what paste i use temps are always the same no diffrence in performance with my cooling, i will only see a +-2 c diffrence maximum , and yes i am a fan of the spread method, marble method, i use heaps of heat sink grease spread it out and when i take my cpu cooler of most of the time the cooler will rip the cpu with out of the socket with it since its so air tight, so i am gonna try right now to use the dot method taking my thing appart and do some testing with it at 1.55V core, currently with spread method cpu metal temp (ihs)is around 30-38C diffrent from water in the loop, and core is around 6C higher or lower than cpu metal temp(ihs)
 
#26 ·
We will be testing at 27c out water temps and this is test one with (CUSTOM Black silicone + white + .....)mixed silicone mixture

max core temps are about 4-5c higher than ihs
Water flow was the same throughout the test and same fan speed they are both externally powered and have seprate voltage controlers for pumps and fans
Water flow very slow minium pump speed of the two brushless pumps in my loop cause i only use soundless computers or extremly silent computers
Same with fan speed very slow about 600-700RPM
http://valid.x86.fr/374eav

also stuff has been cleaned with pure alchool before applying new grease

also after that used micro fible cloth to get rid of dust

during each test cpu heat sink has been insured its put on proprerly same screw tension

This test is not just for you guys its also for my self since i am crazy with oc and i wanna find out whats the best way and whats the best heat sink grease to use on cpu
thumb.gif


CPU voltage is not constance since i am using a 970 pro 3 that has no vrm heat sink and it refuses to out put enough for a constant voltage


 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top