Replacing thermal paste - a quick guide in 4 steps

[Faust]

I posted this on another forum years ago but can't remember which one now

I've lived and worked as a tech in many hot climates and I've replaced thermal paste on more machines than I could count.
Over time the paste dries , shrinks , then cracks .
Once it has dried out it is useless , and my first aim is to remove every trace from the surfaces of the heatsink and the chip .

a] Surface prep - I always use a small piece of plastic cut from a bank card as a "safe" scraping tool
Note - Use an "original" edge of the card to scrape , not the edge that you cut .

b] Cleaning - after step [a] .... with iso-propyl alcohol .... why ?
Ethyl alcohol ( eg methylated spirit ) always contains some water that can never be removed by distillation ,
but propyl alcohols contain very much less .

Iso-propyl alcohol is sold in many countries as " Rubbing alcohol " but for me , by far the best way to buy and use it
is in the form of sterile "Pre-injection swabs " ..... ask at a pharmacy .
They are just perfect for this job ( and also for cleaning the heads on tape recorders , bill-readers etc )
You get foil sachets containing a square patch of synthetic fabric which will leave no fibers or lint ,
and they are pre-soaked in propanol .
Cotton buds are rubbish by comparison ..... a disaster even.

c] Applying new paste - how much to use , how to apply it , and which one ?

My rough measure is the size of a grain of rice , applied to the top of the chip , in the dead center .
Note : the first portion out of the tube often has an abnormal amount of liquid , so it's worth wasting the first dose
by putting it on some paper .

Personally I like Arctic Silver 5 ; it's not too thick or thin , and has above average thermal performance ,
and a ten dollar tube is good for 6 or more jobs ( and it doesn't dry out in the tube ) .

d] Fit the heatsink .... using gentle pressure and some very slight twisting , then put all of the screws in place ,
and tighten each one gently , in a cross-wise pattern.

WARNING ! .... Over-tightening the screws = throwing out the baby with the bath-water .
Even if you don't bust the screw threads , there's a good chance that the heat sink will get warped or skewed ,
and give very poor thermal transfer .

I had a link to someone who must have been either very dedicated ( or strange ,or bored ) who had set up a gaming rig as a test bed
and they had tried just about every commercially available paste ( and some can be very expensive ! ) and then moved on ,
and tested with other things , eg. nothing at all , chrome polish , toothpaste and even Nutella ! .... I'm not kidding !
And in each case they had stress-tested the cpu/gpu and tabulated the results
.... I now wish I could find THAT link also

Take your time and take care .....
Good luck ( but you shouldn't need it ) .

[all41]

Here is a link to efficiency tests of thermal compounds.
Interesting to note that that there is scarcely 3/4 degree C difference among the top ten.

https://archive.benchmarkreviews.com/in ... itstart=12

How to apply:
https://archive.benchmarkreviews.com/in ... mitstart=2

All pretty well supports the above methodology

[trytip]

5 degrees can make a lot of difference. most of the temperature is lost in the transfer copper plating of heatsink not flat against the cpu/gpu
@Faust i'm not sure why there isn't a better system in place to attach the heatsink to the motherboard. i lost a screw over tightening it, but in my case it was ok since i only did it to test the pea drop by then the card was already displaying strange characters in bios i could not see anything everything looks as if it was in ascii

[vladtepes]

A lot of sold assembled computers - particularly the cheaper ones - often have scant attention paid to this aspect when they were built.

I've seen very small amounts of thermal compound used (none in one case), some horrible thermal 'pad' sticker things, and all sorts.
I've pulled CPUs out to see the thermal paste fully dried and crumbly in a computer well under two years old.

So as a general guide - it's something worth doing in most cases BUT be sure that you don't do any damage to the cpu when removing. If you aren't confident on that then get someone else to do it, or just leave it alone.

[srq2625]

I have found, as demonstrated by that article linked to by all 41, that it is nearly impossible to simpley add a drop of grease to the cooler, press the cooler to the CPU and expect the grease to completely cover the mating surfce between CPU and cooler (I am assuming that the greater the contace area, the greater the thermal conductivity - all else being equal).

So I do it a bit differently (and I guess I'll take a ration of bovine excrement for this this process, but ...):

• Prep the cooler: I use a 1-foot square piece of tempered glass (it's the flattest surface I can cheaply acquire), a sheet of 600-grit wet sandpaper, and liberal amounts of water. Lap the mating surface of the cooler until it's pretty.
• Clean the cooler really well. Isopropyl alcohol works very, very well for this when soaked into a lint-free laboratory wipe.
• Apply some thermal interface material (TIM, aka - grease) to the cooler mating surface. I just dab the applicator all over the mating surface, leaving just enough TIM that I can see it. Lots of little dots.
• Using a single-edged, straight razor, held at a very acute angle, "squeege" that grease all over the mating surface and fill in all the gaps - especially if it's a copper heat tube in aluminum mounting bracket. The idea here is to remove as much of the TIM as possible with the razor while, at the same time, leaving enough to fill in all the perfections.
• Now, do virtually the same with the CPU. Clean it with isopropyl alcohol to remove all surface issues. Apply the TIM using the above described technique, but here, I want to leave enough TIM that I can just barely see through it - just barely make that there is writing on the surface of the CPU.
• Now, mate the cooler to the CPU, applying as much pressure between them as I can reasonably expect to be enough. How much? It's a lot like applying torque to a nut on a bolt - you just know when you have enough and not so much that you cause damage.
• Every cooler mounting device I've seen allows for a little "wiggle room" - I use it to give the cooler a bit of a twist back and forth a couple of times. Then I square up the cooler and call it done.

So, why the razor blade and not a bank card? A couple of reasons, but the most significant one being that a razor blade is a finely machined edge - very, very straight - with no gaps or nicks in the surface. Held at a very acute angle to the surface of the cooler mating surface and, again, to the surface of the CPU, there is almost zero chance of causing damage. But, I only use new blades, thus no nicks in the blade egde, and thus virtually eliminating the chance of causing damage.

Anyway, this has worked for me since about 1990 when I built my first computer and I've never had a problem.

[vladtepes]

Well I won't fling bovine excrement at you, but I do think that's taking things a bit OTT.
But if that's what you want to do, go your hardest.
I'd suggest to people in general though, that level of precision isn't actually required.

The reason thermal paste exists is precisely to fill small surface imperfections and allowing heat transfer to occur rather than have numerous tiny (and I do mean tiny) air gaps in there.

No off-the shelf product will be so bad that it actually requires sanding prior to use of thermal paste.

[HaveaMint]

[*]Now, do virtually the same with the CPU. Clean it with isopropyl alcohol to remove all surface issues. Apply the TIM using the above described technique, but here, I want to leave enough TIM that I can just barely see through it - just barely make that there is writing on the surface of the CPU.

Please note not to do the sanding part here !

[srq2625]

[*]Now, do virtually the same with the CPU. Clean it with isopropyl alcohol to remove all surface issues. Apply the TIM using the above described technique, but here, I want to leave enough TIM that I can just barely see through it - just barely make that there is writing on the surface of the CPU.

Please note not to do the sanding part here !

^^^^ What he said! Missed that part!

[Faust]

.....
So, why the razor blade and not a bank card? A couple of reasons, but the most significant one being that a razor blade is a finely machined edge - very, very straight - with no gaps or nicks in the surface. Held at a very acute angle to the surface of the cooler mating surface and, again, to the surface of the CPU, there is almost zero chance of causing damage. But, I only use new blades, thus no nicks in the blade egde, and thus virtually eliminating the chance of causing damage.
.....

A single-edge razor blade is most certainly a precision tool , made in stainless steel , and that's what scares me off it .
With a plastic bank card versus the soft copper heatsink surface ? .... there's just no possible chance of any surface damage ,
.... and I don't trust my hands enough anymore to not make a boo-boo , with the "end" of that stainless steel edge .

I'm old fashioned and probably over cautious , but a very smart tech once told me what doctors swear to :-
" First , do no harm "

I take the point about differing methods for applying the goop .... "spotting " etc. , and it may well work better
for unusually large chips / heatsinks .
Most of my experience has been with transfer surfaces about a few centimeters square , and that might be quite small
compared to some modern gaming or coin-mining rigs .

The only certainty in this thing is knowing when there's been too much paste .
ie. when it's squelching out of the edges , and heading for the pins .... NOT GOOD !
Especially if you are using thermal paste with a powdered metal component .
Even if it is not very conductive electrically , it can have a totally berserk capacitance .

So we should be aiming for the least amount that does the job well , as the best of all possible worlds , right ?
I once got so curious that I pulled apart a perfectly good job , just to look at the spread pattern ..... Ha !
It was absolutely ideal , but I cleaned it all off and started again .... a dollar's worth of goop wasted , but valuable info gained .

But different strokes for different folks , I guess .
I strongly support that notion

[151tom]

.

[srq2625]

FWIW.

https://www.youtube.com/watch?v=EyXLu1Ms-q4

https://www.youtube.com/watch?v=r2MEAnZ3swQ


This guy is strange.
https://www.youtube.com/watch?v=VJq-FbhzOtk


Personally I like Arctic Silver 5 although I use white transistor heat sink compound and have zero complaints.

Based on what I saw in those videos (and others that showed much the same thing - and because everything on Youtube is the absolute truth ) I'm going to have to re-think the way I do thermal paste applications. I may have been working too hard when doing this.

But for now, the CPU temps are well in range so I have no complaints.

[151tom]

.

[lsemmens]

And all I ever do is place a few dots on the back of the CPU put the heatsink on it, move it around a little and clamp it all up. Never had an issue in over 30 years.

[151tom]

.

[catweazel]

People make it out to be a bigger task than it is.

Many years ago I worked in a guided missile construction and test establishment. One walks into a concrete bunker filled with 800 tons of highly explosive Octol with great trepidation, bated breath and a loud heart-pounding in the ears. It was very much like that the very first time I re-pasted a CPU. I think this tutorial goes a long way to helping newcomers step up.