really proper way to apply thermal compound?

[Ian.M]

Take an old credit card or store card and scrape a trace of compound over both surfaces to fully wet them.  Wipe off any excess at the edges.  Apply a dot in the middle (rice grain size for most power transistors, dried split pea sized for large ICs like old Pentium CPUs) and assemble with manual pressure and a slight orbiting motion till you see squeeze-out all the way round.   Fit the retaining clips or screws before releasing the pressure.  If possible clean up the squeeze-out excess with a paper towel or a Q-tip moistened with thinners.

If you pretend you are Scrooge and the paste is loaded with gold dust you'll get the quantity about right - meaner than a Scottish miser spreading Marmite - as the thinner the layer the better as long as its void-free.

[wraper]

People obsess over methods of thermal compound application but real word differences seem minimal to non-existent.

This.  If it were really important, they wouldn't give thermal compound, CPUs and heat sinks to gamers and trust them to not screw it up.

If it's a bare silicon die, it's important. From my experience I can say that I've seen GPU shutting down due to local overheating (but showing normal temp in monitoring app) which happened because of entrapped air bubble (just after paste application). If paste is spread over surface, some small air bubbles are almost guaranteed.

[Mr. Scram]

If it's a bare silicon die, it's important. From my experience I can say that I've seen GPU shutting down due to local overheating (but showing normal temp in monitoring app) which happened because of entrapped air bubble (just after paste application). If paste is spread over surface, some small air bubbles are almost guaranteed.

How did you make sure that was actually the problem you saw? Just reapplying thermal pastes leaves room for many other issues to be corrected at the same time.

The problem with this discussion is that it's surrounded by a lot of myth and hearsay whereas the little repeatable actual evidence that exists shows pretty much all of that is irrelevant. More evidence is very welcome, but we should be very mindful about what is accepted as such.

[Mr. Scram]

whats interesting about CPU is the fact that you can replace the paste between the die and the lid of the IC in a modern intel chip to improve its performance. surprising for 1000$ ic.

but seriously, what about spin coating? Its used for extremely even distribution of photoresist during transistor manufacture.

Most of Intel's higher end offerings once more come with soldered IHSs.

What about spincoating? It isn't currently remotely applicable to CPU TIM.

[wraper]

How did you make sure that was actually the problem you saw? Just reapplying thermal pastes leaves room for many other issues to be corrected at the same time.

Because it was newly applied thermal paste and there was an air bubble. When you are dissipating 250W of heat at 95^oC die temperature over 4cm² without any heatspreader, it matters.

[Mr. Scram]

Because it was newly applied thermal paste and there was an air bubble. When you are dissipating 250W of heat at 95^oC die temperature over 4cm² without any heatspreader, it matters.

That's a bit anecdotal to be honest. By reinstalling the heatsink you change many variables. Maybe the pressure or mounting was different, who knows.

[wraper]

Because it was newly applied thermal paste and there was an air bubble. When you are dissipating 250W of heat at 95^oC die temperature over 4cm² without any heatspreader, it matters.

That's a bit anecdotal to be honest. By reinstalling the heatsink you change many variables. Maybe the pressure or mounting was different, who knows.

You cannot make a different pressure with that heatsink and springy backplate on the opposite side. Pressure applied is a constant amount, period. And how pressure would matter to begin with? And I say it again, there was a freaking air bubble when I disassembled it  .
EDIT: And that graphics card would not go over 95^oC which is target temperature. It would increase fan speed and throttle.

[langwadt]

A good way to ensure even coverage and exclude air is to apply the thermal paste in an 'X' pattern then clamp the heatsink on.
This will make sure the thermal paste reaches the edges but won't capture air.

yeh just like when laying tiles using a notched trowel always make parallel lines so you don't get any closed pockets of air 

[coppercone2]

people will desolder their processor lids and snap off glue with special jigs. look at the process on youtube, I have seen a high end AMD processor go into a pizza/toaster oven.

[Mr. Scram]

You cannot make a different pressure with that heatsink and springy backplate on the opposite side. Pressure applied is a constant amount, period. And how pressure would matter to begin with? And I say it again, there was a freaking air bubble when I disassembled it  .
EDIT: And that graphics card would not go over 95^oC which is target temperature. It would increase fan speed and throttle.

I don't have trouble believing that you saw an air bubble upon disassembly. The issue is the implied causation based on a single uncontrolled observation. That's tentative in the best of circumstances and probably worse in a matter where discussions tend to be endless in lieu of hard repeatable evidence.

[wraper]

You cannot make a different pressure with that heatsink and springy backplate on the opposite side. Pressure applied is a constant amount, period. And how pressure would matter to begin with? And I say it again, there was a freaking air bubble when I disassembled it  .
EDIT: And that graphics card would not go over 95^oC which is target temperature. It would increase fan speed and throttle.

I don't have trouble believing that you saw an air bubble upon disassembly. The issue is the implied causation based on a single uncontrolled observation. That's tentative in the best of circumstances and probably worse in a matter where discussions tend to be endless in lieu of hard repeatable evidence.

It's quite obvious that heat transfer will suck through air bubble. And with such high amount of heat dissipated from small area it's obviously will have severe effects.

[coppercone2]

can you do a ultrasonic reflection test?

[Mr. Scram]

It's quite obvious that heat transfer will suck through air bubble. And with such high amount of heat dissipated from small area it's obviously will have severe effects.

The videos provided before show that these kind of common knowledge factoids in reality don't have the impact some people purport them to have. Things like the method of application, partial TIM coverage or too much TIM turn out to cause differences well within the margin of error. This means the results are effectively the same across the board. Even though conventional knowledge says these factors should make a difference, they don't appear to do so.

That makes it not as obvious that an air bubble would have an effect. The only reasonable conclusion seems to be that there isn't enough information to draw conclusions, although it should be noted that none of the other factors did make any difference. Your anecdotal observation could be the starting point for a more controlled investigation, though.

[wraper]

It's quite obvious that heat transfer will suck through air bubble. And with such high amount of heat dissipated from small area it's obviously will have severe effects.

The videos provided before show that these kind of common knowledge factoids in reality don't have the impact some people purport them to have. Things like the method of application, partial TIM coverage or too much TIM turn out to cause differences well within the margin of error. This means the results are effectively the same across the board. Even though conventional knowledge says these factors should make a difference, they don't appear to do so.

That makes it not as obvious that an air bubble would have an effect. The only reasonable conclusion seems to be that there isn't enough information to draw conclusions, although it should be noted that none of the other factors did make any difference. Your anecdotal observation could be the starting point for a more controlled investigation, though.

Those tests are done with CPUs which have heat spreader, thus bubbles have much less effect. Not to say, you won't see temperature increase in monitoring software if bubble is not above the area with temperature sensor.

[Mr. Scram]

Those tests are done with CPUs which have heat spreader, thus bubbles have much less effect. Not to say, you won't see temperature increase in monitoring software if bubble is not above the area with temperature sensor.

I agree that there may be different circumstances, but without any kind of controlled testing it's quite impossible to say whether a bubble can make these kinds of differences or not.

[james_s]

I just smear a thin even coat with my finger. Too much paste is a common mistake and is worse than too little.

[wraper]

I just smear a thin even coat with my finger. Too much paste is a common mistake and is worse than too little.

Unless it's a conductive crap, too much paste won't do anything bad, excessive amount will just squeeze out. Well, unless you apply so much that it gets into CPU socket. Too little can cause actual problems.

[Mr. Scram]

I just smear a thin even coat with my finger. Too much paste is a common mistake and is worse than too little.

The video already posted shows even a rather excessive amount has no discernible impact. The conclusions really does seem to be that all these hard fought wisdoms of the internet turn out to be made up lines in the sand to fight over. Everyone needs a hobby, I guess.

[james_s]

Too much paste makes a huge mess all over everything, and in the case of power transistors I've replaced failed transistors that had so much paste that they weren't sitting flat on the heatsink. Also those silver types tend to be conductive, I repaired something that was malfunctioning due to conductive paste oozing out on the leads.

[wraper]

Too much paste makes a huge mess all over everything, and in the case of power transistors I've replaced failed transistors that had so much paste that they weren't sitting flat on the heatsink.

If transistors were not sitting flat, it's certainly not because of the amount of the thermal paste. But because they were not mounted properly or were not tightened as should.

I repaired something that was malfunctioning due to conductive paste oozing out on the leads.

You need to buy that crap specially. And those are not even among the best in regards of thermal conductivity of modern thermal pastes.

[coppercone2]

I developed a theory.

The thermal paste is rather abrasive. If you get it into your thread then it will mess with your torque readings. You can't tighten the thread right unless its cleaned and maybe oiled (bit better). If you put gobs of thermal grease over a part that can get into a torque sensitive adjustment screw it might prevent you from clamping.

It should not be an issue with overhead springs though.

Maybe this is how it all started? It seems like you can do it to a TO-220 easily.

Keep in mind standard bolts (not metrological or special use) come in 5 grades of tolerances and probably surface finish (but this is not controlled as well), and I don't know how threads are listed.

If the thread is really ratty and unpolished chances are you will compress a asparity of the metal, which will yield over time and actually reduce spring force. Also even if you tighten by extension you can have some stuff creep out of the threads if it gets in there after a while or from evaporation and cause the torque to change. Also the derivation of torque from rotation requires stable known geometry as a constant I believe to be applicable to bolts. I might check this later

This is why you never use ratty old sealant on nice pipe threads I think. Also, you use teflon tape. Its pretty slippery stuff. Its not aluminum oxide impregnated rough tape.

If you take something rusty apart its a good idea to use new bolts but if not be sure to brush them good to derust and give a slight polish and then lubricate and clean out the thread hole. For small bolts you can kinda wing it by taking a pair of wire cutters to a dremel brush to make a smaller dremel brush. But properly you would wanna use a rethreading tool, or re-tap the hole with a bigger size (re-threader / thread cleaner) is different then a new tap.. or if you need the same size drill a bigger hole and insert a heli spring thread insert.

[Mr. Scram]

I developed a theory.

The thermal paste is rather abrasive. If you get it into your thread then it will mess with your torque readings. You can't tighten the thread right unless its cleaned and maybe oiled (bit better). If you put gobs of thermal grease over a part that can get into a torque sensitive adjustment screw it might prevent you from clamping.

It should not be an issue with overhead springs though.

Maybe this is how it all started? It seems like you can do it to a TO-220 easily.

Keep in mind standard bolts (not metrological or special use) come in 5 grades of tolerances and probably surface finish (but this is not controlled as well), and I don't know how threads are listed.

If the thread is really ratty and unpolished chances are you will compress a asparity of the metal, which will yield over time and actually reduce spring force. Also even if you tighten by extension you can have some stuff creep out of the threads if it gets in there after a while or from evaporation and cause the torque to change. Also the derivation of torque from rotation requires stable known geometry as a constant I believe to be applicable to bolts. I might check this later

This is why you never use ratty old sealant on nice pipe threads I think. Also, you use teflon tape. Its pretty slippery stuff. Its not aluminum oxide impregnated rough tape.

Why are you suddenly going off on a tangent about threads and spring force?

[coppercone2]

I developed a theory.


Maybe this is how it all started? It seems like you can do it to a TO-220 easily.
]

Why are you suddenly going off on a tangent about threads and spring force?

I don't know why you don't understand the context but to explain it in clear detail:

1) There is some crap about using too much thermal grease being bad other then non economic
2) there is a claim insufficient mounting pressure is a result
3) there is a claim this is non sense because it should work and mechanically cannot be explained
4) a partial explanation is offered by the mounting geometry of certain packages being susceptible to contamination if excessive thermal grease is used
5) there is a detailed explanation about the functioning of bolts and screws for those that did not read very boring literature about proper screw function (going off about)

yea im 'going off' maybe try to follow the discussion?

I noticed that people who are not mechanical engineers, and most engineers in general that I spoke to, don't really appreciate the mis behavior of threads from contamination. Plenty of people think "yea its real tight that means its good".

but idk do you say that to your wife when she complains you left your socks on the floor? 'shes going off' ............

[james_s]

Too much paste makes a huge mess all over everything, and in the case of power transistors I've replaced failed transistors that had so much paste that they weren't sitting flat on the heatsink.

If transistors were not sitting flat, it's certainly not because of the amount of the thermal paste. But because they were not mounted properly or were not tightened as should.

I repaired something that was malfunctioning due to conductive paste oozing out on the leads.

You need to buy that crap specially. And those are not even among the best in regards of thermal conductivity of modern thermal pastes.

I didn't mount them, I just cleaned up the mess. These were arcade vector monitors with big TO-3 transistors that screw into flimsy stamped frame sockets.

[fsr]

I just use an old credit card to spread the paste over the entire contact surface. It takes some time, but won't make a mess with the paste when tightening, and seems to work without problems.