The myth of "too much thermal paste"

[wumpus]

I recently replied to this thread, where:

With the case off its about 5-10C cooler. However, my mobo was halting with the case off because the CPU got to 100C with the Nexus AXP3200.

The temperature I reported was 75C. Even now as I sit its 80C with my stock fan, and I'm just sitting in BIOS (i.e. not running anything). I read on other forums that Chaintech measures on-die temp, while most other mobos report externel. This was supposed to account for the extra 10-15 degrees.

Well, I recently installed a new SLK-900 heatsink, and this exact same thing happened to me. Like I always do when installing a new heatsink, I hand-adjusted and eyeballed the fit to make sure the HSF had good contact. It did.

However. After booting up, the machine shut itself down in about 30 seconds. Weird, I thought. Then I went into the BIOS and checked the CPU temps: 70c at idle! That's strange, I thought: I know I mounted the heatsink correctly on all 3 lugs on each side, and I oriented it the correct way-- with the wide edge aligned with the wide edge of the socket.

I am not a newbie by any means, I've been building computers for years and years. The source of this particular problem, I think, was trying to "recycle" the existing ceramique thermal paste on the die. In other words: not enough thermal paste. Even though I scooped up as much as I could with a thin plastic card from the old heatsink, and transferred it to the surface of the CPU where it LOOKED like a sufficient thin layer, it wasn't.

When I removed the heatsink to diagnose the problem, the small amount of thermal paste had spread into a spotty, thin film on the heatsink bottom and CPU surface. I applied a generous amount of new paste and re-mounted the heatsink the same way I did before and-- poof! problem solved. Back to ~40c idle in the BIOS.

This is why the commonly offered advice to use the thinnest layer of thermal paste possible is MISGUIDED and possibly even dangerous.

What's particularly silly about that advice is, even if you use way too much thermal paste, the pressure of a correct HSF mounting will squeeze all the excess out anyway. At worst, all you've done is make a mess... whereas if you use too little thermal paste, your CPU will quickly overheat to 100c BIOS shutdown levels, as mine did! Or, heaven forbid, your BIOS can't detect the die temps and you overheat or damage your CPU. The risk of using too little paste is far more real and dangerous than the myth of too much thermal paste.

[sthayashi]

I have to respectfully disagree due to the exact OPPOSITE experience with my Shuttle SN41G2 w/ Athlon XP 2400+ and Arctic Ceramique.

I put some on the way I normally did, and the damn thing overheated before I got to the BIOS. I put on less and less until finally I had nearly nothing on there. THAT got it to work.

Even the Ceramique instructions on Arctic Silver's website say that it's supposed to be about as large as a grain of rice. My problem is that the Ceramique is always stringy and its too easy for me to put too much on.

[wumpus]

Well, to be honest, I've done some ultra-sloppy installations in my day. Many of them included a rushed "hurry up and install my new CPU so I can play with it" amount of thermal paste, eg, too much. I figured being anal didn't matter since, as I mentioned above, any decent HSF mounting will have enough pressure to squeeze out the excess paste anyway.

And I have never, not once, had a problem with temps on any of those types of installs.

On the other hand, the few installs where I have attempted to use a razor-thin level of thermal paste for whatever reason-- in this case because I was being cheap-- I've had problems with excessive heat. I remember at least one other time I had to do this, go back and install more thermal paste to resolve a temperature problem.

However, I've never had to remove paste. Not once in, geez, at least 30 computer builds over the last 5 years.

[Edward Ng]

I agree on this: Putting too little is worse than putting too much.

I disagree on this: Putting less is a myth.

Understandibly if you've too little on there, you've too little. That's no good. It's like not having any. Having too little does not mean putting, simply, "less." It's putting not enough.

Once you've gotten to the point of understanding exactly how much is JUST ENOUGH, and you are skillfull enough to put no more than JUST ENOUGH, you will achieve the best temperatures you've ever seen yourself achieve. It takes time, it takes practice and most of all, it takes lots and lots of patience!

Even now, after having done about four or five sinks since I started my, "perfect amount of TIM," initiative, it still takes me about 30 to 45 minutes to do the sink for a P4.

Done right, the most minimal amount that is sufficient will always outperform an excessive amount. You can try pressing or squeezing it all you want; no amount of squeezing in the world will be as effective as applying the right amount to begin with. And you know what? Pressing it out with massive amounts of force will work fine on a P4 or K8 system, but if you do that to your K7 (socket A) without a shim, you just wrecked your chip, buddy.

Hey look, I don't mean to be harsh, and I mean no offense, but I find it rather dangerous to recommend pressing with enough pressure to squeeze it down to the right amount when there are people out there who own Socket A chips that don't yet know better and try pressing the sink down to squeeze out excess TIM (without using a shim) and, in the process, break their chip.

-Ed

[wumpus]

Hey look, I don't mean to be harsh, and I mean no offense, but I find it rather dangerous to recommend pressing with enough pressure to squeeze it down to the right amount when there are people out there who own Socket A chips that don't yet know better and try pressing the sink down to squeeze out excess TIM (without using a shim) and, in the process, break their chip.

You don't have to "press" down at all-- most modern heatsinks use a fairly extreme level of mounting force all by themselves! What I was referring to was a little wiggle and mild "reality check" press to make sure the heatsink is truly mounted.

For example. When I removed the AX-7, hardly any thermal paste was left in the square contact patch between the CPU and the heatsink. There was, in fact, a square outline where the CPU block used to be-- that is, 95% of the paste naturally squeezed out due to the standard force of mounting.

Being obsessive is fun and all, but sometimes there's no point. The risk is too great in this case, and ultimately the "perfect" amount of paste doesn't matter: the excess is always pushed out by a proper mounting.

I suppose if you had a HSF that had an ultra-loose mounting, you could theoretically have too much paste between the heatsink and CPU surface, thus increasing temps a few degrees c (though nowhere near what you'd get with no paste, or not enough paste). I'm just not sure in this day and age of 100w CPUs that any "low mount force" heatsinks even exist!

[Edward Ng]

That also depends on the consistency of your TIM. The white crap that came with my Vantec Aeroflow will press out with ease, but the Arctic Silver V I have is far more viscous and it would take massive pressure to squeeze it out even close to as thin as it ought to be.

-Ed

[xarope]

in my personal experience, I first applied arctic silver v to my new K8T w/athlon64 3200+, and was getting temps around 43'C idle. After a week or so to ensure it was stable, I cleaned off the AS-V and the CPU IHS+HS, then re-applied with a much thinner layer (the recommended "grain of rice" amount rather than the "what oozed out when I first opened the AS V tube" amount!), now my idle temps are around 40'C. Note that my ambient temp is prob. around 30'C since I live in Singapore, and the study/PC room windows are usually closed during the day when my wife and I are out.

however, I may agree about the generic brand white stuff the vendors give you, they tend to get squished out really easily as opposed to AS V.

[wumpus]

I kinda switched away from the arctic silver due to its weird, watery consistency. I like Ceramique because it has similar performance without the watery consistency-- and that's what I've been using.

The ceramique definitely squished out on my AX-7 mounting example, above. But that's what I would expect to happen with correct, modern high pressure HSF mounting designs: only what is necessary remains.

[Edward Ng]

This is the fastest thread ever.

Anyhow, back to the topic at hand...

Arctic Silver V, from what I've seen, is much more viscous than my tube of Arctic Silver III. The AS V also does not separate at all; by watery consistency, do you mean AS III has a separation tendency? My AS III was never loose or runny enough to call watery; if anything, the cheap white junk that comes with Aeroflows is much closer to watery than AS III, but I can see how AS III would separate over an extremely long period of time (much longer than I'd personally go before hitting up the goop again); the AS V does not even show a hint of such a possibility.

-Ed

PS As a matter of fact, it's very likely that separation was one of the issues dealt with going from AS III to AS V.

[zak]

Indeed the 'clip force' is quite large these days. And expecially with the small surfae of a bare K7 chip, excess paste will be squeezed out without problem.

To it looks optimal if both sides are "wetted" with paste to fill pores, and a small blob is put in the middle that will be squeezed out, thus preventing inclusions of air.

FWIW the clip force of the K8 family is around 30 kilograms, 300 newtons or 66 pounds force, but the surface is so much larger that paste is much harder to squeeze out. This large force is desired to also keep the

If the surface is not 1 x 1, but 2 x 2 centimeters, the excess paste has to travel twice as far, and the pressure per unit area is reduced fourfold.

But paste is not a liquid, and below certain shear force it will not flow at all. So we are left with a quarter of the pressure. It doesn't matter that there is a larger area to push the paste out - the movement will just stop if the shear force drops too far. Thus, twice the diameter may be left with 4 times the paste, at identical force.

This obviously depends a lot on your paste. The more liquid it is (i.e. ifyou push lightly, it will start to flow) the better. A paste that turns liquid if stirred would be excellent too. Overfilled paste that feels uneven, rather hard or clumpy when moving it around is probabbly the worst here.

New clips have their advantages here... compared to the K6 generation - and so has the small K7 core surface.


Thomas

[sthayashi]

Wumpus, I have a question for you.

Did you follow the Ceramique instructions? I don't mean to accuse you of incompetancy or idiocy, but reading your first post, it doesn't quite sound like you did.

Step 1) Clean everything off w/ alcohol or similar.
Step 2) Put some Ceramique on the heatsink and rub it in through a plastic bag
Step 3) Wipe off entirely with a dry, lint-free cloth (coffee filter, etc). Trace amounts of Ceramique will remain, making the mounting surface appear discoloured.
Step 4) Put a small amount on the die (the rice grain size).
Step 5) Mount heatsink.
Step 6) Wash your hands with dish detergent.

[Edward Ng]

The small surface on K7 cores is great; the fact that the core is so small and the empty space underneath the core in the hole in the socket is absolutely NOT great; it's greatly bad. Ideally the space underneath the core would solid to support pressure from above on the core, and thermal measurements done from an on-die sensor.

This is, unfortunately, not the case for K7. I will not agree with anyone who thinks it possible to apply an excessive amount of AS V to a K7 core and then press the sink down on it without a shim with enough force to spread it out as thin as it would be when applied optimally without breaking the chip packaging and thus rending the poor chip useless...

No way...

No how...

No.

AS III? Maybe. The white junk that came with my Aeroflow? Most likely. The AS V in my little tube here, sitting on me desk? Absolutely, positively, no way.

-Ed

PS Like any other mass-produced chemical product, there will be inconsistencies between batches. Perhaps my batch of AS V is a little, stiffer, than normal, but I doubt it's thicker on orders of magnitude.

[wumpus]

Did you follow the Ceramique instructions? I don't mean to accuse you of incompetancy or idiocy, but reading your first post, it doesn't quite sound like you did.

(.. excruciatingly detailed steps 1-99 omitted ..)

See, that's the problem with discussions on SPCR. No matter how anal you may think you are, there's always someone there ready to raise the bar even further

No, I didn't follow the thermal paste installation instructions. And I'll just get this out of the way pre-emptively: I don't wear a grounding strap when working on my computers, either.

I know. I'm a madman, throwing caution to the wind, with a flagrant disregard for procedure!

[Edward Ng]

I've never worn one of those annoying things, either; keeping yourself properly grounded by hitting off a grounded piece of equipment (like the PSU of your other computer) is just as effective, if not better, since you're not all bound and tangled up by the stupid groundwire armband tripwire.

But, I am a firm believer in doing things the right way where it's the only way to get certain results; wearing a grounding armband and keeping yourself grounded by some other means will have the same results, but the only way to get the temps of a proper application of TIM is, well, a proper application of TIM.

Hey, it's pretty much proven that the difference between a poor and excessive amount of TIM and just the right amount, applied properly thin and evenly, is somewhere around 3-5C. The difference between worthless garbage they attempt to sell as TIM vs. good stuff like AS V can often be greater in difference, like 5-7C. If the difference doesn't matter to you, or your temps are already low enough to keep you happy as it is, then that's fine. If you need to squeeze out every last tiny bit of possible performance out of your heatsink, the proven method is an impeccable application of high performance TIM.

Once again, no offense, and I do not mean to be harsh.

-Ed

EDITS: Grammar, spelling and the such.

[dukla2000]

Edward Ng & zak - you folks patently have miles more expertiese in this area so I would appreciate your critique of my method, please.

1) I dont have temp problems, just if I am not at my lowest temps then I want to learn to get there.
2) Between AS (original), Alumina and now AS5 I have never noticed a significant temp change. Which is the reason for the post: all those lovely paste reviews show up significant differences!
3) All my CPU are AMD (small slug)
4) On removal of the hs I wipe the hs & CPU with tissue paper to remove remaining paste.
5) I shine the slug with a jewellers cloth
6) I check the slug & heatsink for no dust etc
7) I place about a 2-3mm long 'line' of paste directly from the syringe lengthwise down the middle of the slug. Actually with AS5 being more 'liquid' I am finding 1-2mm of AS5 is 'enough' for me. This seems to be a lot less than a 'rice grain' but depends on the variety!
Replace the hs

Whenever I remove the hs there is always a clear 'footprint' of the slug on the hs with raised 'mounds' around the perimeter. With AS & Alumina the residual paste was more visible in the slug rectangular 'depression' on the hs: with AS5 the residual paste in the depression is noticably less.

Based on this visual inspection when I remove heatsinks I figure I am doing OK, except feel free to start retraining

[Ralf Hutter]

I disagree on this: Putting less is a myth.

Understandibly if you've too little on there, you've too little. That's no good. It's like not having any. Having too little does not mean putting, simply, "less." It's putting not enough.

Once you've gotten to the point of understanding exactly how much is JUST ENOUGH, and you are skillfull enough to put no more than JUST ENOUGH, you will achieve the best temperatures you've ever seen yourself achieve. It takes time, it takes practice and most of all, it takes lots and lots of patience!

Even now, after having done about four or five sinks since I started my, "perfect amout of TIM," initiative, it still takes me about 30 to 45 minutes to do the sink for a P4.

Done right, the most minimal amount that is sufficient will always outperform an excessive amount.

This is the philosophy that I follow too, developed from hundreds and hundreds of applications of TIM.

I'd rather have the correct amount for the job, not just plop a big blob onto the die and then have it squeeze out all over the CPU when you bolt the heatsink up. That's messy, and potentially dangerous with the conductive TIMS like ASIII and AS5. It's not so much an issue with Ceramique (which is my TIM of choice these days) but it's still not too difficult to apply the correct amount. Practice definitely does make perfect however.

[sthayashi]

See, that's the problem with discussions on SPCR. No matter how anal you may think you are, there's always someone there ready to raise the bar even further

No, I didn't follow the thermal paste installation instructions. And I'll just get this out of the way pre-emptively: I don't wear a grounding strap when working on my computers, either.

I know. I'm a madman, throwing caution to the wind, with a flagrant disregard for procedure!

I don't wear a grounding strap either, but I follow the instructions right down to the letter, even going so far as to clean off the heatsink re-rub Ceramique through a plastic bag. I've kept a plastic around my desk just for that purpose.

In fact, the hardly any Ceramique that I mentioned previously about my SN41G2 was simply the stuff that I had rubbed into the heatsink.

All this discussion makes me want to take that machine apart again and reinstall the heatsink with varying amounts of paste and take notes and pictures to see what the right amount should be..

[sneaker]

Wumpus, I have a question for you.

Did you follow the Ceramique instructions?

No. He scooped used Ceramique off his old heatsink to re-apply to the CPU for use with a new heatsink, if I understood correctly. Maybe that's OK (I doubt it), but in any case you won't find that conservation tip in the instructions

My own opinion, and everyone seems to have one on this issue, is this--a thin layer is best as long as it's applied carefully to cover the whole core. If you similarly apply a thick later, a lot of it will just get squeezed out the side. And if you just stick a big blob in the middle, you're relying on the heatsink to do the distribution, which will likely be uneven and also result in some of it being squeezed out the side.

There is often a desire to use a thicker later just to "make sure" there's enough, but that's probably only necessary if you're in a hurry or use an awful quality heatsink or have some other wonky aspect in your installation. In a typical setup the space between the heatsink surface and the CPU core is so tiny that, even if you apply a really thin layer of thermal compound, some of it will still get squeezed out the edges. Shine a torch after installing the heatsink and see for yourself. Indeed that's how you best check for a decent installation, IMHO. Install the heatsink with the board out of the chassis, then examine the space between the heatsink and CPU to check the contact (capacitors et al will get in the way but you should still get a reasonable view).

Note that these comments are based on my experience with CPUs where the contact area is small (e.g. Athlon XP). With P4 and Athlon 64 you can probably get away with a sloppy job more often (check out video 10 at http://www.tomshardware.com/site/videos/ for a slightly amusing example of a hurried installation).

[smifou]

in my own experience. i used lots of thermal paste with arctic silver 5 and i got really high temp, a thin layer gave me perfect low temp. i think its cuz arctic 5 is less liquid so it squeezes out less.
when i had arctic silver 3.. i could use lots of thermal paste and it would squeeze out but i hated this paste as it bleeds and after a few months.. i was back on high temp

[Persio]

I have just received my Zalman 7000A-Cu kit and reading this thread I thought I'd ask if anyone can share his experience with the Zalman thermal paste included in the kit.
I intend to use it if possible.
Thanks.

[wumpus]

when i had arctic silver 3.. i could use lots of thermal paste and it would squeeze out but i hated this paste as it bleeds and after a few months.. i was back on high temp

Me too, which is why I switched to Ceramique.. much better consistency. By that I mean more like the regular generic white goop, but performance even better than AS.

[Mr. Thompson]

I am not a newbie by any means, I've been building computers for years and years. The source of this particular problem, I think, was trying to "recycle" the existing ceramique thermal paste on the die. In other words: not enough thermal paste. Even though I scooped up as much as I could with a thin plastic card from the old heatsink, and transferred it to the surface of the CPU where it LOOKED like a sufficient thin layer, it wasn't.

wumpus - What you have is a contamination problem. Any dust that has blown through your case and become lodged in the boundry layer of your old thermal grease installation is now smack dab in the middle of your thermal interface.

Colin Thompson
Arctic Silver, Inc.

[Edward Ng]

WHOA it's the Arctic Silver dude!!!

Can I have your autograph?

-Ed

[flyingsherpa]

don't forget that Dell uses NO THERMAL paste at all and they have tons of pretty quiet computers out there that aren't overheating.

i think the problem wumpus had at first was most likely from recycling the old paste. i don't know exactly what TIM is made of, but most people talk about it "setting" or "breaking in" over the course of a few days before it delivers the best results. which sounds to me that, as it heats up, it probably spreads out a little and possibly changes its properties somewhat (not quite 'curing', but maybe something like it). so re-using paste would be a really bad idea.

someone please correct me if i am wrong, but that would be my guess.

[HammerSandwich]

Every time I think about squeezing the TIM layer out, another picture pops into my head: the crankshaft and connecting rod in an engine. There is no actual contact between these parts (yes, yes, nor the bearings...) in a properly running engine - the rod always is supported by a thin film of oil. I'm confident that Ceramique at 50C is more viscous than any motor oil at 90C, and I know the pressures are higher in an engine.

[wumpus]

wumpus - What you have is a contamination problem. Any dust that has blown through your case and become lodged in the boundry layer of your old thermal grease installation is now smack dab in the middle of your thermal interface.

Possibly, but the root problem was not enough paste, since re-applying more put temps back where they were. No special cleaning occured, I just squirted a bit more ceramique on the die.

I'm sure from a vendor perspective, recycling thermal paste is a VERY VERY BAD IDEA indeed.

All this hand-wringing over perfect thermal paste application is especially comical given the setup of the P4 and the A64, where there is an integrated heatspreader and (probably) mediocre thermal interface material under the heatspreader that we can't even get to. Yes, god forbid we get a few MOTES OF DUST in our TIM layer. Temps might go up by 0.005 degrees C! I think I hear klaxons sounding already!

[Mr. Thompson]

This is not a guess or a ploy to sell more thermal compound, it’s a fact that we have observed in laboratory conditions with test equipment that has far more resolution than the sensors in a CPU or motherboard.

The average particle size in Ceramique is less than 0.38 microns. The particle size range for common household dust is from 1 to 30 microns, the particles you can see are 50 microns or larger, and some pollens approach 1000 microns. It doesn’t take a rocket scientist to figure out that on this scale, a “few motes of dust” can significantly affect a thermal interface.

[al bundy]

Hi Mr. Thompson,

Thanks for advising about the dust issue. That makes sense.

Could you also please comment on another issue presented here, namely: Whether being careful about the initial thickness of TIM on the processor might not really matter, since the excess TIM just gets pressed out when applying the heatsink anyway, etc...?

[fmah]

I suppose Dell doesn't use thermal paste, but they must be using a thermal pad if not paste. The pads are quicker and easier to deal with for volume manufacturing.

Really what you want is an ideal contact between the two surfaces, and just enough grease to fill any voids. If you put too much and don't have enough pressure, you could end up with a thicker layer than you wanted.

[Mr. Thompson]

al bundy - Great user name.

Starting with a thin layer is best. Given a flat CPU and well finished heatsink, a translucent layer is all that is necessary.