Thermal paste patterns

[Flying Fox]

This seems to be a fairly frequent question and I just run into a good opportunity to share some results. Due to me selling the E2160 I can show how I deal with the new fancy HDT and its challenging contact surface.

I think I follow the last method mentioned on this page and after all this time, the following are what I got after removing the OCZ Vendetta HSF:


I would say that the results are pretty consistent and I think I got a good result. With the E6300 coming in I will be following the same method.

Please share more of your war stories (and pictures) and we can build up our own "known thermal paste spreading patterns".

[mortifiedPenguin]

A few months ago I reapplied my TIM using the "spread as thinly as humanly possible with a plastic bag on your finger" method on both the CPU and HSF. It dropped 5-10C from the original application using the "rice grain sized drop" method with a TRUE and an OC'd Q9450. Maybe I just put too much the first time and I don't remember what the exact temps were. I do recall that it also resulted in better contact; I ripped it off to check, resulting in that fractal pattern that shows up when pulling apart two fully mated surfaces. It actually needed some force (albeit minor)to pull off due to some sort of suction (VDW-like forces maybe?). No pictures for now, if I remember to I'll come back with pics the next time I reapply.

[edh]

CPU: Core i7-960 -- Heat sink: Xigmatek HDT-S1283 -- TIM: Shin-Etsu MicroSi G751 (MassCool)

I used a "wipe & triple-stripe" method with my most recent build. I "zip-lock baggied" my finger and wiped both the CPU and the heat sink, making sure that I got TIM paste in all the crevasses. Then I put a very thin stripe along each of the three copper pipes patterned much like that on your example page except that my stripes were a little longer -- probably 3/4ths the length of the contact area of the copper surface (and on the copper surface instead of the aluminum dividers).

[StuG]

What I do is evenly spread it on both the HSF and CPU as little as possible, than put a glob in the middle, and when I apply the heatsink I press down fairly hard, and all the extra just shoots out the sides. I clean up that with a paper towel or what have you, and I've had great cooling performance and results each time.

[Flying Fox]

A few months ago I reapplied my TIM using the "spread as thinly as humanly possible with a plastic bag on your finger" method on both the CPU and HSF. It dropped 5-10C from the original application using the "rice grain sized drop" method with a TRUE and an OC'd Q9450. Maybe I just put too much the first time and I don't remember what the exact temps were.

IIRC the C2Q requires the "line method" to cover the cores better? I wonder if someone has some results to report for the line method.

CPU: Core i7-960 -- Heat sink: Xigmatek HDT-S1283 -- TIM: Shin-Etsu MicroSi G751 (MassCool)

I used a "wipe & triple-stripe" method with my most recent build. I "zip-lock baggied" my finger and wiped both the CPU and the heat sink, making sure that I got TIM paste in all the crevasses. Then I put a very thin stripe along each of the three copper pipes patterned much like that on your example page except that my stripes were a little longer -- probably 3/4ths the length of the contact area of the copper surface (and on the copper surface instead of the aluminum dividers).

Did you pre-fill the "channels" (in between the heat pipes and the separators) with a little bit more paste material?

I also remember that before I locked down the 4 legs, I wiggle the HSF in both clockwise and counterclockwise directions. I might also have done a pair of north-south and east-west horizontal sliding motion as well. Then I lock down the legs.

[edh]

Did you pre-fill the "channels" (in between the heat pipes and the separators) with a little bit more paste material?

Yes, did that with the "baggie-wipe." The MassCool goop I used is more paste-like than Arctic Silver 5.

I also remember that before I locked down the 4 legs, I wiggle the HSF in both clockwise and counterclockwise directions. I might also have done a pair of north-south and east-west horizontal sliding motion as well. Then I lock down the legs.

The heat sink shuffle? Latest dance craze! LOL! Yes, I do that, too. Although I screwed the HDT-S1283 to its metal metal bracket underneath the motherboard instead of using pushpin clips. I don't trust those things except with very light heat sinks (like those provided with Intel's retail packaging).

[Flying Fox]

Did you pre-fill the "channels" (in between the heat pipes and the separators) with a little bit more paste material?

Yes, did that with the "baggie-wipe." The MassCool goop I used is more paste-like than Arctic Silver 5.

In addition to the baggie wipe, I also used those plastic cards they send you in the mail for a straight edge I think.

I also remember that before I locked down the 4 legs, I wiggle the HSF in both clockwise and counterclockwise directions. I might also have done a pair of north-south and east-west horizontal sliding motion as well. Then I lock down the legs.

The heat sink shuffle? Latest dance craze! LOL! Yes, I do that, too. Although I screwed the HDT-S1283 to its metal metal bracket underneath the motherboard instead of using pushpin clips. I don't trust those things except with very light heat sinks (like those provided with Intel's retail packaging).

The HSF dance! That's why I picked the Vendetta because of size and weight. Going heavier I would insist on a backplate too.

Any other doing the line or even the X patterns?

[mortifiedPenguin]

IIRC the C2Q requires the "line method" to cover the cores better? I wonder if someone has some results to report for the line method.

You could be right, I tend to have Core 0 ~10 C higher than the other three. Perhaps I'll try the line method next time and compare. I'm using AC5 though, so I'm not sure I'll be patient enough to just let it cure.

On another note, I suppose the "thin spread method" should be particularly effective on a lapped CPU and lapped heatsink. Maybe I'll do that next time too .

[edh]

I don't have the patience to do the lap dance thing. I figure that by using a baggie to pre-spread ultra-thinly whatever TIM paste I happen to be using (usually AS5, but I switched to the MassCool paste after I started using HDT heat sinks), I fill in whatever irregularities might exist. Then, I use either an "X" or a "||" for regular heat sinks and either "|||" (3-pipe) or "||||" (4-pipe) for the HDT types like my current Xigmatec. With older CPUs and heat sinks -- back in the day when they were smaller, I used the "pea" method, but I still did a pre-spread with a baggie. I've never had a problem with an overheating system.

[just brew it!]

So... does the shape of your thermal paste pattern reveal "the inner you"? Inquiring minds want to know.

[edh]

So... does the shape of your thermal paste pattern reveal "the inner you"? Inquiring minds want to know.

Probably -- but heaven help the person/people tasked with interpreting the meaning. Then again, with the job market like it is, we could be creating a whole new industry....

[Fragnificent]

I've always used the following method, it works great. AS5, squeeze out a little bump about the size of a ball bearing, and use a credit card to evenly spread over entire cpu surface, starting from center, and working out in rectangles. The layer of compound when you're done should be barely visible, and see thru. This has been an awesome method for me so far.

[swaaye]

According to the Arctic Silver folks, the logic behind the bead/line methods is that when you put the cooler on and mush it out, the result will have less air trapped than if you spread it yourself. Also, the critical point of heat transfer on the heatspreader is the area above the core. You want the application as thin as possible over this area and as little air in the goo as possible. The rest of the heatspreader is actually rather unimportant.

With a bare die you are supposed to spread it yourself because you have to get the entire surface covered as best as possible. A small bead might work nicely too though, but you don't want uncovered die.

[fpsduck]

Thanks for sharing.
This is nice! I'll try that method next time.

[Flying Fox]

According to the Arctic Silver folks, the logic behind the bead/line methods is that when you put the cooler on and mush it out, the result will have less air trapped than if you spread it yourself. Also, the critical point of heat transfer on the heatspreader is the area above the core. You want the application as thin as possible over this area and as little air in the goo as possible. The rest of the heatspreader is actually rather unimportant.

Actually AS5 has been talking up the line method ever since we have slapped-together dual/quad cores on the market? A lot of people still talk about the rice/bb pattern based on a little bit older knowledge. Then again, with the advent of IHSes the need to get the perfect pattern has diminished.

[swaaye]

Actually AS5 has been talking up the line method ever since we have slapped-together dual/quad cores on the market? A lot of people still talk about the rice/bb pattern based on a little bit older knowledge. Then again, with the advent of IHSes the need to get the perfect pattern has diminished.

The line method is for the CPUs with very rectangular cores and CPUs with more than one die. They recommend it for P-D, Core 2 and Core ix. When it comes to more square CPU dies, like with AMD stuff, they suggest the bead-o-goo method. Unless of course you don't have a heatspreader, then it's spread it on yourself.
http://arcticsilver.com/arctic_silver_instructions.htm

[CampinCarl]

I usually put on a dallop of thermal paste, and then use an old safety razor blade to spread it as thinly as possible. I have to be careful to not mark the heatspreader/thermal cap/whateveryouwannacallit obviously but I've done a good job so far. My best result was probably my first build--with a ~70F ambient, my A64 3000+ was 77F at idle. I clean with only iso. alcohol.

[SpotTheCat]

I use the line method on newer intel stuff. I run the tube nearly perpendicular to the can so I get very low-profile lines. I also run them as straight and as perpendicular as I can.

[edh]

I use the line method on newer intel stuff. I run the tube nearly perpendicular to the can so I get very low-profile lines. I also run them as straight and as perpendicular as I can.

Perpendicular ( |_ ) ? . . . Or parallel ( || ) ? I'm guessing you intended the latter.

[mph_Ragnarok]

I just put a rice-grain sized scoop in the middle of the CPU, and then the pressure spreads it awesomely.

All you other people are wasting time. It looks exactly the same after the pressure squeezes it out.

[Flying Fox]

I just put a rice-grain sized scoop in the middle of the CPU, and then the pressure spreads it awesomely.

All you other people are wasting time. It looks exactly the same after the pressure squeezes it out.

With the advent of IHSes the need to get the perfect pattern (or the perfect conductive material) has diminished yes. However, there are also 2 new variables introduced over the past few years:
1. Emergence of highly rectangular packages since the introduction of the Pentium D. There is a chance that a circular pattern created by the rice-grain approach may be too small to cover them. Even if you put a larger dab of stuff in the centre, you risk making the centre too thick?
2. Rise of popularity of HDT coolers. If you don't prefill those gaps between the heatpipes and separators, there is a real chance that when the rise-grain spread out, it may stop at the "channels" and you don't get the coverage that you want.

Again, "all you ..." is yet another blanket statement that does not work.

[Neutronbeam]

I may have missed it in the messages above, but isn't anyone lapping heatsinks anymore? Even HDT models? Unless something has a mirror finish [very unlikely] I'm probably going to do a little lapping before applying paste.

[Welch]

I may have missed it in the messages above, but isn't anyone lapping heatsinks anymore? Even HDT models? Unless something has a mirror finish [very unlikely] I'm probably going to do a little lapping before applying paste.

Ohhh man, I hope so The thought of having such a perfectly smooth piece of copper to caress the cap of my CPU is so ohhhhhh...... Ummm.. Opps, hows it going guys!! *Waves as he hides his smooth HS and thermal paste *

I don't see why people wouldn't still lap their HS. If its got the manufacturing groves in it still then it leaves something to be desired. Sure, you might only get a few extra degree's cooler even when using a darn good thermal paste but lets be honest..... That's what separates us, the enthusiast, from the computer "User". When your talking about OCing and all, ill take every little benefit I can get.

Chalk another one here that spreads from a rice grain (even with heat-spreaders I do this) Im used to the old exposed core types like the AMD 32bit Athalon. I don't put the invisible layer like some, that to me is some scary stuff. I put a very even layer that's clearly visible with no metal showing through, not worry too much about the very very edge as the thermal grease expands. I use a credit card to make it looks perfectly even. Thanks Capitol One for all of your useless offers you send that include my new applicators!

[swaaye]

Actually, believe it or not but lapping may actually make things worse. While you definitely want flat surfaces, going for that "crazed sander" mirror finish is apparently not advantageous. Those rough finishes that heatsinks ship with may not always be cheap manufacturing. Arctic Cooling, for example, usually goes with a sort of sandblasted-like rough finish on their coolers.

Here's some talk as to why
http://episteme.arstechnica.com/eve/for ... 1007597931

[Fragnificent]

Lapping is a waste of time...the thermal paste fills in the microscopic crevices in the HSF material regardless. Also, mph_Ragnarok, we're not wasting time...by just applying paste and slapping the HSF on, you're only screwing yourself. Try both methods and measure temps, then come back here and post something useful. I'm far from a genius but I've done this probably close to 3,000 times by now, and yeah it makes a difference how you apply it.

[mph_Ragnarok]

Well, fragnificent, I've tried it as well and I find that it makes no difference in temperatures.

So suck a lemon.

[Fragnificent]

Well, fragnificent, I've tried it as well and I find that it makes no difference in temperatures.

So suck a lemon.

ur doin it rawng...

[Freon]

I generally use the "put as little as possible on both surfaces" method. I use my finger, though, not a plastic bag, but that seems like a good idea so you avoid making such a damn mess and wasting compound in your fingerprint (probably waste more in my fingerprint than is necessary for the CPU!). I'll try that next time. I have the S1283 so I did prefill the gaps as well.

It seems every time I put a dab or three or whatever on the heatsink and let it push itself out it ends up being way too thick. It really requires very little compound if you do it right and if both surfaces are reasonably flat. The high frequency inconsistencies are all you should need to worry about, and those take very little to fill.

[Welch]

And since we are talking about nit-picking anyhow... Your hands natural oils can actually effect thermal conductivity too. So using your finger to spread it sort of contaminates the thermal paste. I always make sure to clean the heat sink and heat spreader with some good 99% Alcohol to remove any other foreign matter/oils from my handling it.

[puppetworx]

Generally I just bukkake the bejesus out of my processor and then slap it with my heat pipe.

I've found the thin line method works best though, worth a good number of degrees Celsius. Lapping showed little difference in comparison with my switch from dollop to thin-line.