Heatsinks: Are they Flat? Answer here?.

[cautery]

Related Threads:
Thermal Tape Selection and Thermal Interface Qualities
http://forums.anandtech.com/messageview.cfm?catid=37&threadid=204693
PIII Coppermine and Heatsinks - quality of the thermal interface...
http://forums.anandtech.com/messageview.cfm?catid=37&threadid=203730
PIII Coppermine Core ? Lapping it flat?
http://forums.anandtech.com/messageview.cfm?catid=37&threadid=203731
FC-PGA Heatsink ?Feet? ? how to get a custom fit
http://forums.anandtech.com/messageview.cfm?catid=37&threadid=203734
Heatsinks: Are they Flat? Answer here?.
http://forums.anandtech.com/messageview.cfm?catid=37&threadid=203739
Thermal Interface Materials: Required Thickness?
http://forums.anandtech.com/messageview.cfm?catid=37&threadid=204688

I have discussed several aspects of achieving a GOOD thermal interface in the Related Threads above, but this post contains a key ingredient to accomplishing this:

Q: Is your heatsink mating surface flat?
A: Generally?. No, it is not.

I have examined a number of commercially available heatsinks: generic (show specials), GlobalWin, Alphas, and even The Arctic Circle, and while some of them are flatter than others, in general, NONE of them meet the generally accepted standard for flatness for use in a thermal interface (which is a runout value of 0.001?/inch).

The solution to this reality is simply to lap your processor flat and smooth? Here?s how I recommend you do it, unless you happen to have a multi-thousand dollar, multi-axis milling machine available:

Materials:
1 each heatsink to be lapped (disassembled from fan, shrouds, etc)
1 each NEW sheets of 3M brand wet/dry sandpaper in the following grades minimum (180, 220, 320, 600, 1500) You can use progressively higher grades if desired and continue progression to polishing compounds if desired.
1 each 8? x 10? piece of quality double thickness (or better) glass.
CLEAN, preferably distilled water. (actually, a high quality, very light grade of lapping oil is preferable)
Several lint free towels?. I use lens cleaning towelettes.

This is the general procedure. There are tons of additional ?anal retentive? measuring tools and techniques that one can use, but for the purpose of this article, you can ?get by? with the above.

Place the glass pane on an absolutely clean and flat surface?. Please make sure you clean off your work surface and remove all extraneous materials. Lapping a top quality heatsink should not be done on a computer workstation cluttered with McDonald?s trash, pop cans, diskettes, Cds, and paper?.

Dampen the reverse side of the 180 grit sandpaper (beginning grade of paper depends on the amount of material that needs to be removed) with a sponge and clean water?.. Make SURE there is no label/price tag on the reverse side of the paper. The damp paper will adhere to the clean glass better, thus reducing the tendency for it to slide around. Do NOT wrap the paper around the glass, or try to tape the paper down, as this will tend to cause the paper to ?bow? up in areas which will radically increase your chances of NOT having a flat lapping surface and causing ?rounding? of the edges of the heatsink?. This would be BAD!!!

Now, place some (amount dependent on the size of sink surface to be lapped) clean/distilled water (lapping oil is preferable since it will reduce the tendency of the sink to ?suck down? to the wet paper as the surface gets flatter?. This ?sucking down? may cause a ?hang? which in turn MIGHT cause you to ?skip? the edge over and across the paper? possibly rounding the edge) in a single ?puddle? in the CENTER of the NEW sheet of paper now located on the glass. It is critical that you use a NEW sheet of paper in PRISTINE condition (no scuffs, or creases?.). Any imperfections in the paper will greatly increase the chances of you getting a less than optimum result.

Place the sink, mating side down, in the center of the water ?puddle?. Grasp the sink with your hand, trying to maintain an even (light) pressure across the entire sink. Now begin making small circular patterns with the sink on the paper. As the sink ?settles in? you can gradually increase the circle size? to no more than 1.5 times the size of the sink, and maintain LIGHT pressure. Too much pressure, and you will ?skip? the sink, plus it?s easier to keep the sink flat with a lighter pressure.

Periodically, check your work, rinsing the paper to remove loose grit and waste material. Clean the sink off, and repeat the process until you have a flat surface. This will be evident as the original surface will gradually ?disappear?, being replaced with a uniform circular pattern of ?scratches? consistent with the grade of paper used. Don?t remove more material than necessary to achieve a uniform, flat surface.

The slower and more patient you are, the better your result will be (eg. Less rounding of the edges, etc.). Repeat the above procedure with progressively finer grades of paper/compound until you get the surface quality desired. Wind up the procedure with a thorough rinsing of the bare heatsink, and clean the mating surface thoroughly with a lint-free cloth and alcohol (use a degreaser then alcohol if you use lapping oil).

Reassemble the heatsink/processor assembly as quickly as possible after completion as the newly lapped surface will immediately begin to oxidize. Use the best non-conductive thermal compound you can get your hands on?.

On FC-PGA heatsinks, you?ll need to remove the feet to use this procedure?..

Hope this helps?. Have fun.

 

[fast_eddie]

Thank you for the very descriptive process. I am getting ready to lap two RDJD K-601 heatsinks for a BP-6 box I am building and your 'retentive' description has answered the last few questions that I have had regarding the best way to obtain the flattest surface.

 

[Goi]

When I lapped my C300A, I just used normal sandpaper all the way up to 1500grit. I taped the sandpaper to my glass coffee table and used the "figure of 8" rule to sand my CPU onto the sandpaper.

 

[randomlinh]

good instructions. Maybe i should give it a try. Do you think the glass from an 8x10 picture frame would work (it's one of those "frameless" ones that have a very thin plastic border, so it looks like all glass)?

 

[cautery]

Inguyen: Most picture frame glass is NOT double thickness. It's imperative that the glass be thick enough not to flex when pressure is applied.

Goi: Nothing personal here.... but this thread is on lapping heatsinks, not C300A cores. Lapping the core is dealt with in one of the Related Threads. Besides, lapping a Cu-Mine core and a Celeron core aren't even remotely the same task. It's vitually impossible to keep sandpaper flat enough by folding or taping it down.... We're dealing with THOUSANDTHS of inches here, and keeping the lapping surface as flat as possible is critical to a good result.

And if you try to get "cute" with figure 8's on a Coppermine core, you ARE going to wind up with a very expensive keychain. Do it the way you want, but my method has been tried and it works well on small cores. So please try not to confuse the readers with a "Well, this is the way "I" lapped my core, blah, blah, yadda, yadda...." comment. Thanks for your understanding...

 

[Frenchie]

Clay:

Long time no see! Good to see you around! How is the Arctic Silver Thermal Epoxy coming?

 

[cautery]

Why, Frenchie... whatever do you mean?

Good to hear from you too, dude!

 

[Descend492]

cautery, you seem to know a little about getting the most out of your CPU (if a little means more than I will ever know), so could you provide some info on exactly how to apply the thermal grease? With a perfectly lapped interface (CPU and HS), you'd think you'd barely need any, but I could be wrong. How think exactly should the layer be?

 

[randomlinh]

do you do the same process to lap the CPU? I'm assuming you do.

Hrm. I don't want to turn my cumine to a keychain How much does a piece of double thickness glass go for?

 

[cautery]

Descend492 - Thank you for the kind words. The thickness of thermal compound required is directly related to the qualities of the mating surfaces. Please read the Related Threads links and especially my comments about thermal interface runout made in the first Related Thread link in my original post to this thread.

The thermal interface compound layer MUST be at least as thick as the TOTAL runout measurement for the thermal interface to insure an air gap free mating of the two surfaces. For instance, assuming that both surfaces are, or have been lapped to the maximum acceptable flatness of 0.001"/inch, then the TOTAL interface runout in 2 x 0.001" = 0.002"/inch. This measurement expresses the maximum possible gap that may exist between the two surfaces (assuming that "valleys" in the two surfaces oppose each other upon mating).

Thus in this instance, you would need a uniform layer of just >0.002" of thermal interface material (compound or tape), and apply sufficient mounting pressure to cause the "peaks" on the two surfaces to meet each other solidly. Thus the thermal compound not actually required to fill the voids will be forcibly relocated to the edges of the interface as "waste". Any less than 0.002" layer runs the risk of leaving a partial void (of air) in the case that two "valleys" oppose each other. Significantly more than a 0.002" layer MAY through various means (not discussed here) prevent the two surfaces from properly mating along the peaks in the surfaces across the entire area of the thermal interface, and less significantly contributes to a larger amount of wasted compound.

Generally, unless the two mating surfaces have been prepared (by hand lapping or machining), they will NOT meet the maximum allowed runout of 0.001"/inch and to the extent that the TOTAL interface runout exceeds this maximum, a correspondingly thicker layer of thermal compound will be required. Thus, I highly recommend reading the Related Links and giving consideration to properly preparing the thermal interface surfaces.

Now as to the actual application of the thermal compound, different compounds require differing methods. As an example, I will discuss the method "I" use to apply Arctic Silver:

1) CLEAN the surfaces thoroughly - If the surfaces are freshly lapped (no thermal compound having ever been applied), rinse thoroughly with water and dry. Then clean the mating surfaces completely with a low residual solvent (isopropyl alcohol will work) and a LINT FREE cloth. I use lens cleaning towels. If the surface has had thermal compound applied previously, it is essential that the surface be thoroughly scrubbed/cleaned with a quality degreaser and then followed with the alcohol cleaning step. It is essential that you keep the surfaces free of foreign materials and that you NOT touch the surfaces (a hair, piece of lint, and even dead skin cells can significantly affect the thermal interfaces performance, especially on PII Cu-Mine and other small cores/surfaces as the surface area is already severely limited. In addition, oils from your finger tips/skin can adversely affect the performance by preventing direct contact of the Silver fill from contacting the metal surface. (Finger prints can be as thick as 0.005" (half the acceptable runout max.)

2) Apply the thermal compound to the SMALLER of the two mating surfaces in the appropriate thickness. DO NOT use your finger to apply or smooth the compound (skin cells, and oils again). I have begun using a modified single edged razor blade as my application tool. You may use whatever tool you choose as long as it is CLEAN and allows you to control the application area and thickness.

3) RECHECK to make sure no foreign contaminants are present on either surface, and assemble the two surfaces. Here is where most people screw up an otherwise GOOD thermal interface:

4) PRESS the two surfaces together ONLY. Minimize at all costs any "twisting" or lateral "sliding" in either plane in an attempt to mate the "peaks" of the surfaces together, and this is why.... When you twist or slide one surface against the other, "peaks" on one or both surfaces may/will travel over areas where two "valleys" should come to rest. The peaks will scrape away compound that is needed to fill the void between the valleys that will oppose each other when the surfaces are in their final position possibly/probably causing small voids (air gaps). ANY air gap will significantly increase thermal resistance in an otherwise GOOD interface. If the thermal compound is properly made and of the proper consistency, direct )non-sliding) pressure vertical to the mating plane will be more than enough to insure that the surfaces "bottom out". Additionally, "sliding" the surfaces together may/will depending on the materials, cause one surface to scratch/gouge the other possibly opening up a larger inter-surface gap which will increase thermal resistance.

5) Secure the thermal interface with the supplied clips or other hardware.

Hope this helps,

PS - I will post this response as a separate thread in my thermal interface series, and update the references.

 

[Descend492]

thanks big time - it helped fill in the small but important details that I needed. You're the man

 

[cautery]

lnguyen - Look at the Related Links in my (updated) original post to this thread and read the thread on lapping the PIII core. Actually, I highly recommend reading them all.... lot's of good info, IMHO.

And you should be able to get an 8 x 10 double thickness pane of glass for a couple of bucks max....

Descend492 - Glad I could help...

 

[Goi]

cautery, i had no intention of misleading users into thinking my way is the "correct" way or "best" way. I DO know that you can't lap Coppermine cores that way, nor would I even attempt to lap Coppermine cores at all given their packaging. All I wanted to share was my experience on lapping, lapping C300As may be slightly irrelavant to lapping heatsinks, but then again I didn't say "do it my way, its better" right? Guess I should have worded it better, or maybe I shouldn't even have posted.
Peace, man

 

[cautery]

Goi - No problem, dude... I could tel that wasn't your intent. I just wanted to try and keep the heatsink stuff here in this thread, and the CPU lapping in that thread to avoid any confusion of the differences in the technique. Re-reading my post, I could have been a little bit more diplomatic in my wording. I can see how my response COULD be misinterpreted. This was NOT my intent at all.

I'm trying my best to get top-notch, accurate info out to those who need it, and I definitely don't want to shut down discussion on the subject... discussion is one of the best ways to learn!

I apologize if my response to your response, to my post rubbed you the wrong way...

Thanks for letting me know,

 

[BlazingSaddles]

Hey Cautery, you said "Minimize at all costs any "twisting" or lateral "sliding" in either plane "... What should I do with a golden Orb then? You have to twist it so it will lock in place... Also, how much of a temp difference will fingerprints and skin oil make? I used my fingers to spread the thermal compound on the CPU. hehe.

 

[cautery]

rking - Will respond in the Thermal Interface Materials: Required Thickness? thread: http://forums.anandtech.com/messageview.cfm?catid=37&threadid=204688

Thanks,