Heatsink Lapping

[S!X]

I just heard about this technique on the ncix.com forums and wonder if any of you guys have done it, and is it worth doing?

http://www.overclockersclub.com/guides/heatsinklappingguide.php

[Virtualbody1234]

Notice that the article is from 2002.

The gains from this are far less now than when the article was written. Processors now have much larger surface area for the contact point between the heatsink and the core (compare images below). Also the newer smaller technology and lower power consumption designs generate less heat.

http://www.pandoragraphics.com/ebay/pics/amd.jpg http://img5.pcpop.com/ProductImages/Leader/0/164/000164047.jpg

Notice the (old) tiny contact point (left center) vs the (newer) full area contact surface.

[suprafreak6]

now thats some improvement

[Appzalien]

None the less, getting your heat sinks surface to a mirror finish can cool your cpu 5-10 degrees C more than the typical sink the way it comes from the factory. I have seen it with my own eyes. I have pulled a heat sink that was running at 52 C, polished it and got 44 C afterwords and I didn't even go all the way to mirror. You can find some techniques online by googling "heatsink lapping" but basically you need a glass plate (like from a picture frame) wet/dry sand paper varying from 100 to 2000 grit (the grit in the 1000's is hard to find and expensive) and some time. The best I have done (52 C to 44 C) was done with a final grit of 600. The best technique I have found is to sand on the plate starting close to yourself and pushing away in a strait line then lifting come back towards yourself and start the next pass, I tried sanding in a circular motion but got less pretty results. You definately do not want to sand pushing away and again back towards yourself, the back and forth rocking can cause irregularities. If you can find a glass plate at least as large as the sandpaper you get, you can make the paper stick to the glass when its been wetted, and it will stay in place fairly well as you push the sink across the paper holding it flat and in the same position all the time. Not only will this give you a smoother surface but will take out and irregularities in the level of the surface of the sink. Most times I have done this the sink sands at the outer edges of the square sink first, leaving a round space in the middle of the sink untouched. The worse this effect is the more important it is to start with a coarser grit to attain the level before going for shine. In other words it can take 6 hours of sanding with 600 grit paper to clear the space in the center of an off level sink, but using 80 grit or 100 grit will clear the space in 10 or 20 min. and then going to 200 grit then 400 grit then 600 grit over the next hour will finish it off. 600 grit will not give you a mirror finish, you need the grit in the 1000's range for that, but I did see one tutorial where the guy suggested a final polishing with just plain paper to get the shine. I tried it but it seemed it would take too long to achieve anything worth the effort. I do want to try using a liquid polish (car polish maybe) on paper to see if I can get the effect of 2000 grit some what cheaper.

[spirex]

automotive stores usually have sand paper with grits in the 1000's. you can also lap the heat spreader on the cpu or gpu but that requires alot of prep work for gpu's.

[clocker]

Sanding to the point of a mirror finish is a total waste of time.
The object of the exercise is to achieve a flat surface, not a shiny one.
Most quality heatsinks are already flatter than is easily achieved at home.

BTW, this procedure is NOT lapping.

Lapping refers to abrading the two mating surfaces together to achieve optimal surface contact. Thus, instead of sanding each piece on a glass-backed piece of sandpaper, you'd apply lapping compound (available in a variety of grits) directly to the CPU's IHS and rub it directly against the bottom of the heatsink.
This would result in a perfect match between the two surfaces, so perfect in fact that, taken to the extreme, it's possible to create a vacuum seal between the parts. You'd be hard pressed to separate them without destroying one of the parts.

If you are still determined to sand these parts, I'd suggest you start with the CPU- not the heatsink.
The IHS on today's chips is not especially well finished and would probably repond better to touch up work than a decent heatsink.
Early generation Conroes are notorious for having concave heatspreaders.

[S!X]

I finally got this thing and installed it, but being that its pretty huge and my case is TINY everything is hitting in like my HD and PSU. I cant even assemble my case back together cause the HD will be in the fan on the freezer. Im stumped on what to do.

E/ There actually is an alternate way of mounting a HD in this case which is to fasten it to the bottom of the case, the only bad thing about this is its causing a fair amount of case vibration, a lot more then it had being setup the other way. Now my question is, is there such thing as silicone washers or anything like that which I could put between the HD and the case to dub down the vibration?

[zapjb]

Bring the screws to an auto parts store. They'll have plastic washers or rubber hosing that should work.

[S!X]

Alright, one other thing that id like to mention is ive noticed the base of this cooler doesn't over the the whole CPU diode. From reading review on it, they claim that its normal. Now should I be concerned about that?

E/ If you guys have any other ideas on how to silence my harddrive by all means shoot some ideas. The whole suspending it idea wont work because of where its mounted to. Here is a pic just to show you what im talking about http://img343.imageshack.us/img343/1206/img004cy9.jpg

[clocker]

E/ If you guys have any other ideas on how to silence my harddrive by all means shoot some ideas. The whole suspending it idea wont work because of where its mounted to. Here is a pic just to show you what im talking about
Do you move your case frequently?
If not, simply laying the drive on a couple of these pads (http://www.petrastechshop.com/pegviabbl.html) will do the trick.
You could use Velcro to make the drive even more secure should you find it necessary.

Be sure to enable SMART monitoring of your drive and keep an eye on the temps.

[lynx]

You could use plastic P-Clips as an alternative.
http://base.google.co.uk/base_media?q=hand4808370447738430928&size=1

Cut off the 90 degree bend (on the right in the picture) and make a new hole. You've now got a flexible U-shaped bracket.

Now fasten one side to the drive using the standard fixing screws, and fasten the other side to the floor of the case using a small screw and nut.

The natural springiness of the plastic should dampen any vibration, and at the same time it will be fairly secure.

[S!X]

Some nice ideas, I'll check into all of em and see whats easiest to find. I ended up remove the cooler and switch the locking mechanism around but even when its vice versa it's still in the way because the locking lever in way of the PCI connector on the mobo :( It took me atleast an hour if not more to put it back on. Clocker any comments on that?

[S!X]

Just got the badboys for the HD, http://www.silenxusa.com/productcart/pc/configureprd_z.asp?idcategory=17&idproduct=55

It looks like Iill be boring out the holes to get the screws to fit :dry:

[zapjb]

Imo you don't need those. I'd get a mousepad free or at the dollar store. Put the HDD on the mousepad.

[S!X]

Imo you don't need those. I'd get a mousepad free or at the dollar store. Put the HDD on the mousepad.

These things are awesome... almost completely silenced the HD, I can barely hear it. It's cut down the vibration a lot. :cool:

[S!X]

Sorry for this mega bump, after reading this lapping guide (http://www.modsynergy.com/article6.htm) im determined to lap my AC64 because the contact base of it is very rough which surprised me when I got it. Now I have to go up to canadian tire and get some sandpaper and metal polish unless there's an alternative way to polishing copper.... ah. I'll report back on the results after all is said and done.

[clocker]

Do NOT use metal polish on your heatsink.

[S!X]

Do NOT use metal polish on your heatsink.

:( I already did. It's "Blue Magic Liquid Metal Polish" which claims to be safe and nonabrasive. Here's the only decent link I could find for it (http://www.autobarn.net/ch200.html). On a lighter note, the lapping did help a bit. The idle temps gone down about 6 or 7 celcius , im not sure about load though.

[clocker]

The idle temps gone down about 6 or 7 celcius , im not sure about load though.
Either your heatsink's surface was warped beyond usability or something else is in play here- such a dramatic difference is just not possible by simply lapping.

Probably a combination of mount pressure, TIM application/surface finish/flatness and room ambient.

[S!X]

Either your heatsink's surface was warped beyond usability or something else is in play here- such a dramatic difference is just not possible by simply lapping.


I believe the TIM application was too thick prior to this one, to the point of where it oozed out the sides quite a bit when I put the heatsink on. Why do you recommend not using metal polish on the heatsink? The only thing that I could find that is an apparent disadvantage to using is in this quote:
The disadvantage is that almost all metal polishes contain additives such as corrosion inhibitors and other stuff to shine it up. The problem with this is that those things hinder heat transfer and must be removed with the chemical cleaner.

[lynx]

Most metal polishes work by filling the microscopic pits in the metal surface.

These are the very pits that you need to be filled by the thermal compound, so it can't work properly if you've already filled them with metal polish.

What you really need is some proper lapping compound.
Apply this directly to the glass sheet, and repeat as before.

Ideally you should also do this with the processor, it's no good having a shiny heatsink if the surface of processor cap is like a mountain range.

I think it's been said before, but none of the above is actually lapping.

Theoretically the final process should be to match the processor to the heatsink - this is true lapping. Personally though I reckon that stage is probably a waste of time.

[S!X]

Theoretically the final process should be to match the processor to the heatsink - this is true lapping. Personally though I reckon that stage is probably a waste of time.

I did see someone who posted their results on a forum from CPU lapping and apparently the temp drop was about 10 degrees. I can't seem to find that link in google anymore though :pinch: Even if it pays off that much who the hell can be bothered with that? I'd imagine it'd be quite a delicate and time consuming process because any screw ups could cost you dearly.

[lynx]

I don't see why you would make a mistake doing that, or what sort of mistake.

If I did it, I'd push the pins into a block of expanded polystyrene foam or similar, to protect the pins and give a better grip.

I doubt whether you would see a change of anything like 10C though, unless you were talking about the complete process and you started off with particularly poor examples of both cpu and heatsink.

I think this (http://www.overclockers.com/tips70/) is probably the review you were thinking about. Notice that it is a slot 1 processor. The original cooling on those was pretty abysmal in the first place. Simply dismantling and re-assembling with decent paste would probably have given quite a temperature drop. Notice also that the 10C drop is actually only 8C, nice bit of rounding there.