Lapping: It really works! (my experience)

X

XacTactX

Supreme [H]ardness
Joined
Dec 13, 2010
Messages
4,134
Background Info About the Heatsink

Well, I just had my first experience with lapping a heatsink and I have to tell you all about it :). Sadly I didn't take any before or after pictures of the work but the results speak for themselves.

The heatsink that I lapped is the Zaward Vapor 120. This is a fairly capable heatsink, but I had a feeling that it was being held back by the crude finish on its base. Here is a picture of what the base looks like. The photo does not do justice to the scratches on the base, they are much deeper than they look.

Source: www.overclockersclub.com


I also learned through FrostyTech's review of this heatsink that the base has a surface roughness of 32 microinches, which is considered good by FrostyTech. For comparison, a high quality mirror finish like the one found on the Thermaltake Frio Extreme (review here) has a surface roughness of 8 microinches, which is considered excellent. Clearly, this heatsink had room to improve. This was one of the main factors that led me to lap this heatsink. One of the most important variables in lapping is the initial quality of the heatsink base. A high quality base with a mirror finish will not be improved by lapping.

Lapping Materials

I lapped the heatsink using a mix of sandpaper and lapping compounds. Lapping compounds are supposed to create an exquisite finish, better than what sandpaper can do alone. I read somewhere on the Internet that these compounds are comparable to 20,000 grit sandpaper, an incredibly fine finish. After shopping around and doing some research, I settled on this kit from PCViper because of its low price and the inclusion of three different lapping compounds, something I couldn't find with any other kit. I also read this old review from 3DVelocity and the final picture at the end was very convincing. Personally, I wasn't able to get rid of all the scratches, but the scratches are now very small. If the scratches on the heatsink were the size of coarse hairs before (like armpit or public hairs :p), they are now very fine, like the hair on your head. It may not sound like much, but the difference is dramatic if you see it in person. But don't take my word for it, look at the results!

Test Setup and Results

  • Intel Core i5-750 (3.2 GHz, stock voltage 1.20 V, LLC On)
  • Gigabyte P55M-UD2 motherboard
  • 2 x 2GB DDR3 1600
  • Radeon HD 5450
  • NZXT Lexa S (side panel OFF, all fans at 100%)
  • Stock Zaward "Golf" fan at 2000 RPM
  • 10 minutes of OCCT CPU stress test

Maximum Temperature BEFORE Lapping*** 54 - 55 - 55 - 54 (Celsius)
Maximum Temperature AFTER Lapping***** 51 - 51 - 52 - 50 (Celsius)

All four cores show an improvement of either 3* or 4* C. This is a 6.5% decrease in temperature, a noticeable improvement. I'm happy. :D

Source: www.3dvelocity.com
complete2.jpg
 
Last edited:
I think you meant to say "pubic", not "public" unless you are in a nudist colony, a streaker, stripper, female ferengi, etc. :D
 
Not that I doubt your results, and I know it's a personal review - but - some other things may not be constant: thermal paste application, heatsink pressure, ambient temperature...

Yes, I am in the group that thinks lapping helps out by a couple degrees, but it's hard to measure outside of a very controlled test environment. I lap my waterblocks and CPU's to get a good contact surface. Every little degree helps. Been slacking lately, though. Dust buildup. :)
 
I didn't list all of these specification in the first post (because I thought most people won't care) but:

  • I used Arctic Silver Ceramique 2 with the dot method of application in both of the tests. I tested immediately after applying the paste both times.
  • I also set my thermostat to 77* on a hot day so that the temperature is only fluctuating within about 1* F.
  • Finally, this heatsink uses hand tightening for all of its mounting hardware, and I tried my best to tighten it as much as possible so it would bottom out.

To be honest I can't be sure about the mounting pressure but I'm pretty sure that ambient temperature and thermal paste have been constant throughout my testing. But thanks anyway for bringing up these issues, I know you have good intentions too. :)
 
Yo guys, I have a technical question about lapping and polishing. If you are working on heatsinks that are copper and aluminum, how do you know what increments to use for sandpaper (and polishing compound) so that you get rid of the bigger scratches and get a mirror finish at the end?

I did some Googling and couldn't find any guidelines on this kind of stuff.
 
For both the chip and the heatsink base I use 240 grit to do most of the work and get it flat. Then I'll work up to 800 or 1000 for a final polish. I've had one combo machined on an industrial lapping machine that didn't turn out any better temp results than doing it by hand. I don't bother with any liquid compounds.

I never got any further improvement going past 800 grit which isn't quite mirror finish, but having a mirror is not important and sometimes counterproductive. Getting both surfaces making flat or even contact is what counts. That's why some heatsinks are concave or convex on purpose, it theoretically makes it have better contact with the not-flat stock cpu surface. So for best results the two surfaces need to match each other, whether flat, curved or whatever. Sometimes lapping just one side makes things worse because one side is perfectly flat now but the other is still curved, so you end up with a worse fit than stock.

My SB and 212 didn't improve any from stock which seems to be the consensus on Sandy (I did it before anyone else posted results anywhere, else I would have saved my warranty^), but all the other chip/hs combos I've done have netted 2-5c.
 
XacTactX said:
Yo guys, I have a technical question about lapping and polishing. If you are working on heatsinks that are copper and aluminum, how do you know what increments to use for sandpaper (and polishing compound) so that you get rid of the bigger scratches and get a mirror finish at the end?

I did some Googling and couldn't find any guidelines on this kind of stuff.
Click to expand...

I usually started with 600 grit and then 800 1200 1600 2000. If I recall correctly, lapping both my TR Ultra ex 120 and my Q6600 helps lower the temps around 6°C
 
jthomas said:
For both the chip and the heatsink base I use 240 grit to do most of the work and get it flat. Then I'll work up to 800 or 1000 for a final polish. I've had one combo machined on an industrial lapping machine that didn't turn out any better temp results than doing it by hand. I don't bother with any liquid compounds.

I never got any further improvement going past 800 grit which isn't quite mirror finish, but having a mirror is not important and sometimes counterproductive. Getting both surfaces making flat or even contact is what counts. That's why some heatsinks are concave or convex on purpose, it theoretically makes it have better contact with the not-flat stock cpu surface. So for best results the two surfaces need to match each other, whether flat, curved or whatever. Sometimes lapping just one side makes things worse because one side is perfectly flat now but the other is still curved, so you end up with a worse fit than stock.

My SB and 212 didn't improve any from stock which seems to be the consensus on Sandy (I did it before anyone else posted results anywhere, else I would have saved my warranty^), but all the other chip/hs combos I've done have netted 2-5c.
Click to expand...

Agreed.

If I am going to lap something, it's going to be HS and CPU, the results of doing just one can often be worse than when starting. After lapping my Q9550 and TRUE some time back I saw almost a 10C reduction across all cores. Granted the base on the TRUE at the time was very very poor, I also noticed another large drop in temps by going to a super thin TIM such as liquid pro, as both lapped surfaces had very little space that needed to be filled and the thin TIM allowed for better/more metal to metal contact.

Been thinking about lapping my i5 to see what I get, as I am still running that lapped TRUE, I would be willing to bet I would see a drop, but I really don't see the point as my temps are very good with a mild OC of 4.5GHz.
 
You guys should look into starting of with contact pressure paper to see what areas need to be sanded down to make the best contact.
 
I'm thinking that why don't company already lap their CPU Cooler before selling it as an advantage for us consumers.
 
Illyria89 said:
I'm thinking that why don't company already lap their CPU Cooler before selling it as an advantage for us consumers.
Click to expand...

Many do but I think most of the problem with getting a good flat high pressure contact is because the IHS isn't perfectly flat.
 
I agree lapping does work!!!

My 5 year old TRUE120 lapped does a better job at cooling then the Antec 920 kuhler.

Worst $95 mistake I ever made (but my g/f sure loves the 920's cooling!)
 
Both lapping and tinting improve cooling. Rub TIM into cooler base and CPU and wipe off with lint free cloth or coffee filter (kinda like waxing your car), then install as you normally would. Not sure what all it does... fills imperfections in surface and/or give TIM easier bonding to metal?? Kinda like tinning before soldering I guess??
 
SonDa5 said:
You guys should look into starting of with contact pressure paper to see what areas need to be sanded down to make the best contact.
Click to expand...

I've never seen this for sale but if it's not too expensive it sounds very interesting. :)

I've read about a DIY way to do something very similar to this. When you are lapping a heatsink with sandpaper, take a permanent marker and scribble on the base of the heatsink. Then, as you lap the sink, you can see which parts of the permanent marker wear away first and it will tell you how flat the base is. When all of the permanent marker wears away, you have removed an entire layer of the heatsink and you can continue onto the next grade.

There are a few limitations though if you use this method. You cannot wet the sandpaper because the ink from the marker will dissolve. Also, you should not do this with high grade sandpaper (above 1200 grit or so) because high grit sandpaper might not wear away all of the ink.

Source: UltraKit lapping kit instructions.
 
i thought it was lapping the cpu that mattered, most heatsinks come with a polished and lapped face anyway. its the cpu that has a slight rounded face, which needs lapping.
 
BlueFireIce said:
Agreed.

If I am going to lap something, it's going to be HS and CPU, the results of doing just one can often be worse than when starting. After lapping my Q9550 and TRUE some time back I saw almost a 10C reduction across all cores. Granted the base on the TRUE at the time was very very poor, I also noticed another large drop in temps by going to a super thin TIM such as liquid pro, as both lapped surfaces had very little space that needed to be filled and the thin TIM allowed for better/more metal to metal contact.
Click to expand...

This!

Someone should have told me this before I lapped my heatsink (back when I was still air cooling). It actually made my temperatures worse. I had to lap my CPU just to get temperatures back to where they were before. Apparently one surface had been convex and the other concave so that they actually fit together well. Having one convex and one flat made things worse. So if one surface is flat, then lapping the other surface should help. Or if both are convex or both are concave then lapping might help.

Of course, I ended up just going with water cooling which knocked about 23C off my load temperatures anyway.
 
Its also important to turn the object being lapped 90 degrees every few strokes to average out whatever error you are putting on it with your squishy biological lapping machine.
 
You must log in or register to reply here.
Back
Top