Lapping a Heatsink

[bront]

I'd like some recomendations for lapping a heat sink. What paper(s) to use, any tips, techniques, ect.

I'm sure I'm not the only one here who's never lapped one before, so give as much detail as possable

 

[chuckshissle]

Get a 2000grit sandpaper and a polisher. Place the sandpaper on a flate surface and tape it down securely. Then evenly start rubbing/lapping the hsf on the sandpaper as level as possible. You can use water or honing oil for better result. Once finished to the desired result then use a soft cloth and metal polisher compound to polish the hsf for a mirror like surface.

 

[wun911]

Strangely enough I use a strop and oils.

A Strop is the stiff piece of leather that is used to sharpen cutthroat razors. Strops can be found in many antique shops.

 

[Datman]

A flat piece of glass and some wet & dry sand paper in a few different grades.
That's what I've used for things in the past.
Change the direction that you are rubbing in from time to time.

 

[OEM-NIB-OBO]

search easypc heatsink lapping kit on google and then buy on eof the premium kits with the glass. It'll work optimally.

 

[306maxi]

search easypc heatsink lapping kit on google and then buy on eof the premium kits with the glass. It'll work optimally.

I've always wondered why you'd lap a heatsink when all you do is eliminate spots where AS5 can go into. The thing is that AS5 is generally more conductive than copper or aluminium. Obviously if the bottom of the heatsink is so rough that it doesn't sit on the processor at all then it's best to lap the heatsink but I wouldn't bother otherwise.

 

[botchog]

If I plan to lap my zalman 9500 what griit sandpaper should I start with?

 

[Crashman]

Zalman 9500 is probably already smooth and flat, so finishing it with 1200 should be good.

Most other coolers...you can cut them flat with 600 and jump straight to 1200 if you would rather spend a little more time than money.

 

[chuckshissle]

Get a 2000grit sandpaper and a polisher.

You'll be wasting your time starting with 2000. *shakes head*

Well, my mistake but it depends on the surface smoothness/roughness of the heatsink contact surface.

As for my heatsink was smooth out already but not mirror like surface, so I just used the 2000grit sandpaper and polisher.

 

[306maxi]

That's because the silver particles in AS5 aren't large enough to fit into the machining groove on a typical HSF.
According to calculation the roughness created by 1200grit sandpaper fits the particle in AS5 perfectly.
http://www.overclockers.com/articles1303/

It's the same reason why people get better performance with Artic Silver Ceramique or MX-1 when the metal surfaces is treated at 2000 grit due to smaller particle size than AS5.

The thing is that AS5 is generally more conductive than copper or aluminium.

Wrong. Metallic to metallic contact is the most thermal conductive method.
Overall thermal conduction of AS5 is same or less than actual metal and is only useful in filling in the micro pits.

AS5's conductivity is higher than that of Aluminium and is only just behind copper in conductivity There will always be microscopic pits on both surfaces and this means that you're always going to need to have "goop" in the middle to provide a connection and as you say.

My unlapped heatsink with AS5 that I applied with my bare finger is keeping my CPU nice and cool at 25 degrees celsius (reported by mobo which of course could be wrong) when the temperature inside is probably about 20 degrees.

With the labour involved with lapping my HS and a going to the trouble of reapplying AS5 by other means you would want to hope that it cools your CPU to a temperature that is lower than ambient.

*sniggers*

Rather than spending a couple of hours lapping your heatsink just put the damn thing in your fricking computer. The increase in productivity from the fact that your PC will actually be turned on will outweigh that extra 3.338758mhz that you get out of it with your "increased cooling"

I could understand someone doing this if the bottom of their heatsink was rough as anything but most decent coolers are fine. and if you bought a low quality cooler then you probably wouldn't care about overclocking anyway!

Just my rant-tastic 2 cents

 

[clue69less]

Rather than spending a couple of hours lapping your heatsink just put the damn thing in your fricking computer. The increase in productivity from the fact that your PC will actually be turned on will outweigh that extra 3.338758mhz that you get out of it with your "increased cooling"

I could understand someone doing this if the bottom of their heatsink was rough as anything but most decent coolers are fine. and if you bought a low quality cooler then you probably wouldn't care about overclocking anyway!

Well, you have a decent perspective about the time/benefit angle. I know that lots of people lap their heatsinks - at least they call it lapping when most of them actually just grind them. Lapping traditionally involved using a "lap" substrate along with a polishing compound and a lubricant as opposed to sandpaper, which is considered grinding. Anyway, having taught metallography to numerous people over a period of about a decade, I can tell you with a high degree of certainty that very few people can grind and polish something the size of a heat sink and keep it anywhere near flat. My comment is based upon teaching the techniques to students, then actually measuring the flatness of their ground and polished surfaces. Even if you use an automated grinding/polishing device that costs upwards of $100K, flatness is not guaranteed, but you stand a far better chance with a good machine than you do by hand. It's almost an art form to grind and polish by hand.

I have used one device that works very well for soft metals like copper when flatness is a priority. It uses a metal bed that has a powerful sonicator below it. The metal bed has some sort of cloth stretched across it that supports the grinding/polishing media, which can be anything from silicon carbide to diamond or even colloidal silica. You add a lubricant that can be water, kerosine, hexane, etc., depending on the application. The surface to be ground/polished is mounted in a jig then set onto the bed and it rides around on the vibrating surface like a bumper car. To do it up right, you'd want to polish your CPU the same way, to the same finish and I've never had the guts to do it just yet, but I'm thinking about it. I'd think that if you went down to 0.1 micron diamond or colloidal silica at 0.05 micron, you'd need an even finer thermal grease than Arctic Ceramique.

 

[michaelahess]

I use my 5 and 1 micron fiber polishing paper, if it's good enough to terminate single mode fiber, it's good enough for a cpu cooler

 

[clue69less]

I use my 5 and 1 micron fiber polishing paper, if it's good enough to terminate single mode fiber, it's good enough for a cpu cooler

The medium is fine, no doubt about it. A true one micron finish on CU looks like a mirror. It's the proper application that's tricky, especially by hand. The thing hand lappers (sounds porn, eh?) have going for them with CPU HS's is that the most important area for flatness is the center and the center is typically the one of the last areas to go out of flat if you screw up.

 

[michaelahess]

I use a small piece of very flat metal and wrap the paper around it, it's pretty flat, better than sandpaper and a block of wood And since it takes off such a small amount of material, making the surface curved would be pretty hard.

 

[samir_nayanajaad]

I looked into lapping a little while ago and determaind that the most effective way to get a lower temp is to remove the heat spreader from the cpu then slap your hsf right onto the cpu with some as5 inbetween.

this is how to do it for an amd a64 cpu. anyway dont think any others would be much different
http://forums.extremeoverclocking.com/showthread.php?t=178897

more polishing of the hsf surface could help but removing that ihs will defiatly help if you like to risk you cpu for it

 

[michaelahess]

Ah!! I'm not that gutsy!

 

[clue69less]

I use a small piece of very flat metal and wrap the paper around it, it's pretty flat, better than sandpaper and a block of wood And since it takes off such a small amount of material, making the surface curved would be pretty hard.

Think so? I recommend a little trip into the literature. It's easy to curve a surface by sanding, lapping, whatever. With HS's, it probably isn't as easy to notice because the surface you mate with isn't totally rigid and the HS is most likey to crown if you get it out of flat - so it just deforms the IHS when you draw it down.

 

[OEM-NIB-OBO]

Look I know a lot of people are giving you a lot of recomendations. But unless you are a serious Overclocking diehard then the difference you'll get by using any sanding paper above 1200 grit is very minimal. At best you'll probably see only about 1C - 2C of temp difference on your Cpu. So unless you have ABSOLUTELY nothing else you can upgrade that has too do with you cooling then dont worry too much about it.

So all you have to do is get a Lapping Kit from "EasyPckits.com" and you'll be in business. Oh and to anyone who says that you'll benefit from using AS ceramique for 2000 grit and Arctic 5 for 1200grit should really look at the extremely small difference in temp that it will give you. I mean "C'mon"! , just sand it to 2000grit put on "ANY Arctic Silver product you wish, and you will have guaranteed low temps!

 

[OEM-NIB-OBO]

Oh and if you get a kit from Easy Pc, just make sure you get one that comes with the glass, so that you can warap the sandpaper around it. That way when you sand the heatsink it will be perfectly smooth.

 

[Datman]

How do I properly lap a heatsink/ waterblock?

Another guide just to read.

 

[samir_nayanajaad]

I agree with what OEM-NIB-OBO had to say, unless you are seriously into ocing this is just not worth it. If you’re into ocing that much just take off the IHS like I suggested above because it is way more effective and though this may sound dumb I would be inclined to believe it is less likely to mess something up.

I haven’t done this but from the directions I posted above it doesn’t seem nearly as complicated, takes less time, and works better than lapping, but that’s just my ballpark guess no experience to back it up.

If you want to do this just for the experience though go for it, there's a first for everything. One more thing that you might want to think about is actually lapping the IHS to the HSF using lapping compound. You can buy it at Auto Zone or any automotive parts store, I even think wal mart might have it (they own auto zone don’t see why not).

Lapping compound is a lot like as5 because all it is this grease with metal in it made to be like liquid sand paper. All you need is the two parts to lap together. Put some on the parts then gently rub them together in O / | - 8 patterns and they are done. The good thing about that is the two parts match each other perfectly if you do it right. Might want to practice a little on something that doesn't matter first as with the sanding too

 

[OEM-NIB-OBO]

Yes you are very right, thankyou

 

[wun911]

I must agree that laping by hand would not give a even surface...

I like the idea of using sonicaction, we use a set up similar to "Clue69Less" for giving the edge to surgical tools and microtomes. Microtomes used to slice thin tissue sections including bones for microscopy etc.

The sonicator is just like a water bath, however the liquid (and the heat sink) is moved rapidly in a short motion by high frequency soundwaves (ie 'Sonic -action'). We usually use some sort of low molecular weight highly crosslinked sephadex gel as an abarisive media.

However, the people at Zalman do a pretty godd job when they pump out those heat sinks, laping one of these will give you only a very marginal increase in heat transfer. I laped my HSF only because a scratched it by accident.

 

[wun911]

Why not put the Zalman 7700 HSF under a microscope and look to see if its a flush surface or not? A 0.1nm disparity will come out looking like a Mt Everist.

Basically we can use the microscope to magnify a surface by about a 400 times. Microscopes do go up to about x1000 but then we need immersion oil and light need light to pass through the HSF which is physically impossible.

If we cant figure it out using a microscope we will move to SEM (scanning electron microscopy) a magnification of about a million times.

eg people have used it and have noticed that the human hair is not smoth surface it is actually rough etc etc.

If its not flush from the Zalman factory we can lap it and see if there is any difference or if there is room for improvement!

Because im sure that some of you out there using sand paper and harsh abarisive materials will make the surface worse!

And yes we do have a microscopes at our university......

 

[michaelahess]

I can get a fiber perfectly flat to the eye through a 800x optical fiber scope, course you don't want fibers flat, but I did one just to see.