Heatsink Lapping Guide

[Bosco]

The guide can be found here.

[Drobanir]

Do you still use thermal compound after lapping?

[Mendayen]

Thanks for the guide. Definately helps a lot better than some other guides I've seen out there (The ones that suggest figure 8 motion). Only went to 600 grit, but works great.

[maelstrom]

im a machinist by trade and anotherway to get a mirror finish on a a lapped part that is heat malable is to simply apply a heat gun to surface.

 

take part and hold heat gun about 1-2 inches away and wave in the same direction back and forth .this will tend to lower the crests and raise the valleys of any heat affectable part,ie copper,al,silver etc.

 

another helpfull lesson about lapping is that its best done in one direction ,not a back and forth motion.then as stated in guide do 90 degree rotation and repeat.

 

using the above u can achieve a 4-10pu finish fairly quickly

pu is a surface standard for finishs and is millionths of an inch

 

anyway goodluck

[d3bruts1d]

Do you still use thermal compound after lapping?

Yes, I did. Even a lapped HSF will still have defects and groves that will cause air to be between the CPU and HSF... and the CPU itself has grooves in it. A lapped HSF will make better contact, but it won't be perfict.

[DarkStar]

Loved the guide!

 

I have read that also lapping the CPU can help even more, is this really something one would want to do? I mean it is one thing to risk screwing up a $40 heatsink but I am not sure I would want to wet down and sand a $500 P4!

[d3bruts1d]

Loved the guide!

 

I have read that also lapping the CPU can help even more, is this really something one would want to do? I mean it is one thing to risk screwing up a $40 heatsink but I am not sure I would want to wet down and sand a $500 P4!

You know, that same question came to mind. I've seen people lap the CPU, but I've never attempted it myself. I may give it a shot here in a week or two and update the guide.

[psykos!s]

While I've never Cup O' Noodles'd my CPU with nitrogen or dunked a complete PC into an inert liquid, I've had my fair share of amazing overclocks via a variety of methods and have definitely lapped *more* than my fair share of heatsinks big and small...

 

After reading your article and 3 other recent ones covering the same topic, I have to say that either we've all become totally incestuous in the hardware community and just reword other peeps' articles and ideas, or everyone currently writing articles on lapping has completely missed the point.... Case in point;

Quote:

"We lap the heat sink to make it smooth, allowing us to get the best possible contact between the CPU and heat sink. Even after lapping, there will still be grooves for the air to be trapped in, but the grooves will be much smaller, and cause less of a problem." quoted from http://overclockersclub.com/heatsinklappingguide.php

All of the articles I looked at today said pretty much the same thing in different words. It's kinda rubbing me the wrong way, as lapping is far from an infant performance technique and a quick Google for "lapping how to" or "lapping guide" tuned up well over 1000 relevant hits.

 

For the record, you lap a heatsink and/or the mating surface of the IC that it's going to cool not to remove grooves, or even polish the surface. You lap a heatsink or CPU/GPU slug to make it as !!!FLAT!!!! as possible in order to make the heatsink and chip mating surface as flat and parallel to each other as possible. There's really no other reason. In my experience, it's also not a good thing to polish it to a mirror finish, and in reality you could probably leave the surface scuffed with 500grit wet paper.

 

I think the flat and parallel mating surfaces is self explanitory, but some of you might not understand why a mirror finish is bad on the bottom of a heatsink. Other than being a waste of time since noone will see it (it does look pretty cool though!), it's actually inhibits the heatsink from making contact with the chip substrate in almost all instances. The reason lies with the thermal paste that you use. Next time you lap a heatsink and you think you are finished, apply some thermal grease to the bottom of the heatsink and take it to the sliding glass doors to yer back yard if you have them--glass dining of coffee tables work even better. Now, take your heatsink with your usual amount of thermal paste on the bottom and press it to the glass. Now look on the other side. You'll probably see the goo spread in a nice, even pattern. Most likely, the now almost perfectly flat and parallel heatsink will stick to the glass--even a Swiftech MC462 behemouth won't move if you've lapped it correctly.

 

While you admire the laws of physics being expressed in the forms of stiction and hydraulics, look *really* close at the blob of white paste. How much metal do you see? Chances are, you've got relatively THICK layer of goop between the glass and the heatsink--4 or 5 thousandths anyways. No matter what kind of $15 a gram ultra thermal paste you've got there, it does not transfer heat with anything near what you want and need it to do. At that thickness, it's also getting pretty close to being insulative instead of conductive....not good--and most likely your thermal transfer delta is orders of magnitude *WORSE* than the TIM and warped heatsink that you are trying to fix. Thermal compound, when properly applied isn't really *that* much better than not using any at all unless your heatsink surface is extremely un-flat. It would also most likely take several hundred PSI of pressure to displace enough of the compound to bring some metal into contact with some silicon--most spring clips exert maybe 10lbs at best and using the motherboard holes and springs might get you 15lbs of pressure.

 

Here's the deal; You *WANT* the bottom to have scratches. The scratches help "poke" through and displace the heatsink compound. Just like the treads on a car tire, the scratches will help displace fluid out from under the mating surface helping to bring metal into direct contact with silicon.

 

So, I've stated that the goal of lapping wasn't to polish the bottom, that's the last thing you want to do-instead you are trying to make it perfectly flat. Going the extra mile would be to make the silicon you are attaching the heatsink to not only flat as well, but perfectly parallel. I also covered that some texture is a good thing to help displace the over-abundant amout of thermal paste that we all apply.

 

So, now going back to the article on OverclockersClub, we can see that he chose to use a Swiftech heatsink. I happen to have an MCX462 warming up my office with a TB1333@1550 as I type and am extremely familiar with both the Swiftech product line and Gabe Rouchon having had several lengthy discussions about topics such as this. I can confidently say that beyond a shadow of a doubt, every heatsink that is manufactured by Swiftech is as close to perfectly !!!flat!!! that you can realistically make a chunk of metal. It is also no mistake that NO swiftech heatsink has ever left the factory with a mirror-lik finish. It's slightly textured from the (no kidding), optical-grade industrial lapping machines that finish the bottom of the heatsink with a scratch and dig rating that closely resembles telescopic and microscopic lenses prior to final polishing and coating. I hate to say it, but d3bruts1d pretty much ruined the bottom finish on the Swiftech--it's not as flat, it's most likely not parallel and the carefully engineered microtexture finish that's been carefully applied to that particular slab of copper is no more.

 

The last bit of my rant has to do with the thermal paste. The concept of having a fliud fill in the imperfections on the surfaces of both heatsink and silicon isn't new, and it appears to be well understood. However, it seems the *scale* at which this is desirable isn't really all that clear. Say we take the heatsink above and re-lap it to restore some texture to to bottom--I usually stop at 800grit, if not 600grit. What's proven to work the best for me is to apply a dollop of goo as usual, but instead of spreading it evenly with a finger, I literally scrape the surface with a fresh razor blade. This puts the paste *only* where I need it, filling in the grooves and imperfections leaving as much metal as possible to contact the silicon. If done properly, the surface should look milky or cloudy, hardly resembling what we are used to seeing as a thermal paste application and should be done separately to both the chip surface and the heatsink surface before mounting. Remeber, you bought and took the care to lap things FLAT (not smooth or shiny grrrrr!), you want the METAL of the heatsink to conduct the heat NOT the paste. I had a hard time with this the first few times I used this method as there just doesn't seem like there's enough there--if you don't trust it, closely monitor the sink and die temps if you can during setup or a defrag (or FoldingAtHome or 3dMark loop, whatever) to get that heatsink working and you'll be surprised at the appreciable difference in not only lower max temps but a substantially improved return to idle temps after load.

 

Hopefully this helps clear a few things up about lapping for some of you. I gotta throw in a last jab here though; shame on you guys for not taking advantage of the extremely long, detailed and generally useful information-filled tens of thousands of hardware tweaking and overclocking related articles that are all over the web. When I starting doing this way back when, there's wasn't much of a web let alone a fairly large a supportive community of PC hotrodding information like there has been in the last 5 years or so--and you're silly not to at least do a cursory search before anything else when you have just about any kind of hardware problem or questions about something.

 

Shame on the author for not informing himself more thoroughly on the subject before posting an article and effectively spreading disinformation about lapping--I *know* you didn't mean to misinform your readers, but it's still irresponsible. >whew< I feel better--lets all get to cooling!

[d3bruts1d]

When it rains, it pours, doesn't it?

 

Let's see if I can comment on this without having to write a book... First, a flat surface will transfer heat better than a non-flat surface. Metal to metal is better than metal to air to metal, which is what you would have in a non-lapped (or poorly) heat sink.

 

Mirrors are flat surfaces, which is why I went for a mirrored finish. Remember.. metal-metal is better than metal-air-metal. While you don't have to achieve a mirror finish to see improved results, I decided to do that, it's my guide, I can do that, right? I suppose I could have included a line that stated

[posey]

Well put d3bruts1d

[psykos!s]

wasn't a knock -- although i cam off harsh. It's just a peeve.

 

Thank GOD you said FLAT. You CAN have a MIRROR finish without it being flat. What did you use to measure the FLATNESS of the heatsink before and after? I have a VERY hard time beleiving that you improved the performance of the Swiftech at all, and most likely performance is worse--although if you POLISHED it correctly, you would most likely not have the equipment to measure the difference.

 

My point is, POLISHING does not LAP something FLAT....the way your article reads, you state that the aim of lapping is to POLISH out all the grooves and ridges and what have you -- you do NOT emphisize FLAT, you emphasize SHINY. Take any crappy extruded heatsink and POLISH it with your dremmel or buffing wheel -- don't lap it, just polish it. By nature, extrusions are NOT FLAT. After the bottom is all purdy and mirror-like, compare it to the lapped and polished swifty and take a few pics. They look alot alike, both are shiny and the surface imperfections are very hard to see, but one IS NOT FLAT. Flat is good. SHINY is a waste and misleading.

 

I don't know how long you've been in the hardware community, but a while ago, Alpha started shipping some of their heatsinks with a polished bottom surface. Alot of web reviewers oohed and ahhed, and it took about 4months before anyone really noticed and published info on how un-flat their stuff was at the time. They covered up bad quality control with making it purdy. This irks me.

 

Lapping = making something FLAT. POLISHED to a mirror surface will produce WORSE cooling if ANY paste is used. Polishing is a useless step in lapping. I've done alot of it dude. Alot. I *STILL* to this day have a hard time applying paste correctly too--I want to always use too much. It's not as big a deal as it was when CPU dies had heatspreaders on the top--like the old P]['s and K6]['s and such--flipchips have a very small die footprint, but still, attention to detail goes a LONG way. It becomes absolutely NECESSARY when you start supercooling to *properly* lap both the heatsink and chip slug with a bare minimum of paste.

 

Flat, NOT shiny bro.

[d3bruts1d]

I don