Adventures in lapping

L31mr0d · 03-12-2006 12:03am #1

Ok, so since reading about the ins and outs of lapping a few weeks ago, I decided with my next cooler upgrade i'd give it a go. I've been through a few iterations since and think I have the best method for doing it down.

First off here is my first attempt, I didn't want to chance messing up my new Tuniq, so I practiced on a spare stock AMD HS.

First off you will need a flat surface (i.e. marble or glass) I had a few spare photo frames around so I decided to use those, and some wetordry sandpaper, in the picture below I have a selection of grits that can usually be bought at any local motor factors (i.e. grits 220, 400, 600, 1200, 1500), any higher than this and I found I had to order abroad.

Next get your motherboard and HS

remove it

Do the penny test for composure

Now, i've since reworked this setup, but i'll get to that later. But for this I duct taped the piece of sandpaper to the glass in the frame before starting

I then proceeded to go up to the grits moving the heatsink up and down on the sandpaper, rotating it by 90' every 2-3minutes. This particular HS was VERY convex, I wish i'd taken a photo, but it took a good 20 minutes of lapping to get rid of the bump in the middle of it. The picture below is the result of this method of sanding up to a grit of 1500... not that good I know

But at least its flat)

Next I ordered some lapping kits from Dave at easyPCkits.com, for the AMD HS I used the heat sink lapping kit as i'd already gone up to 1500, I only used the grits 2000 and 2500 that where included. I also reworked the setup by removing the glass from the frame and duct taping the sandpaper directly to it. I also found that putting a few drops of water on the glass helps the sandpaper stick in place.

Here are the results

Better but not great, the one thing I found is that on the easypckits guide it tells you to wet the paper thoroughly before using, when I did this I found the grit would clump up on the sandpaper and leave deep scratchs in the heatsink, that would require going back to a lower grit to remove. So imo, its better to leave the sandpaper dry whilst sanding as the metal dust can easily be wiped/blown away if it is dry. Also I started sanding the heatsink by rotating it in a counterclockwise manner on the higher grits which I found to be effective.

Next i'm going to show you the next method I used for sanding the Tuniq Tower 120 with better results. Then i'm going to be lapping my Opty 165 which i'll log also.

Any opinions ideas welcome.

hanky · 03-12-2006 12:22am

An engineer told me that Mirrors are much flatter than standard glass so would probably be much better than glass for honing

mathias · 03-12-2006 1:48am

Excuse my lack of knowledge but ,
Can anyone tell me why on earth you would want to do this and the benefits?
And some links to the theory of why its a good idea please ?

mathias · 03-12-2006 2:05am

OK , just looked this up , It dumbfounds me to think that there are people out there who will go to all that trouble for a system temp thats 1- 5 degrees lower , what utter lunacy.

I notice no one shows any performance improvements in terms of flops , FPS , 3d mark scores , just some temp graphs that look very similar before and after , and no real world benefits at all !!

You'd get the same or better results using a higher grade heatsink paste , like arctic silver or something , thermal conductivity in no way requires that any surface be flat , just that they have a good thermal connection , thats what the paste is for !!

Jeez !!

Flaccus · 03-12-2006 2:26am

L31mr0d,
thanks for the info. Something I have not tried myself yet but may do in the future. Interesting stuff about using dry sandpaper because I always read you need to wet it.

mathias - while the gains to some are not perceivable, some heatsinks and even heatspreaders can be quite concave and lapping fixes this. It can also mean less thermal paste since the whole purpose of paste is to fill in the microscopic hills and valleys on the heatsink so the cpu can make good contact. If you had a perfectly flat heatsink and cpu heatspreader and I mean perfect, you would not even need thermal paste. Also there are those that are into overclocking as a hobby so every degree counts. I think L31mr0d went to alot of effort to show those that want to lap, how to do it properly. I also think your comments were unhelpful and obviously based on a few reviews you just read.

mathias · 03-12-2006 5:35am

while the gains to some are not perceivable

Perception in the case of engineering is a measurable result of some consequence , 1 to 5 degrees celsius is of no consequence in terms of a processor !!

I have a big interest in overclocking and modding , Im also a mech/Elec engineer , when I first saw this I thought , no way could this be just to improve thermal conductivity , but lo and behold it was , working in heat systems for years tells me this is way too much effort for very little return ,

Now changing from stock cooler to say , watercooling or a sub zero refrigeration system , they are worthwhile projects giving maybe 20 - 40 degrees improvement , even changing from a stock cooler to something like a 120 mm Zalman gives 10 degrees or so improvement , and thats worth it !

This is too much for too little if you ask me ! There are way better projects to improve cooling.

L31mr0d · 03-12-2006 1:32pm

mathias wrote:

Im also a mech/Elec engineer

Thanks for that, although I don't see its relevance, i'm also an engineer by profession, but I prefer to back up my opinions with facts rather than titles

mathias wrote:

even changing from a stock cooler to something like a 120 mm Zalman gives 10 degrees or so improvement , and thats worth it !

Can I refer you to the second paragraph of my first post where I said I was using this HS to "practice". The temperature benefits of lapping a stock AMD HS are negligable, but I wanted to get the process down before experimenting on some more expensive air coolers, an example of which you will find below. Also, I overclock, so temperatures DO matter, my opty running at 56'c under load compared to 61'c makes a big difference. Those extra 5'c could mean an extra 100Mhz of clock speed (i.e. giving you your flops, fps and 3dmark06 scores)

Lapping a Tuniq Tower:

Anybody who knows anything will know that atm (I use that term liberally

) the Tuniq Tower 120 is the best non TEC aircooler out there. But compared to the likes of your Scythes and Zalmans the finish on the base is horrible. Almost as bad as a stock AMD HS. In fact I found the copper on the sides and top of the copper plate to be shinier than the bottom

Heres the euro test (not good

)

First I removed the fan unit, which, to my surprise is not held in by anything, you just push the fan up through the HS to remove it.

This is the setup, I used: Glass, duct tape, premium lapping kit, Articlean, some Qtips and lint free cloth (lens cleaner)

The method I used for lapping this time I had perfected on the previous AMD HS. I do 2 minutes going up and down, then I wipe off any metal dust from the paper with a lint free cloth, then rotate the HS by 90 degrees and repeat for a 360 degre rotation, I then do a number of minutes moving the HS in circles, checking the surface and corners to make sure everywhere is being lapped evenly.

The way I held the heatsink is in the pictures below. I found the weight alone of the heatsink was sufficient for sanding so I applied no extra pressure down on it. I also found applying pressure would cause the edges to snag the paper, you can solve this also by beveling the edges by stroking along them a few times on the sandpaper.

here are the results, I used the premium kit off easypckits which included girts of (400, 600, 800, 1000, 40 micron, 25, micron, 20 micron, 15 micron, 10 micron) This kit IS a lot superior to the original sandpaper I was using and I found I started to get a good mirror finish after using only the 400 grit paper.

This is at 600 Grit

1000 Grit

10 Microns

I'm very happy with this finish, I'll be lapping my Opty today so I should have the pics up by this evening. Theres something sexy about having bare copper on copper between the HS and CPU

, although there'll be some coollabs liquid pro to buffer them

Cuddlesworth · 03-12-2006 3:57pm

Somebody give this man a Mod status.

Cheers for that, its going to come in handy when I go to lap my G5 and E6600 in a weeks time. I still need to figure out how to get to hi-tide to pick up some cool labs liquid pro.

Oh and mathias, a pro lapping finish like that, a good mount with good past(liquid pro/silver 5), and high end wc/air gear can see temps drop by only a few degrees according to temp probes. Actual temp's will be even lower, as most probes are not at the source of the heat and better thermal efficiency across the entire Ihs will cause the temp probe to register a higher temp then it should.

This could mean anything from 100mhz to 200mhz oc, which is the equivalent to spending 200 euro on a better proc.

Anti · 03-12-2006 4:38pm

Very impressive. This is what poggy did too. He also ordered sandpaper from abroad, As he couldnt get 2000 or 2500grit paper in ireland. And if i remember he got 7C out of it.

mathias, You say your into overlcocking and modding. Then you really should know 5c makes hell of alot of a difference when it comes ot overlcocking. My old amd 3500+ would run 2.93ghz at 62c but if it hit 65c it became unstable. And thats only 3c of a difference.

I too will be lapping shortly, And not just the cpu hsf. The graphics card too. be it vga waterblock, or a new cooler.

Creamy Goodness · 03-12-2006 4:42pm

holy ****, that's a great job. fair play to you

Psycopat · 03-12-2006 5:26pm

excellent guide, nice detail and simplistic, 10/10, i may even try it on my arctic cooling freezer pro when i go to water cooling after christmass

TonyM. · 03-12-2006 6:04pm

why not use valve grinding paste i believe it comes in different grades
and it would be easier to use.

L31mr0d · 03-12-2006 6:42pm

TonyM. wrote:

why not use valve grinding paste i believe it comes in different grades
and it would be easier to use.

Easier than pushing a heatsink around on a piece of sandpaper? This process serves 2 purposes, 1: It flattens the surface, 2: It removes any grooves on the surface. Have you got a link to this paste, i'd like to see how its actually used. I ordered this finishing paste as well from easypckits and tried it with the AMD HS. I didn't mention it because the effort needed to get it to work wasn't worth it. Its really meant for application by some rotary tool with a felt head, i'll post results of it once I get this felt bit for my dremel.

All in all the Tuniq took me approx. 40mins to complete. A lot of the time was spent just stopping and checking that it was being sanded evenly, i'd advise to have a lamp nearby to shine directly onto the surface to pick up any small scratchs which will let you know it is catching on some dirt on the sandpaper.

All in all, a chimp could be trained to do this, and the rewards are knowing that you are getting the best possible temps out of your HS. Its one less thing to be thinking about when OC'in. The kits are dirt cheap with shipping from the states, i'd recommend the premium kit as the sandpaper is reusable and the best quality out there. The other kits are good also, but won't give you the same finish as the micron stuff will.

TonyM. · 03-12-2006 6:57pm

http://www.flatlap.co.uk/consumables.asp#paste

You can probably get Holts Grinding paste in any Motor Factors its used to seat new valves when rebuilding car engines.

http://www.edirectory.co.uk/pf/880/mia/d/holts+valve+grinding+paste/pid/439242

mathias · 03-12-2006 9:29pm

Hello L31mr0d ,

Nice work by the way , Im not disputing that ,I just dont see where the math comes in with regard to improvements , I'd be grateful if you could post results before and after ,

Thanks for that, although I don't see its relevance, i'm also an engineer by profession, but I prefer to back up my opinions with facts rather than titles

Here is its relevance , see attached word document and picture , this is one of my work projects thats about 2 years old and naturally some names have been removed , but you get the idea ,

I work in Semiconductor final module test , thats testing the finished chip , the finished product must be tested at its most extreme expected operating conditions , this can be up to 150 degrees C die temp. This attachment shows an evaluation plan for a cooling system from a company called Kryotech.

In particular note the section on thermal resistance ... Appendix A ( by the way the systems are for testing large processors , typically big iron server processors , ATC stands for active thermal control and is a system to ramp up a processor to its maximum heat dissipation for test purposes before being shipped to a customer , I had a lovely powerpoint presentation but cant upload it here, its too big, the word document covers all the parameters for accurate thermal modelling. )

What I would like to understand is what section of thermal resistance ( or any of the other parameters ) does lapping improve , that cant be done with paste ?

Heres a drawing of the test site hardware , not dissimilar to some of your photos above ,

L31mr0d · 04-12-2006 12:39am

Here ya go. You might need to read a bit of german but you get the drift. This graph shows the different thermal conductivities of different pastes, materials and metals that we are concerned with. Kupfer is copper (go figure

) I'll let the graph speak for itself, basically I am aiming for little or no TIM and more copper to copper contact. The benefit? Well as you can see copper is 42x more thermally conductive than AS5. I'll also be using a coating of liquid pro which as you can see has the highest thermal conductivity of all the TIMs. I hope this clears things up.

Cuddlesworth · 04-12-2006 1:08am

Btw, Silver rates higher for thermal conductivity then copper. Just cost's a **** load more.

Wasn't there something about copper to aluminum heat sinks, Where the thermal efficacy of copper was great at removing heat off of the IHS but the aluminum was better at transferring the heat to the air. Hence copper heat pipes with aluminum fins.

mathias · 04-12-2006 3:56am

Thanks for the Graph L31mr0d and it looks an impressive reason at first glance but then you need to look at the formulae , not forgetting that those values are conductance and need to be inverted ,

http://en.wikipedia.org/wiki/Thermal_conductivity

http://en.wikipedia.org/wiki/Thermal_resistance_in_electronics

Plugging in the values into the resistance equation linked , which is good enough for this purpose , you'll see that the difference is going to be very small , even between the worst and best on the graph.

The system of chip casing / TIM/ Copper / Fins is not throttled by any of the materials conductance values as it does not produce enough heat to do this , the biggest heat transfer comes into effect from fins to air !!

Add to that the fact that chips are not made flat , ( there are no strict requirements on the vertical dimensions of processors , only on the X a Y dimensions along with the ball grid. ) and you have a situation where you have a nice flat mirrored surface on your heatsink , and the surface of the chip is nothing of the kind !!.

Unless of course the chip is lapped as well !

http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/32800.pdf

Even so , the recommendations from AMD still recommend a TIM of some sort , basically the calculations dont support lapping , the units will do as they stand and with the correct TIM there is absolutely no need to lap !

This to me is one of those half correct ideas like keeping left and right speaker wires the same length , etc . It looks right , but in reality makes little or no difference ! The math backs that up.

feylya · 04-12-2006 4:00am

mathias wrote:

Unless of course the chip is lapped as well !

Check the last line in post seven. I'll give you a hint - "opty" is short for opteron, an AMD cpu.

L31mr0d · 04-12-2006 11:38am

mathias wrote:

Even so , the recommendations from AMD still recommend a TIM of some sort , basically the calculations dont support lapping , the units will do as they stand and with the correct TIM there is absolutely no need to lap !

Then you clearly aren't out for the pursuit of performance. I'm am fully aware that lapping my heatsink and CPU might only shave off 5'c, and that this might only give me enough headroom for an extra, say, 100Mhz clock, which in real world terms in not noticeable except in synthethic tests. But why not? In the time it took you to write that last post I could of been half way through lapping another heatsink.

The magpie in me finds it enjoyable to make something dull and concave, flat and shiny. I want the best possible performance out of my heatsink and cpu, and if that means an hour or so of doing something I enjoy then so be it. I really don't want to turn this thread into a philisophical debate, but life is full of pursuits that in the big picture are meaningless to the whole, but important to the individual.

Something that you may deem to be irrelavent is viewed by myself and many of my peers to be a paradigm shift towards aircooling. Its all down to perspective, 5'c to you is a drop in the pond, to me its a tidal wave of difference. Now where did I put my dremel?

Anti · 04-12-2006 9:00pm

Agreed, Just a few degrees can be the difference between a stable cpu overclock and a unstable one. As i pointed out earlier in my previous pc.

I myself today ordered two lots of the kit. One for praccy on a coolermaster vortex hsf. And another for my tuniq. And the cpu.

Only difference is , i will be using some mercuary. A friend of mine is a lab tecnician in the local uni. So he says he can rob me some on the QT loike

If there is enough left over. I might try lapping the 8800gts stock hsf too. This b.astard runs hot !

TonyM. · 04-12-2006 10:33pm

http://www.3dvelocity.com/reviews/pcviper/ultrakit_2.htm

Some more info on using paste.

gline · 05-12-2006 8:28am

would you not just go phase :rolleyes:
hehe

Cuddlesworth · 05-12-2006 4:12pm

gline wrote:

would you not just go phase :rolleyes:
hehe

Speaking of which, was think of treating myself to a Christmass present.

Any good? I dont know if I should buy it or not, since I have a good oc with water with better gear on the way.

gline · 05-12-2006 5:09pm

krazy_8s wrote:

Speaking of which, was think of treating myself to a Christmass present.

Any good? I dont know if I should buy it or not, since I have a good oc with water with better gear on the way.

id go for it man, sweet system

Psycopat · 05-12-2006 6:24pm

krazy_8s wrote:

Speaking of which, was think of treating myself to a Christmass present.

Any good? I dont know if I should buy it or not, since I have a good oc with water with better gear on the way.

that looks quite nice, anything similar in shops?

Cuddlesworth · 05-12-2006 8:34pm

Nope, I have looked into this for quite some time, its costs around 400-600 Euros for a custom phase change unit. These out perform by miles units like the Mach2 and Vapochill LS which will retail for 790+. Only thing close performance wise is one that extreme phromethia sell for 1200 sterling. This unit is not the only one you can get, there are lots of builders in the UK and States that make a living out of making these units for people, and are willing to do some amazing fits inside of case's. What you lose is support without cost, as most builders will warn you. Its rare you see units breaking down if they are well made.

If you want more info feel free to post up in the xtremesystems phase forum about this. Most builders including this one ply their trade there.

Anti · 05-12-2006 8:49pm

Just read that guys ebay sale, dual sli chillers ..........WTF ? This is overclocking gone to far.

Psycopat · 06-12-2006 1:06am

That is ridiculous, i want one

just for the crack like. You'd want a board with huge voltages though. is it just for the cpu?

L31mr0d · 06-12-2006 5:17pm

First let me tell you that i'm glad others are going to learn from my mistakes in this thread and start off with a better understanding oh how to lap either your HS, or IHS... I am NOT however glad that my Opty is now lacking 1 pin :eek: but it still works, after replacing it back in my mobo with sweaty palms and booting up all seems fine, I guess I lucked out and accidentally broke off a redundant gnd pin or something. It happened because I had put the black square of foam that comes with AMD64 CPUs on top of the pins and was sanding when one of the edges snagged the sandpaper and I pressed down a little too hard on one side, thankfully only one pin broke. I then reevaluated how I was going to continue. I placed the plastic covering over the CPU and foam and sanded it very gently (picture below)

Heres how you can avoid this:
1. Bevel the edges of your IHS BEFORE beginning to sand (do this by stroking along the edges on a low grit sandpaper)
2. Protect the pins with by using the package that came with your AMD CPU
3. Press down VERY lightly on the CPU, the movement alone should be enough to sand it. If you find it isn't sanding fast enough, move down to a lower grit sandpaper.

L31mr0d · 06-12-2006 5:35pm

First off you will need all the same supplies as for the HS guide, but you will also need some kind of protection for the CPUs pins as outlined in my last post, also i'm adding to the list "Brasso", I got recommended this on another forum and I have to say it works very good for removing very fine scratches after you have finished sanding.

First heres the euro test with the stock nickel plating. You'll want to sand this off until you see only copper, I used about 1 A4 sheet of 230 grit to achieve this.

After a while you'll accept that your warrenty is gone for ever

and you'll start to see all that coppery goodness. From this picture you can see how horrible concave this IHS was, as the copper around the edges has already begun to show and the nickel in the middle is still barely touched.

Almost gone...

voilá, naked copper

Here it is after 400 grit

after 800 grit

20 microns

and the finished article... after 15 grit and a rub down with some brasso.

Here's a few more glory shots... was it worth it, hell yes

. It might of almost been a very expensive copper pocket mirror for the missus but for now its a perfectly flat,shiny thermally efficient Opty 165

Adventures in lapping

Comments