Last message on previous page:
cool, when you were lapping the CPU, did the surface get hot?

when i lapped my heatsink  i just kept the sandpaper on a flat surface and
and went crazy  moving it up and down the paper really fast, going to higher grits to get a  mirror finish  but the surface was also really hot when i was done  (which is why i  didnt do it to my cpu)

cleaned up your image a bit

 
No it didn't get hot.  I'd discourage anyone from lapping to the point of the surfaces getting hot... not because the friction generated heat is bad, but because you probably aren't getting a nice flat lap by moving that fast.  The key is slow, precise movements.  Also, you can lap your CPU with no fear of the friction you're generating hurting it; after all, how hot does your chip get from CPU load?  My Q6600 can hit 65 °C depending on the settings I'm using which is probably much hotter than it gets when I lapped it!

Is there such a thing as lapping the CPU or HSF too much?

 
 
yep, if you start seeing come of the chip through a hole made in the  heat spreader then you know you sanded it too much
 
when lapping, your not really looking to make it super thin, your just looking to make it as flat and smooth as possible.
 
 
(i wonder why  don't they just give the heatsink and cpu a smooth finish to begin with)
 
PS when it comes with some AMD stock coolers, lapping it is vital for overclocking  
 
here is why you should  give your amd heatsink a smooth finish
 
(large image, click for full view )
 

 
here is a amd stock heatsink  when placed on a scanner

holy crap.  I was thinking of giving this a shot on my next build, but after seeing that pic, maybe I should lap my current 3700+ with my stock HSF.  If I try, I guess the best method is to sand a little, look at it, sand a little more, look at it..., making sure you don't do it too much.  Very interesting topic.

Also, I've read some of the other guides on how to go about lapping.  Some talk about doing this vertically rotating 180 or 90 degrees every few minutes to make sure you sand evenly.  My best guess is that the figure-8 method does the same task without all the shifting and turning....right?

Maybe I'm not all that 'knowledgeable' on lapping but...
 
Does everyone 'lap' off their nickel plating?  You know why that plating is there, right?  Because copper corrodes in the presence of Oxygen.. even the airborne oxygen.....  You might say 'it won't be exposed to Oxygen". but my comment would be 'yes it will... oxygen will diffuse into the thermal paste'.  Any around that has lapped their CPU a year+ ago and can comment on what their CPU looks like now?
 
The only thing stopping me from lapping my CPU is the corrosion problem.

 
I dought after two years theres going to be soo much corrosion that it will affect the cooling, show me proof that after two years it will affect cooling. Aren't we nit picking here a little?

not all copper has this problem, and if it doesn't it may take longer than 2 years, not sure if this is true for the CPU but My mom has some pots and pans that are copper and it has been years and they still look almost new
 
 
probably for the CPU, the damage to the copper will probably take like 5-6 years and by then you will most likely  have upgraded  
 
generally lapping surface of the heat spreader on the CPU is smooth but the shape can be a little distorted which can cause higher temperatures
 
(I will scan a high res image of the surface of the heat spreader on the heatsink)

Just to clarify:

• smoother does not always = flatter* (a chromed ball bearing is very smooth and shiny, but is far from flat) and
• flatter does not always = better contact UNLESS BOTH mating surfaces are "flatter".

The ultimate goal in lapping is to get both HSF and IHS to mate perfectly. This can be done with any shape profile**, but it is incredibley more difficult to create mating shaped profiles than it is to simply make both surfaces flat. So you want both your HSF and IHS to be smooth AND flat for better contact. The glass/mirror provides a flat surface***, and you use progressively high grit sandpaper it make it smooth.
 
The reason you really need to lap both HSF and IHS is to get the two surfaces mated (i.e. exactly the same shape). It's very possible to increase your temps by only lapping one of the two (as in the case of the TR U-120, which has been designed to better mate the concave IHS) because you could detrimentally effect the surface match.
 
 
* - flatter DOES = smoother, though, because the flatter a surface becomes, the smaller the "bumps" become and the smoother it becomes (which also makes it shiny).
 
** - as mentioned above, that's why the people at TR make the U-120/Extreme with a slightly rounded bottom, they are trying to mate the HS to the IHS because the IHS is not flat.
 
*** - if you're REALLY fanatic about a flat surface, you can get granite surface plates that are manufactured to have very flat surfaces (within one ten-thousanth of an inch). I'm not suggesting you buy one (they're rather expensive) but if you ask around you machinist friends, you might find someone that has one you could access. Just thought you might like to know. 8-)

I thought to myself that if I'm going to lap my HSF then why not do the CPU to. It doesn't make since to only do one.

 
Very true.  You're talking about two surfaces mating to form a thermal transfer.  Doing just one is kinda silly; you need to do both halves!

I only did the heatsink  on my main pc (stock cooler was really bumpy like the image I posted)
 
so i  sanded that to a smooth finish and I got around a 3-4C reduction in temperature,

AutoZone, Advance Auto Parts, and the like.  Call ahead and ask what grits of refinishing (car paint) sandpaper they have.  I got 400, 800, 1000, and 2000 at a local Advance.

A very nice post.   Good pics and explanation.   I am now confused,though,  because I just read another post where a thermalright extreme ultra 120 reviewer found a convex heat sink, but after lapping both the cpu and the heat sink,  got no improvement. http://www.cryo-laboratory.com/for [...] eview.html  
Is there some explanation?

Dunno what to say about that link...

Interestingly, after lapping and such (razorblade test shows it's flat as can be), while all cores dropped substantially thus far, cores 2 and 3 are 2c higher than cores 0 and 1 under prime95 load.
 
I'm thinking maybe I should reapply the AS5 maybe my line wasn't long enough.
 
Zalman 9700, 3240mhz right now, 49, 47, 51, 51 after an hour of Prime. Can't complain 1.35v I plan to go for 3.6 at least though which is why I'm lapping, etc. now. 3240 is just the "burn-in" speed.
 
I did notice the Zalman is seemingly a bit convex, at least the outer edges are a bit higher than the center area (using the razorblade test, and we're only talking about a human hair higher--just a sliver if light comes through), but since the chip doesn't cover the entire base of it, I am thinking that if ANY heatsink was pre-lapped and not in need of a lapping, it would be that one. For all I know the part that actually makes contact with the IHS could be bear perfect (it's certainly machined better than most other HS's I've ever seen before).... so I dunno.
 
I did a test where I installed the heatsink, waited a few mins then removed it... the grease was spread very evenly. Then I cleaned everything and reapplied the grease to the IHS and re-installed.

It settled down a LOT after another hour... 44c straight across Couldn't be happier. It probably is just taking a bit for the grease to finish spreading.
 
If anyone out there is on the fence about lapping, DO IT. I've dropped about 8c already and I haven't even begun to "cure" yet!