My lapping story

[CompuTamer]

Okay, so, two things happened today. I got my contacts (I can see!!!) and i lapped my CPU and my DK. Here's the story!

Before temps:

This is at 3.9. I think i did lower the voltage some between now and then, but it didn't impact temps all that much. I was still seeing 99C spikes when the heater came on. (it blows right into the front of my computer :/)

So, on to the lapping. I went between the CPU and DK. I'd get bored with one, and lap the other some, then get bored with it, and lap the other, ect ect.

After 10minutes or so on the CPU:

Already i know this may take a while...

After about a half hour...

Hour and a half...

I don't really know how long now... lol

That little spot on the bottom about drove me out of my mind...

Still working on the spot...

Okay, enough of the CPU.
After 20 minutes or so on the DK

 

[CompuTamer]

A half hour...

And after 6 hours of total work, here is the final result

Would've liked to go down further, but i was worried i may break through the heat pipes on the DK, and i ran out of sandpaper anyway, and my arm was REALLY sore, so i just called it a day lol



EDIT: Hmmm, just noticed that the fans cycle up and down now instead of running flat out all the time. When they peak temps drop to roughly 69-71 and then slowly climb back up to 75 as the fan slows down, then it speeds up again, and the cycle continues lol

 

[emceepecks]

holy crap, 99c spikes???? yikes!!

nice lap job btw, very shiney

 

[CompuTamer]

Yeah. I was running 66C under load when i first put it back together but then it went up to 75 and i can't get the load temps back down now

Thanks. I wish that i had some more 1000 grit sand paper that way i could've gotten the little scratches out, but i think that can wait a bit lol

 

[Rich'[ard]]

so you got a 5 degrees difference out of that?!
that's pretty damn good i reckon.
nice job

 

[xtreeme]

Depending on the flatness of the CPU ; if it is "almost" flat I start with 1000 grit - if it's not I start with 800 grit.
I use a glasplate on which I fasten the emery paper, and I use a "8" motion during the sand job - which I can see you haven't....
I can also see that you have been "tipping" during the sand job since your corners on the ihs aren't sharp.

If this was mine cpu I would relap it, started with 1000grit (8 motion) - when it is "pretty ok" I use water - just a little bit which you apply and spread evenly on the emerypaper with your fingers.

At last I use 1200grit and water - and change the paper often.

With older cpus as socket 478 I used wd40 instead of water since that cpu have a little hole in the ihs

 

[CompuTamer]

I did try to have a figure 8, but i kept on messing up the pattern :/

It was my first lapping job, and i'll probably redo it in a few months/weeks. I just ran out of time today, and figured that i'd go ahead and see what it got me. I got a 5 to 20C difference depending on the load and if the stupid heater was running.

 

[onefstsnake]

Try picking up some 1200-2000 grit at an auto parts store.
Thats where I get mine.

 

[CompuTamer]

I got mine at Autozone

I figured i wouldn't need anything over about 1000 grit. That was the last one that i used, but i will probably redo this soon.

 

[Conumdrum]

Methinks ya did great. 5C for mindless 6 hours back n forth is priceless.

You've lapped enough if it's flat. Tests actually proved that over 600 grit is more than enough, but it's sure prettier when shiny eh?

 

[TaCoSwInDeZ]

Yes it is hehe, and AMD is the epitome of mirror because after 2000 grit I buffed to a mirror shine with Turtle Wax liquid Rubbing Compound. You can get it at any Automotive store. It comes in a green squeeze bottle with a black flip top. Its an excellent metal polish. So far i've tried it on Aluminum and Copper, it works very well on both surfaces.

 

[CompuTamer]

Methinks ya did great. 5C for mindless 6 hours back n forth is priceless.

You've lapped enough if it's flat. Tests actually proved that over 600 grit is more than enough, but it's sure prettier when shiny eh?

Lol certainly mindless... once i got the figure 8 down, i just started listening to music until i ran out of sandpaper.

 

[Mpegger]

As has been said before, anything over 800 grit is overkill and counter productive. Temps can actually start to go up because there are no microscopic valleys for the TIM to fill in, which is what its designed to do. Your goal is just a flat surface, not a mirror surface, so you did just fine with your lap job. <insert pole dancing smilie here>

 

[CompuTamer]

i actually started it up with no TIM at all, just out of curiosity, and i found this in the BIOS temp monitor... really astounded me:
This is with no fan or TIM right after startup

and then after about 10 minutes (while i was cleaning up the mess i made in the kitchen)

and here's the mess i made in the kitchen lol

 

[bing]

Nice result & +1 to what Mpegger said, its not the mirror finish that most important, but the flatness and/or mating surface to the IHS that is really matter.

Since you messing the whole table, and if you have a piece of thick glass, suggesting alternative flatness checking here -> "IHS + Ketchup/Ink Dye are Tasty !" ?

 

[xtreeme]

huh.... that's new

A perfect ihs / cooler surface will be the one with no pores whatsoever and using no tim at all, but that is impossible to achieve.
The goal lapping a cpu is to get it flat and to remove as much of the pores as possible, when you look at the ihs through a microscope you will still see a lot of grooves after you are finished lapping with 1200.. hell yeah even 2000 for that matter.

Another cruicial thing is to choose tim wisely, as5 and alikes just don't cut it on lapped surfaces, the thumbrule regarding tim is and will always be ; a thick tim fits coarse surfaces and a fine surface needs thinner tims.

I have tested this - I don't make this up

The word is : you can never get a too fine surface..


Some of the heatpastes I have tested :



Now I'm using Indigo Xtreme on a lapped 920 D0 @ 4600 24/7

 

[bing]

Actually this topic is not new, here, an oldie but still goodie imo -> Why Heatsink Polishing Might Be A Bad Idea (this article dated back in 2001)




.

 

[xtreeme]

Actually this topic is not new, here, an oldie but still goodie imo -> Why Heatsink Polishing Might Be A Bad Idea (this article dated back in 2001)
.

haha - I had that cooler, by modifying the fastening a bit I was able to put a 92 mm fan on top of it

In respect I deleted this link

Which tim did he use ??? he never says .... and lapping only the cpu and not the cooler too is a thing I wouldn't have done. Another thing not to forget - is that the firmer contact between ihs and cooler the better heattransfer (until a certain limit of course).

So modifying the cpu fastening to get firmer contact always improves heattransfer - but take care not overdoing it since damage can occur (and by lapping you will of course loose "some" of the pressure between the cpu and cooler).

EDIT: A tim will never increase heattransfer as many seem to think, the tim's primary goal is to remove air between the ihs and the cooler (fill in scratches and pores). So naturally - the less scratches and pores - the better heattransfer.

I have a very special procedure on how I put on the tim that I don't have seen anyone else use. I might do a write up on my procedure one of these days ---- next year

 

[TaCoSwInDeZ]

Hmm...I just tested this theory myself. Before I installed the stock HSF onto my I5 the first time I mirror polished just the heatsink like I mentioned using IC7 TIM. But out of curiousity I managed to find a peice of 400 grit sandpaper and so I removed the heatsink and quickly hit it with the 400 grit to get rid of the dullness and what do ya know, using a think TIM like IC7 defintely likes the rougher surface. I've gained about 2-4°C in full load temps. No doubt that a slightly rougher surface yeilds better temperatures.

 

[xtreeme]

Hmm...I just tested this theory myself. Before I installed the stock HSF onto my I5 the first time I mirror polished just the heatsink like I mentioned using IC7 TIM. But out of curiousity I managed to find a peice of 400 grit sandpaper and so I removed the heatsink and quickly hit it with the 400 grit to get rid of the dullness and what do ya know, using a think TIM like IC7 defintely likes the rougher surface. I've gained about 2-4°C in full load temps. No doubt that a slightly rougher surface yeilds better temperatures.

Which tim did you use ? - EDIT : sorry now I see which tim you used...

You are really talking against physical facts - the best heattransfer were if you could lap the surfaces so there were none groves/pores left and use no tim at all.
TIM does not really increase the heattransfer vs. two perfect surfaces - not even CoolLaboratorys MetalPad does that - and certainly not a ordinary tim

I suspect your cooler/cpu is lapped flat ; but it doesn't have the same height over the ihs anymore, that can easily happen if you don't use the 8 movement when lapping.
Or are inexperienced with the procedure.

In the first year at marine engineer school lapping was one of the things we had to learn, and I promise you - it is not THAT easy to get it correct as many of you seem to belive.
OP here have tipped the cpu under the sanding - I can see it of the pictures