Need help lapping

[Particle]

I received a pair of socket F heatsinsks last night for my 1U server, and I noted load temperatures were exceeding 75C after about 10 minutes.

Since I only have one CPU installed, I thought I'd take the other heatsink and lap the bottom. Big mistake. After about two hours of sanding, I've yet to level out the surface. The middle of the block is still being quite stubborn--only now starting to take scratches from 60 and 100 grit. The ripply machining marks are getting harder to see, but in the only place that matters on the whole block they're easily felt.

THERE MUST be a better way to do this. I've gone through about six sheets of 60 and 100 grit sandpaper, and it seems like they're only good for about 15 seconds. The next ten minutes accomplishes about the same as the first 15-30 seconds on a fresh sheet. Am I using the wrong kind of sandpaper? Am I doing something wrong? What gives?

[HuffPCair]

yeah i have wanted to see a guide on how to do this for some time now, but have never seen one. Hopefully someone here can show or atleast inform

[hokiealumnus]

yeah i have wanted to see a guide on how to do this for some time now, but have never seen one. Hopefully someone here can show or atleast inform

Guide to lapping (http://www.theraptorpit.com/forum/index.php?showtopic=1549).

Particle, since you're lapping a heat sink, you could get some automotive grade wet/dry sand paper and use a little water to help you along. Use a paper towel to sop up the water and metal particles when it stops sanding as well.

[HuffPCair]

Thanks a lot for that.

[Particle]

It probably doesn't help that my heatsink has a bottom area about the size of six normal heatsinks. The bottom is 4.3" long and about 2.5" wide, flat.

[Jor3lBR]

There must be a 'professional lapper' around the Forum, like you send him the cpu and he put's the chip on a lapping machine then send you the chip back. You pay for his work.

I know how to lap but it's just too much time consuming for me.

[Ozzfest05]

well you should not be using more then 2 sheets of each grit IMO make sure you are using metal sand paper or "automotive" you can use a couple of drops of water in the middle of the sand paper also make sure to use a flat surface like a piece of glass and to get even a horrible base lapped to a flat surface should take more then an hour of full sanding I do it 15 minutes at a time while im watching tv if you sand for 15 minutes straight after an hours worth it will be on the spot

[Particle]

Temps dropped by about 10C or more after the sanding. That's impressive considering I essentially gave up as early as I possibly could. I can live with 66C under load in a 1U server.

No more danger of my Opteron melting into a pool of molten silicon! :up:

[Jor3lBR]

Lapping is quite simple...:up:

Just start with high grit wet/dry paper, use straight lines for 15 minute intervals, rotate 90*, continue...:up:

hope this helps...:up:

Sweet Lapping there R3. I wish I had your patience. :up:

[zanzabar]

any rsx pics i thought that they had a bare die

[Particle]

Holy crud batman...don't forget the water. Water makes the process infinitely faster and more effective I've found. Anyone looking at this thread at a future date for lapping advice: Benefit from my woe--use water. It should not be treated as optional!

Start with a grit that matches just how bad your heatsink is. If you've got a 1mm bow, start with 50 or 100 grit. Use "emory cloth" instead of wet/dry (silicon carbide) or wood (normal aluminum oxide) sandpapers. It consists largely of corundum (aluminum oxide), but it's wet-proof, more flexible (eg: harder to ruin by accident), and available in lower grit numbers than normal wet/dry. It also seems to allow more sanding between unloadings (getting rid of the copper fragments).

[emoners]

is there a way to properly hold the proc when lapping?

[Particle]

I would be tempted to hot glue a knob to the processor and remove it when I'm done.

[Gautam]

Where might I ask do you find such fine grit sandpaper? Tried two hardware stores around town, one only went up to 400, the other had 600 and 1500, but nothing in between. (both in very small quantities too, but lots and lots of 220 and lower)

[Ozzfest05]

automotive stores carry high grit sand paper 1000 1500 2000 you dont need the in between for those three steps if you looking for performance you dont even need to go that high going to 1000 is fine after that 0 performance gain is had just mirror finished for bling

[Hypediss]

I remember reading another thread about lapping but the highest grade of paper the OP used was 800.. and saw that nething above 800 would produce negative results... can neone confirm that? or is it the higher the paper grit the better?

[tiro_uspsss]

could ya run it over a belt sander with the first few grits? :D

[HuffPCair]

LOl then it would only take like 5 seconds instead of 5 hours LOL

[noinimod]

I don't think it's necessary to go up to 2500.. unless you have tons of time on your hands
i did 250\500\1000 and it still dropped my temps by alot

[Vapor]

I remember reading another thread about lapping but the highest grade of paper the OP used was 800.. and saw that nething above 800 would produce negative results... can neone confirm that? or is it the higher the paper grit the better?The number of tests and mounts that would need to be done in order to confirm that is unfathomable to me....dozens at least. That's how hard the difference is to measure reliably. It's also totally dependent on TIM.

Long story short, past a certain point, the gains are mostly for looks.

[tiro_uspsss]

LOl then it would only take like 5 seconds instead of 5 hours LOL

:D yes siree thats what I want :D :up:

[Hypediss]

The number of tests and mounts that would need to be done in order to confirm that is unfathomable to me....dozens at least. That's how hard the difference is to measure reliably. It's also totally dependent on TIM.

Long story short, past a certain point, the gains are mostly for looks.

thanks for the much needed confirmation! :up:

[Jsunn]

is there a way to properly hold the proc when lapping?


I am looking for recommendations on this as well? Will masking tape on the pins cause issues?


Thanks,
Jason

[TheGanG]

I am looking for recommendations on this as well? Will masking tape on the pins cause issues?


Thanks,
Jason


May be... Use a paper mask for pin area...

[hokiealumnus]

I am looking for recommendations on this as well? Will masking tape on the pins cause issues?


Thanks,
Jason
I just used the plastic cover that came on the CPU. Unless you threw it away of course, which would be unfortunate.