[Exempt]

Hello OC techies, Ive been coming to this site since it first came up waaaay back when so I finally decided to register after reading the Tips part on Lapping. It was a nicely composed article, maybe a bit to in depth for noobs, but hey at once we were all noobs...so no knocking them. Anyways, I wanted to add to that article because lapping, which is simliar to lathing is COMMON practice for me on all my Intel CPUs. I will explain why..

1. Lapping or sanding the CPU can greatly decrease heat when overclocking.

2. Lapping is relatively safe and if you have the correct sandpaper..its free..

3. I said its free and it reduces heat from the CPU. do I need to say more?

ok so heres the deal, this is what youll need and from here on, know that I am referring to INTEL cpus not AMDs...

1. A intel CPU socket 478 or 370 if your taking it old school style, preferably the socket 478

2. 220 grit & 600 grit, and if you can find some 1200 grit sandpaper, they cost about 1.69 US dollars for 6 sheets each pack, and you can get them at your local hardware store.

3. a neoprene foam piece large enough to cover the surface area of the gold pins on the back of the CPU, or a piece of foam that came with the CPU.

4. preferably a piece of glass at least 12 inches by 12 inches wide or a really flat surface, I mean really flat, not concrete flat lol.

5. any good tape, scotch, duct, you get the points

ok thats it, not much materials needed. Heres what to do.

1.Take the Neoprene foam or CPU protective foam, and gently place it on the back side of the CPU's gold pins, make sure that it covers the pins decently.

2. tape a full piece of the 220 grit sandpaper, and tape the edges of the sandpaper to the piece of glass.

3. place the glass on a flat surface, I use my granite kitchen countertop.

4. now take the CPU and gently move it in a circular motion on the sandpaper.

"NOw wait a sec, how do I know how much I should lapp this sucker before I ruin my CPU?"

thats easy, on the 370 chipset, you should NOT use the 220 grit sandpaper unless your hardcore, and is you do decide to use 220 gritt on the 370 chipset I wouldnt do it for more than a few rotations cause you have to remember you still have to hit it with the 600 grit in order to smooth out the visible scratches.

on the 478, use the 220 grit sand paper until the entire surface is copper, then stop and get ready to switch sandpaper.

If your hardcore like me, I go for quite a while, when I use this method, otherwise I will take the CPU and go get it lathed but thats a whole nother' story...

5. Once you reach the copper on the CPU take off the 220 grit sandpaper and switch it to the 600 grit, tape it down, and start going in circles again.

If you continously look at the surface of the CPU you will notice that the deeper looking scratches go away rather quickly and after a few minutes it will start to look smooooth..oh yah thats what we want.

6. Now if you went the extra method, getting the 1200 grit sandpaper, (this isnt neccessary at all ...just a perfectionist thing), then you will follow up with this last switch out of the sandpaper to the 1200 and move the CPU in circles again until you have a semi mirror finish.

now clean up your mess, make a double check to make sure that there is not metalllic dust on the gold pins from the sanding. ( I use a compressor or compressed air ( Dust off ) ).

7. clean the CPU with cotton swabs and isopropyl alcohol nicely.

8. apply thermal compound and overclock that sucker.

only bad thing to this is...if your like me, which a lot of you are, I have more than 9 rigs, and I am constantly swapping out CPUS for newer ones or whatever...I sometimes forget which CPU is which because now there is no way to find out but to boot the CPU at default to figure out its real clock speed lol.

My latest experience, I took my new 3.0 800FSB CPU and did this to it, using artic silver III and a prometeia mach II unit I have overclocked it to 4.1 , idle temp is - 46 with load it is -38 -35, on a Abit IC7-G MAX II mobo vcore at 1.8, mem will not OC at all at this speed, neither will PCI or AGP slots, but man it is FAST... I can post pics if you guys want..more than glad too...

With Air cooling, on a volcano 7+ HSF , I lapped my 2.2A , I overclocked it to a nice 2.8 with vcore at 1.675, temps are 28C idle, and 34-36C loaded. using one vantec stealth intake and one vantec stealth exhuast....


ANyone have questions lemme know

[TC]

Nice how-to here. I wonder if anyone has tried lapping the actual core of a P4 with the HS removed? That would be hard core.

[stan03]

thats a nice guide... sticky?

[OnDborder]

Originally posted by Exempt
I can post pics if you guys want..more than glad too...

That would be very cool of you..
Nice post btw.

[NiTrO bOiE]

Nice guide. Sticky material in deed. You would have to have alot of time and courage the lap the actual core of the chip. I don't think it needed though because it is mirror like already, in fact I can see myself in it.

[Exempt]

[Exempt]

that was before, this is after.


notice the reflection... srry but I didnt take pictuers of the entire process, but I can in a week, when the 3.2 comes...

[Exempt]

BTW I dont know if you guys check out Deviantpc.com forums, but there awesome, my name over there is "skuba lee" short for Skuba lee dungaree ..just a funky name lol

[Maxvla]

can't view without registering.

send your pics to maxvla@maxvla.com (me) and i'll host them for you so you can post them here.

[tio]

Nice writeup Exempt. From my experience though, lapping a processor's IHS in a straight line in a single direction (and then turning the processor 90° after every 10 or so passes) yields slightly better results than lapping in a circular motion.

Lapping in a circular motion actually makes the IHS on the processor slightly convex in shape. For me, this was really noticeable when applying thermal compound to the IHS. I found it virtually impossible to apply a thin and even layer of AS3 because the edges of the IHS were actually lower than its centre.

After re-lapping the chip in a single direction and rotating it every...ect (you get the picture), I not only found it much easier to apply thermal compound, I noticed about a 1 - 2°c drop in my load temp.

[NiTrO bOiE]

Originally posted by Exempt

only bad thing to this is...if your like me, which a lot of you are, I have more than 9 rigs, and I am constantly swapping out CPUS for newer ones or whatever...I sometimes forget which CPU is which because now there is no way to find out but to boot the CPU at default to figure out its real clock speed lol.

Would it hurt cooling performance if you were to write on the chip with a permanent marker that says its speed?

[tio]

Yes, in theory it would. The oils from the marker would only harm the transfer of heat from the IHS to the heatsink or waterblock... ect.

[DOUG]

Use a sharpie,there's no oil in the ink................

[Bob_Sherrill]

My experience in lapping other objects for high performance jet engines in the military, call for lapping in a figure 8 motion. It results in a flatter result.

Bob

[Tical]

I've lapped both ways, in 1 direction, & figure-8 & believe going in only 1 direction achieved the desired results in less time.

[NoxioN]

should sand in a figure 8 motion.

why not write on the green area of something of the chip to label it?

[tio]

Lapping in a figure eight will make a surface slightly convex. Do a google search on lapping surfaces to find out more about this.

[Flakk]

what is the benefit.. like, before lapping and after? you only give the lapped results, so theres no way to measure the gain of lapping (so i can see if it's worth it or not )

on average how much would you say you gain from a procedure like this compared to an un-lapped cpu

[Exempt]

Use a circular motion, go about 10 times in one direction then reverse the direction, as long as the surface is flat, you should not have any type of convex or uneven lapped surface on the CPU, and if there is, it would be so minuscule that it would not affect performance.

All depending on how much you lap the CPU , differs the beginning and end result, if you do a minor lap to the CPU, then your temps might drop 2 degrees, but a good lap can take it down up to 7-8 degrees, and every one counts,

[ol' man]

I reiterate.

from me once upon a time
Here is a good lapping kit.

1 12 inch metal file
1 8 x 12 mirror or piece of glass
various sheets of 600grit~2000grit
1 flat wood table or desk to set glass plate on.
1 glass of water to hold down wet/dry sand paper
1 towel to wipe away the aluminum and copper filings mixed with water on the lapped HS.

Instructions.

For IHS simply use 600~1000 grit on wet/dry. It is not necessary to lap all the way to the copper. I find lapping the IHS does not help much anyway.

For most heatsinks I use a 12 inch super flat file and lap in 90 degree alternating directions till surface is flat going both ways. Then I lap HS on flat glass/wetdry apparatus till shiny and no ridges left. Lapping is done in a back/forward motion. No circles until final buff using ultra fine wet/dry. Every say 8 lapps I then turn the HS 90 deg. and lap again for another 8 lapps. I have used this method for over three years now and it is the only way to not get that dang bowl affect going. You know when the edges wear down faster than the middle. If this happens the best way to get rid of the "sphere" affect is with a file.

I am the only one I have heard of that uses a file(1"x12') to get the initial amount of material down but it works alot faster than using heavy grit and it does not leave large pits in the copper. It actually leaves the copper quite shiny.

In my experience going in a fig 8 will leave your IHS or HS bowl shaped. Don't ask me ask my "pentium pro" I lapped for like 15 hours into a bowl

When I use a file it takes about five minutes to make a HS flat no matter what it is copper or aluminum. Copper is easier to work with.

Here are two short movies explaining what I mean.

http://www.shackspace.com/~chemhaqr@.../takedown1.wmv

http://www.shackspace.com/~chemhaqr@...om/polish1.wmv

That is 1000 grit wet/dry sand paper which is wetted and stuck to the mirror using the simple adhesive and cohesive properties of water.

[DaWiper]

How much lapping can the core take before you kill the cpu? Just wondering?

[tio]

The core? Not much at all. And anyways, you won't be able to make the core any flatter by lapping it, since cores themselves are typically extremely flat.

[CeleronO/C'D]

a bit of mother's metal polish from the auto parts store and a finger's worth of elbow grease will get the thing smooth to a mirror finish, be sure to get polish, not a "filler" because the polish is "3000 grit or so sand paper" in theory and will put the smoothest finish on it. do the same to the heatsink. results will be greater than any sandpaper. dont forget to seal that combo with Arctic V

[CeleronO/C'D]

there is also less risk involved

[ZGOZZ]

cool I think I just read a new project to do.

[ZGOZZ]

Originally posted by CeleronO/C'D
a bit of mother's metal polish from the auto parts store and a finger's worth of elbow grease will get the thing smooth to a mirror finish, be sure to get polish, not a "filler" because the polish is "3000 grit or so sand paper" in theory and will put the smoothest finish on it. do the same to the heatsink. results will be greater than any sandpaper. dont forget to seal that combo with Arctic V

without sanding?

[gone_fishin]

Lapping the IHS is not a good idea, here's why. The IHS has a phase change tim material under it which compresses under the load applied when attaching the heatsink and the die becomes hot when first used. If you notice, a brand new processor will actually be convex (or protruding) out from the center of the IHS. When load is applied, the phase change material under the IHS has a certain amount of presure applied to it. If you remove this bulge in the new IHS by lapping, not the same amount of presure will be applied to the phase change material and a thicker tim will result between the bare die and the underside of the IHS. The edges of the IHS will transfer the full load to the chip package and not to the phase change material on top of the die.

Claims of 1 or 2C improvement are within the margin of error between mountings. Nobody ever reports an increase in temp, they just remount till they get a better reading.

I stubbornly lapped one and was proud of how "flat" I made it, thinking how shoddy Intel was for making a wavy topped IHS. In reality the only thing it did was make me feel better at the time. I ask you, what good is having the secondary tim improved if the initial tim is degraded?

[Exempt]

You are correct when you say that 1-2 degree differences could be mounting error differentials, but that is mostly with air cooling, and it also depends on how far you actually lap the IHS. the IHS plays the part of protecting the acutal core of the CPU, some people just pop off the IHS completely, BUT you are at a much higher risk of crushing the core and rendering your CPU useless. When I use my before and afters, Im using a few methods, I use air cooling (the least though), water cooling, and a prometeia mach I and mach II unit. I notice the largest improvements with the vapor cooling units, due to the fact that I lap the s**t outta them, polish them up to a mirror shine... but it all depends on how far your willing to go for those few extra mhz, or to cool it just that little bit extra...

[gone_fishin]

You missed my whole point, you degrade the first tim joint to improve the secondary one by lapping an IHS.

quote "but that is mostly with air cooling" end quote

Waterblock mounting will have just as much error as air heatsinks between mounts.

[Exempt]

lol I think you missed what I was saying, when I said that using air cooling a 1-2 degrees difference is very small, and can be blamed for by mounting, but when you lap the CPU and use water cooling or vapor cooling and the tempurature differences are more along the lines of 8-10 with water or 10-12 degrees with vapor, then it goes beyond calling it mounting mishap...

irregardless of the tim, whether it is not anatomically flat, you are essentially shaving off more metal inbetween the core and the heatsink or waterblock or Vapor unit copper block. If it was a bad method then macci or G-Vex, or any of the top world reknown overclockers who blow away futuremark or 3dmark scores...wouldnt use this method.