since copper is so much more heat conductive than AS3, i think the claims of poorer performance with a lapped block may be errors caused by poor installation. the main issues i see:

1) rougher surfaces have channels that let the AS3 flow out while it settles, so ppl who goop on AS3 do better with rougher surfaces (supports claims of cooler temps after a "break-in period").

2) excess AS3 reduces the total area of copper in direct contact.

3) theoretically a smoother & flatter surface conducts heat better, but the gap between the HSF & core is so small that the performance is almost entirely dependant on the quality of the installation. i mean, it's easy to add 50 or so microns overall by abusing AS3. it is more difficult to reduce the overall gap the same amount by lapping.

example:
Mr. n00b laps his core to 1600grit, but goops on 2 "BB's" worth of AS3. Mr.n00b has an average 200 microns of AS3 between his core & block.

Mr.modson left his block with a 600grit finish & applies a nice thin film of AS3. thus modson has an average 150mm gap.

n00b spent 1hr extra on lapping, & now knows why his block doesn't work as well as modson.

let's change the focus for a bit to P4 procs. Should the P4 heat spreader be lapped? I know I used waaaay too much AS3 on my core, simply because it's hard to get even coverage on such a large surface... I'm going to pull the WB off and use a razor or something to scrape the AS3 off, and whatever remains is OK. I'm thinking that should lower temps sharply.

Where can I buy fine sandpaper. ...finest I can get around here is 600. After that, I have to use jeweler's rouge.

Impressions?

Yikes! Jeweller's rouge is not just an abrasive, it's a wax, kinda like polishing compound!

I had a hard time finding anything above 600 grit myself. I found 800 at a hobby shop, but I moved since:cry:

Anyone have any ideas?

Honestly, 600-grade paper, used properly, leads to good results. Not mirror yet but nearly.
To get further if you cant find 1200-grade, try lapping with an abrasive paste. If my memory is correct Dremel makes such a paste which goes with special 'wool' tools.
Lapping the p4 heat spreader: since it IS possible without taking any risk, i'd do it, definitely. Just to correct surface geometry.

Auto parts stores ussally have good paper. Mine stocks up to 1400 IIRC.

Very good point. I once looked at a Duron under a steromicroscope and was amazed how rough the surface is. Unless you lap the core, I think a SMALL amount of compound will always be needed.

Here's a link to get up to 2000grit sandpaper at reasonable prices:http://www.wicksaircraft.com/gotopage.php?page=29. $1.33 for one 2000grit.

It can be done, with the block I was telling you about earlier, their was a version of it that supported it. The pelt was just clamped inbetween the block and a coldplate, similar to the maze series. The probolem is that you actually don't get much surface area compared to a block, so you need a ton of flow to get it to work.

-WireX

I found up to 2000 grit in a local construction store, it's used in lots of applications where you want paint that looks good and also is extremely durable (cars being the most obvious).

I have to support the "Finer is better" variation on this, the counter-argument being that there are no microscopic crevasses for the goop to fill, which is simply wrong. Imperfections in a 2000grit sanded block can be seen with relatively simple methods:

http://www.procooling.com/~iceheart/2000_scratches.jpg

This block looks perfectly mirror finish in normal lightning conditions:

http://www.procooling.com/~iceheart/2000.jpg

The fact that many people has reported lower temperatures with less sanded blocks have to be related to the "too much goop" issue which was mentioned earlier. At some point in the future I intend to do some experiments with this.

Thanks to everyone for their contribution.

I think that we can safely say that the best lapping (to a mirror), combined with a PROPERLY applied thermal paste (preferably Arctic Silver 2 or 3) will achieve the best results.

Well, here's something for ya.

I had a P4 w/ a semi-mirrored Maze 3. I was using way too much AS3. So, I removed the block and used a credit card to scrape away any excess AS3, and the tested things out. Well, temps were a few degrees worse. Not only that, but as soon as the CPU had any load, the temps shot up, and as soon as it started idling, the temps fell quickly.

So, I took everything apart, and looked at the chip. The built in heat spreader was not flat... not even close. It was high around the edges, and a little high in the center. So I lapped it. I tried 600 grit, the 400, and finally 320 started helping. It was waaay off. FYI, the heat spreader is made of copper. I don't know what Intel did to make it silvery, (aluminum perhaps), but now its fairly shiny copper, and flat.

I put things back together, and a used a Kleenex to daub on AS3... a veeery thin coat, on both the CPU and the WB. I worked the WB around to get everything to settle in, and then cranked the springs down all the way.

*Temps haven't changed a bit. Maybe just a tad higher, but it feels warmish in this room, so who knows. Maybe I'll add more AS3... maybe I'll take it all off. I dunno. I only used enough this time to make it look a little silvery. hmmm. If nothing else, I probably voided my warranty. All is not lost. :)

hey ECU, try lapping the block with some 600 grit and some WD-40 and take temp readings. and see if there is any difference.

NEVER do that. DO NOT use WD-40 to facilitate lapping.

http://www.articsilver.com/arctic_si...structions.htm

Use plain old water.

BB2K:
You had the answer on that site from the beginning !
Quote:
"
The flatter the mating surfaces, the thinner the layer that is required. Stock processors and/or heatsinks with normal surface irregularities will require a layer 0.003" to 0.005 thick as shown below to fill the resultant gaps. (Equal to the thickness of about 1 sheet of standard weight paper.) Properly lapped heatsinks with mirror finishes will only require a translucent haze.
"

To the last sentence one may add "together with a properly lapped chip".

Yeah, but everybody's got a theory, even Arctic Silver. I started this thread to get everyone's opinion, and to make it perfectly clear that the mirror finish is best.

Still, it seems to me that the surface of the core is now the issue. I don't want to get into the Durons, but the current Athlon XPs, especially the latest one with the Thoroughbred core, seem to be very flat, even shiny. I guess someone is doing something right, I'm just wondering why it took so long.