[Guide] Lapping your HSF+IHS (Pictorial). - Overclock.net

TriBeCa · 08-16-07

Hi everyone. I recently did my first lapping job, and I thought it might be useful for me to share the experience with those who haven't done it before and are considering it, and for those who have done it to compare their experiences against.... I should also mention that I'm usually exceptionally bad at this kind of thing, but I ran into no serious problems (except that it took longer than expected). So unless you're seriously inept or suffer from some kind of motor disorder, you shouldn't have any trouble

I took a TON of pictures, so you'll have plenty to look at as I go through it

Here's the rig I was starting with:

In the first pic you can see the 80mm intake fan mod I did on the case to drop temps. But with ambients in my apartment reaching >26C sometimes, my OC had ceased to be stable. Load temps on the OC where up to 64C, and it would sometimes idle at 50C

So I've been running my rig on stock settings for a while, where it idles at 36-40C and loads over 50C.

So...I decided to replace that Tt Golden Orb II with something a little heftier...a Thermalright Ultra-120 eXtreme!

Now, I'd done my research around here and learned that the base on these things is concave (by design). This was done to improve contact with CPU IHS's, which are usually a bit convex. However, several people around here have reported good results from lapping, and I suspected my CPU would not make good contact the heatsink. Turns out, I was right:

If you can't make out that pic, it's the 120-X sitting upside down, with the CPU (also upside down) resting on it, with a lightsource behind them. All that light you see coming through in between the CPU and the heatsink = terrible contact. I definately wanted to do something about that....

I ordered the arcticlean bundle from easypckits.com, as this is aneconomical way to get both course and VERY fine-grain (10 micron grains, anyone?) sandpaper with glass (to use as a flat surface), a tube of AS5, and arcticlean 1 and 2 (arcticlean 1, in particular, is awesome for dealing with the some of the nastier thermal compounds out there). As an aside,

in a big way to Dave at easypckits. He was awesome to work with. When I emailed him expressing concern that my order wouldn't arrive before the weekend, he was very responsive and ultimately ended up shipping my kit overnight at no extra charge to me. And then when it got here, there was a free bag of cable ties as well

So, here's the 120-X before I started:

First thing I did was to add a slight bevel to all the edges of the contact area, to reduce stuttering and the chance of it catching and tearing the sandpaper.

I started with 180 grit, as it's the coarsest and will remove the most material as I try and get the base flat. The lower grit stuff is for polishing the base to a mirror shine.

Here it is after the first bit of sanding:

Some things to note in the above pic: a) I'm wearing gloves, as the heatsink fins cut into your skin and are quite uncomfortable--the gloves make it much easier to hold, b) the fact that the outer rim is turning copper first verifies that the base is indeed concave...BUT c) copper at the very center of the base is being exposed as well, suggesting that in the middle the base is actually conVEX. This makes some sense, as contact directly over the processor core is the most important, and this ensures that the two will touch despite the HSF's concavity. But it begs the question as to why to make the thing concave in the first place. It would seem that a flat base would ensure contact at the center of the convex IHS just as well...and would result in greater overall contact area in the center.

Anyway. The sanding motion I settled on for >90% of the time spent flattening the base was a two-handed circular motion, with virtually no downward pressue (the weight of the HS is sufficient, and downward pressure tends to cause stuttering). I'd switch every few minutes from clockwise to counterclockwise, and every so often I'd rotate the heatsink in my hands. You need to switch it up like that to ensure that slight imbalances based on the motion you're using don't result in a base that isn't flat.

Figuring out how to affix the sandpaper to the glass was a bit tough. At first I tried taping the glass and sandpaper down to the newspaper, but the water ended up lifting it right off. I only had electrical tape and scotch tape available, and the solution I settled on was to do four strips of scotch tape, one on each side, affixed to the bottom of the glass.

An added advantage of this setup was the the tape actually held the water on the sandpaper better, as you can kind of see in the above pic. Copper filaments in a pool of water look pretty damn cool, imho

Here's how the lapping progressed:

Now. Before starting, I had read around these forums and seen people saying their lapping had taken anywhere from 1-2 hours. I don't know if I was doing something wrong, if I got a particularly bad HS base, or what, but it took over 3 hours to get to the point you see in that last pic. I was kinda shocked. I suspect I should've moved to a finer grit sandpaper, only because I'd worn down the 180 too far. A fresh sheet of 220 would probably have been courser than the 180 after 3 hours of use. But that didn't occur to me 'till the very end.

And ya know what? The last bit of zinc coating you see remaining in that last pic.... That took over an hour to get rid of, from the point you see there. This is what makes me suspect I might've had a particularly bad surface to start with. At any rate, I FINALLY got here:

From this point on I only spent 5-10 minutes at each grit. This went quite quickly. Getting the mirror finish once you have a flat surface is actually quite easy. From this point on I used a straight back-and-forth motion, although between grits I switched up which direction I did it in. This is because a striaght motion produces a better mirror finish. Here's 400 grit:

1000 grit:

25 micron. You can see it's starting to get a mirror finish:

Based on a post here, I didn't use water with the micron papers (finer than 1000 grit). I can't remember who it was, but someone said they could only get a mirror finish by keeping these papers dry. You can see these finer papers, when dry, respond to the heatsink surface VERY differently than the grittier sandpapers:

From 25 micron I went straight to 10 micron (the finest available), and this worked fine. You can see it doesn't pull nearly as much material off the HS:

Here's some pics of the finish:

So...I'm done (%$#^$@#ing finally!). So how's the contact with the CPU IHS?

Still crap!

Ok. I had planned to lap the CPU as well...but I did want to check. If you've got a lapped 120-X and you haven't seen much improvement in temps...just look at the above pic to see why

TriBeCa · 08-16-07

It's surprisingly hard to toss your CPU in a puddle of water....

Here we go...

You can see the CPU has a very different effect on the sandpaper. It's much easier to move it around because it's smaller, and you need a bit of downward pressure because it's so light.

Now on to the finer papers:

All this took under 2 hours. Here's the mirror finish (I think this look particularly cool on a CPU

):

I had a chance to verify that getting the finer grit papers wet actually ruins the finish. Some water got on the paper as I was finishing up with the CPU, and it invariably made the finish quite muddy looking. Here's the 10 micron paper when I was done:

All of those solid patches are places where the paper got wet...you really want to avoid that. And, if you can believe it, those oddly shaped lines at the top of the paper...were caused by a couple of HAIRS that got between the glass and the sandpaper! The SLIGHTEST bit of anything on the glass will show up.

One problem did arise here. Copper filiment laden water got between the CPU and the plastic backing:

I rinsed this off with tap water, dried it with a coffee filter, and then went out and bought some dust remover (that weird gass stuff that comes in a can) and sprayed it off. This was also on the back of the CPU (no pic because I'm nervous about having the contact area exposed), and I wiped that down with a coffee filter and sprayed it with the dust remover as well. I had no trouble once everything was installed and running, and there was nothing visible on either the CPU contact area or the plastic backing.

So...how's the contact between the HSF and IHS now???

IT STILL SUCKS!!!!

I finally hunted down something in my house that I trusted to be a straight edge...the back side of our chef's knife.

The CPU is totally flat...thank goodness. I didn't want to mess around with it any more if I didn't have to....

So I didn't get the heat sink base flat

I'm starting to run out of good sandpaper at this point, but I went back and started sanding again. In less than an hour I get here:

YAYYYYY. Finally they make good contact! You can see on either side it's not perfect, but it's a TON better than it was, and I'm not willing to throw another hour or two at it to get the contact on the sides better. Besides, it's the contact over the processor cores that matters....

Time to install the 120-X....

Used the line method with AS5:

It's a really impressive looking cooler, I must say:

...and it's REALLY huge with a pair of 117.5CFM panaflo's on it:

I've heard of people using all kinds of stuff to attach an exhaust fan for a push/pull effect...but if you just use one of the included clips on each of the fans they stay on fine...as long as you don't kick your case around much.

The 120-X *barely* fit in my case, and unfortunately the intake fan for the 120-X interfered with my intake fan mod

So I just moved the fan to the outside...I wasn't gonna let all the work drilling holes in the plexiglass siding go to waste

So...results? Unfortunately I didn't test the cooler before lapping...but over Tt Golden Orb there was a whopping 15-16C drop in temperatures across the board. The one exception was idle temps at stock, which obviously couldn't go below ~2C above ambients (stock idling was at 28-29C, lower than I've ever seen on this system). Stock load temps didn't go much over 36C, which is around where it was idling before. I went back to my previously stable OC (stable before the summer), and it loaded around 45-46C, though it appeared to be climbing a bit. Unfortunately it still wasn't orthos stable at 10 minutes, so I couldn't find peak load.

I decided to get to work trying to get that OC stable again. First thing I decided to do was update the BIOS, as I was still on the stock version. I flashed the BIOS using the BIOS EZFlash utility...and got a bad flash which fubared the board. You can see my discussion of this in other threads here, but the board has been RMA'd. So I can't give further results until I get the board back from ASUS

Anyway. I did take some pics of the distribution of AS5 on the HSF and IHS when I stripped down the mobo for RMA.

You can see it's quite different than you usually see. Part of why you see it in that particular pattern, I think, is because of the way the 120-X is mounted. When I released one screw, the whole thing shifted to one corner, then the second screw and it shifted again, and then a third. So I think that's what created the semi-triangular shape you see. Nonetheless, you can see VERY clearly that the line method does distribute the TIC quite well, and despite the distortion from the unmounting procedure it does appear that there's *very* little TIC in the center of the HSF-IHS contact. This is ideal, as you want straight copper on copper over the cores as much as possible.

So, that's all for now. Once I get my board back I'll work on my OC and post back with how high I'm able to get it with the new cooling. 3.2 GHz (400x8) was DEFINATELY my limit before, based on CPU temps (even in the winter I was at >62C under orthos load). I suspect part of the instability I'm seeing may be due to heat on the NB, so I've got a new chipset cooler coming, and I have a tiny active cooler for the southbridge.

That's all for now!

MGX1016 · 08-16-07

Wonder why the EZ Flash fubured?

Very nice though otherwise.. I may end up lapping my future rigs

TriBeCa · 08-16-07

I dunno...but someone else said he had the same problem flashing to that bios on this board.

If you can believe it, when I called ASUS the tech support rep tried to tell me that 15% of BIOS flashes go bad like this. I didn't bother arguing with him about it, but everyone on this forum would have their boards in RMA half the time if that were true

Asce · 08-17-07

Nice job i may lap my TT120, if i knew what i needed

TriBeCa · 08-22-07

just a piece of glass and at least one course grain sandpaper and several fine grained sandpapers. I'd highly recommend getting a course lapping kit (without glass) and a premium lapping kit (with glass) from www.easypckits.com. They're inexpensive and will arrive promptly.

TriBeCa · 08-25-07

ASUS says my board is on the way back...so in a week or so I may have new OCing results to report

kevg73 · 08-27-07

what did you do differently the second time you lapped the heatsink to get better results

TriBeCa · 08-28-07

I was more careful about not letting it stutter, and I used a slower, more deliberate motion on the paper. Also, the second time I changed my grip, though I'm not sure that really made a difference. Instead of using a two-handed grip I used one hand to hold the glass down and I hooked my middle and ring fingers on my other hand under the bottom fins between the heatpipes (and my index and pinky fingers went on the outside of the heatpipes). Without gloves on you'd probably want three fingers inside the heatpipes. Anyway, I suspect this grip made it easier not to apply downward pressure (which is bound to be not totally straight down), and I think I even lifted it off the sandpaper very very slightly, so that it just glided around more and stuttered less. I think the first time I got impatient (because it was taking so long), and I was inadvertantly rolling the heatsink very very very slightly while making the circular motion. This lead to it developing a slight convexity which, of course, ruined the result. Basically...just be really patient and deliberate with it.

v!p3r^ · 08-28-07

i thought u only need the premium kit, dont think you need two kits

		Overclock.net - Overclocking.net > Cooling > Air Cooling
[Guide] Lapping your HSF+IHS (Pictorial).

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)