Lapping the Heatkiller 3.0 [Top tip: DON'T DO IT!]

[HESmelaugh]

It came up in a thread somewhere that the Watercool HK 3.0 has an un-lapped, unpolished base while several other blocks like the Apogee GTZ or the Koolance CPU 350 have polished bases.

This, of course, begs the question whether a lapped and polished surface will provide improved temperatures.


Only one way to find out...





Before lapping

Here are some pics of what the HK 3.0's base looked like before lapping:



There are some scratches on the base because it's already seen almost a dozen mounts. Also, sorry for the dust on the base on that second pic.
As we can see on the third pic, the base does have something of a mirror-shine to it.

Anyway, let's get the sandpaper out.


During lapping

After a few minutes of lapping I had a first look at the base and was a bit surprised. See for yourself:



EDIT: The pic is actually a bit misleading. The darker spots is where the block made contact with the sandpaper, the brighter areas are the concave ones.

We are clearly looking at a bowed base here. The area in the middle is making contact with the sandpaper, as are the edges, but there is a "ring" around the middle that isn't toughting sand yet.

Nickel020 checked the base of his HK 3.0 and reported that it was completely flat, as far as I recall. So what's with the base of mine? This is a retail-sample, by the way, so no need to come up with conspiracy theories.

I don't know if this has to do with the fact that this is a used block or if it's coincidence or whatever. Seeing that the block is bowed and that I am basically flattening out that bow makes me guess that I won't be doing this block's performance any good. But I started it, so I might as well go through with it.


After lapping

Here are some pics of what the base looked like after lapping and polishing:



Not perfect, but I guess it can pass.


Temperatures before and after

The block is currently on it's second mount after lapping.
I'll post the results when they come in.


Okay, results:
All temps are the difference between CPU and water temperature.


Average temperature of all mounts before lapping: 30.4

Temperature after lapping, Run 1: 31.7
Temperature after lapping, Run 2: 31.7
Temperature after lapping, Run 3: Well, what's the point, really?


Conclusion

As expected, the performance suffered from lapping down the bow. Though I never expected it to suffer this much!

What did we learn?

- The HK 3.0 has a slightly bowed base.
- It's nice to see your reflection in a block's base but not necessarily a good thing.
- Shane is willing to wreck his stuff in the name of science...

Cheers,
Shane

[MF Traum]

Seeing that the block is bowed and that I am basically flattening out that bow makes me guess that I won't be doing this block's performance any good. But I started it, so I might as well go through with it.

But it's bowed wrong way isn't it? That's kind a weird to have it bowed inward.

edit. Anyway I'm eagerly waiting for your (and others) results as I'm coming back to watercooling soon.

[G-UNIT91]

Imho...-1°c?:d

[faster3200]

But it's bowed wrong way isn't it? That's kind a weird to have it bowed inward.

I don't think it is bowed the wrong way.

[Sparda]

Indeed. It look like concave to me ...unless my eyes are deceiving me.

[MpG]

But it's bowed wrong way isn't it? That's kind a weird to have it bowed inward.

No, the center portion is bowed outwards (towards CPU), so it's making full contact with the sandpaper. The scratches around the outside are where the sandpaper wasn't really making anything but accidental contact during the lapping.

[HESmelaugh]

No, it was bowed the right way (convex). I had a look at the picture and see what the problem is: The light is reflecting off the parts that made less contact with the sandpaper and so they appear bright. The darker spots are actually more finely sanded because they made good contact with the sandpaper, but because of the angle, the light isn't reflecting off of them.

EDIT: What he (poster above me) said.

[nikhsub1]

I suspect performance to be worse now... It will really depend on how bowed it was before you lapped... always best to lap with the base removed from the block if you want to try to keep a bow.

[Hondacity]

now its bowed...hmmm what does the manufacturer say about it?

[Nickel020]

My guess is that you'll see no performance improvement or even a little worse performance, since a little bit of convexity puts more pressure on the die itself which may be a good thing, even on a lapped CPU. Differences will be marginal though, less than 0.5C for sure, more like 0.2C so you can't even be sure whether it's not a measuring error.

Nickel020 checked the base of his HK 3.0 and reported that it was completely flat, as far as I recall.

I can't say it's completely flat but it is the flattest i have seen. And I do ahve quite a few blocks here: Alphacool Nexxxos Xp + Nexxxos X2, EK Supreme + Ek Supreme LT (and an old Wave^^), the Watercool Heatkiller 3.0 CU, D-Tek Fuzion V1 and V2 as well as a Swiftech GTX. I don't want to go into the air coolers as well...

I'll try to visit my dad and borrow his camera, my camera is broken and the camera in my mobile phone isn't good enough.

PS: If you wonder why I have so many coolers, it's simple: I plan to review them all But I need all the testing equipment (bench table, digital temp sensors, rotameter...) first before I will publish results. And I need to finish building my website... I don't want publish any results or even do much testing with analog temp sensors only, because I'd like to attain a high standard and provide accurate results.

[Sparda]

I can't say it's completely flat but it is the flattest i have seen.

Did you check it by plain eyes or by razor blade or something like that ?
iirc I can see the bow using razor blade on Fuzion V1 when I bow them with o-ring.

[MF Traum]

Ah, okay, I thought wrong. It looked like the middlepart wasn't touching sandpaper and I clearly read HESmelaugh's comments wrong.

Wouldn't it be easy to check flatness with a piece of glass and a drop of water? I mean if razorbladetest is difficult with such a large base.

[NaeKuh]

OMG HES, how can you lap a bowed block.

You did the biggest goof any noob can do.

Razor test everything b4 you lap.

Bleh its hard to get rid of a bow, takes MANY MANY HOURS of lapping, and you need to keep the entire center even.

[HESmelaugh]

I can't say it's completely flat but it is the flattest i have seen.

Thanks for clarifying.

Tell you what,

Unless the block is really bowed, you can't tell by laying a str8 edge on it. The bow in the OP's picc is only showing up because of the lapping. A lot of people are reporting flat blocks, when in fact alot of them have slight bows just like this one. The base might be flat when they are manufactured, and it may be intended not to have a bow, but most always distort their shape when attached to the blocks top plate. Copper is a very flexible metal. It distorts it's shape very easily when pressure is applied form any one direction.

the original D-Tek Fuzion v1's bow was actually more of a mistake found after assembling the block IIRC the story. That mistake caused the base to bow out, or convex, which applied more mechanical pressure to the IHS thus improving thermal performance. The post describing the anonmaly is still on the forum somewhere dating from around Feb 2007.

People like nikhsub1 expanded upon that, and found increasing the bow even more improved the thermal performance. Thus, a bowed or stepped base is better. But evidently, that's changing somewhat with the newer blocks. At least it would appear so.

andyc

Interesting, I didn't know any of that.
I didn't disassemble the block for lapping because the base is quite thin and it would have been very difficult to hold and move accross the sandpaper.

OMG HES, how can you lap a bowed block.

You did the biggest goof any noob can do.

Razor test everything b4 you lap.

Bleh its hard to get rid of a bow, takes MANY MANY HOURS of lapping, and you need to keep the entire center even.

You seem to have a tendency to overreact, good sir.

The base was only slightly bowed and it didn't take very long to lap it down. I realize that this will probably worsen performance, but I started this to see if there would be a difference and what would the point be if I just pussied out as soon as I saw the block was bowed?

I'll survive having lapped a block without benefit to temps.

[Waterlogged]

It's been known for a while that anything beyond 800 grit does not improve transfer. Polished bases are purely cosmetic (and worthless IMO).

OMG HES, how can you lap a bowed block.

You did the biggest goof any noob can do.

Razor test everything b4 you lap.

Bleh its hard to get rid of a bow, takes MANY MANY HOURS of lapping, and you need to keep the entire center even.

Give me a mill and a fly cutter and I'll have it gone in a few mins.

[nebuchanezzar]

Thanks HESmelaugh, I at least was curious about the effect. That base looked pretty flat to me when I put in on yesterday. Like I mentioned I just used the slide on my caliper and id did not rock on the base. Not a razor blade flat I'll admit but flat enough imo. Interested in what, if any effect, that is actually going to have.
Hopefully it's still flat. I never did any lapping myself because I always felt that I would "lopside" it if you know what I mean.
Thanks for taking the leap on an awesome block, imho. I know I wasn't the only one curious, or who perceived a flat base on it either.

[HESmelaugh]

First post updated...

[Sparda]

Oh that sux Shane. What will you do to the block ? Sell them at luxx or is it possible to bow them back ?

[systemviper]

thanks for the effort, now I won't lap my 2 LT's...

thanks

[Sadasius]

First post updated...

Took one for the team .....Now we know....ummmm was going to do the whole GiJoe thing but never mind!

[NaeKuh]

Meh... you need to LAP it til its all FLAT....

unless u have a mill good luck.

This is why i said many many hours of lapping..

[HESmelaugh]

Oh that sux Shane. What will you do to the block ? Sell them at luxx or is it possible to bow them back ?

Yeah, I'm gonna sell it somewhere: "Special edition HK 3.0 with super mirror-polished base! Only 10% more expensive than the standard HK 3.0!"

Hope the buyer never finds this thread...

No, seriously: I don't know what I'll do next. Maybe try andyc's suggestion.

Dude, I commend your effort non the less. Now that you've gone this far. Try removing the base, lapping again, reassemble and see if the bow performance comes back. Would be interesting to see if the distortion comes back after removing the base.

Or, you could just forget I ever mentioned this if it's to painful.

andyc

I might give this a shot next time I'm bored. Thanks for the suggestion.

[tool_462]

What grit was used for the final finish?

I've found that about 800 grit performs the best, I have gone up to a perfectly polished 2000 grit surface and lost a couple degrees of gain.

I have a thread somewhere showing E8500 IHS lapped with different grits. The real key is FLAT not polished

[zlojack]

Is the CPU lapped?

I bet you get better performance again if you lap the CPU.

[nikhsub1]

Just as I thought. Try to relap unassembled... it may fix it.