Sanding Megahalems, order another one?

[aamsel]

Have a Megahalems that has been clamped-down
on more than 5 CPUs. I know that the surface is supposed to
taper up toward the edges so that the clamp-down will flatten
it out, but mine seemed to have roughness at one or two edges.

I decided to take some 600+ wet/dry and smooth it out (not using
a sanding block) with very light sanding.

Now it looks ok, but the copper is starting to peer through the stainless
finish of the contact surface.

Is it fine?

Should I

a.) leave it alone?
b.) sand or polish additionally?
c.) order another heatsink?

Please advise.

 

[aigomorla]

its fine...

Did u make sure it was lapped flat tho?

U dont need to buy a new sink.

worst comes to worse just lap it down.

Its noted that u can typically end lapping a sink using 700 Grit SP.
However i would end it using at least 1000 grit SP.

That way u get a mirror finish.

Go in 1 direction only, and use a flat surface to lap.

But there is nothing wrong if ur seeing the copper.

I used to lap TRUE's all the time because they were horribly uneven:

NEVER use polish.
Polish applies coat.
The coat interferes with thermal transfer.

Just end it with sand paper, use rubbing alcohol to clean up abit, and remount.
Then win!

 

[aamsel]

its fine...

Did u make sure it was lapped flat tho?

U dont need to buy a new sink.

worst comes to worse just lap it down.

Its noted that u can typically end lapping a sink using 700 Grit SP.
However i would end it using at least 1000 grit SP.

That way u get a mirror finish.

Go in 1 direction only, and use a flat surface to lap.

But there is nothing wrong if ur seeing the copper.

I used to lap TRUE's all the time because they were horribly uneven:

NEVER use polish.
Polish applies coat.
The coat interferes with thermal transfer.

Just end it with sand paper, use rubbing alcohol to clean up abit, and remount.
Then win!

OK, I will lap, but...
This Megahalems (and all of them, I thought) is not flat.
It is nowhere near flat.
My understanding is that they are not supposed to be flat.
If you are talking about a sanding block and making this thing
flat, it is going to take forever.
Otherwise, is there another method, and can I get the finest paper
I would need at a Home Depot, or do I need a specialty place?

Thanks,
Andrew

 

[Jimmah]

its fine...

Did u make sure it was lapped flat tho?

U dont need to buy a new sink.

worst comes to worse just lap it down.

Its noted that u can typically end lapping a sink using 700 Grit SP.
However i would end it using at least 1000 grit SP.

That way u get a mirror finish.

Go in 1 direction only, and use a flat surface to lap.

But there is nothing wrong if ur seeing the copper.

I used to lap TRUE's all the time because they were horribly uneven:

NEVER use polish.
Polish applies coat.
The coat interferes with thermal transfer.

Just end it with sand paper, use rubbing alcohol to clean up abit, and remount.
Then win!

Just like to add, the use of lapping polish you get at a machine supply shop IS allowable, as it is only grit and the only way to achieve a 'real' lapped surface (by real I mean a 0-1 surface grade and the ability ti wring two surfaces together).

Better yet, finish the 700 grit sanding off using oil or water, will give a better finish than 1000 grit.

 

[aamsel]

Thanks.
How do I sand with the sandpaper on a sanding block,
and keep the original curvature of the heatsink?

 

[Rubycon]

Do NOT lap, grind or otherwise change the profile of these heatsinks! These are NOT flat for a reason per engineers. Lapping these is ruining them! An aggressive scrubbing pad will refresh the surface. Sanding may be possible with a sanding sponge but do NOT take off too much material or otherwise change its profile.

Much testing goes into this design and thus is the reason why the surface is not flat. Believe me if they wanted them flat they would be flat!

 

[aamsel]

Do NOT lap, grind or otherwise change the profile of these heatsinks! These are NOT flat for a reason per engineers. Lapping these is ruining them! An aggressive scrubbing pad will refresh the surface. Sanding may be possible with a sanding sponge but do NOT take off too much material or otherwise change its profile.

Much testing goes into this design and thus is the reason why the surface is not flat. Believe me if they wanted them flat they would be flat!

That's what I thought,
I was forced to very finely sand the bottom surface and edges
or toss it right now. There was some roughness around the edges that
might have kept decent contact from happening, regardless of TIM used.

One installation too many, apparently!!

Please tell me if a new one right now is the better option, that
is no problem.

To this point, I have just lightly sanded and the copper is just
barely peeking through. Again, if that is ruined, just let me know.

 

[Rubycon]

Best bet is to refresh the surfaces and mount it and check temperatures. They should be about the same as before. The only difference is if you used TIM such as Coollaboratory Liquid Metal. That stuff loves copper.

 

[aamsel]

Will do.
Thanks!

 

[Rubycon]

"Ruin" may be a little strong here...

However if lapped don't expect to see as good of a result as the original. Now if you happen to have a CPU with a slightly different IHS it is indeed possible that temps could actually improve.

I've had several TRUE, a Venomous-X, and a Megahalem professionally faced (milled) down to a flatness range of 1/10000" verified with a CMM inspection machine. On my chips all were no better - often a degree or two WORSE than stock (un milled) products! The Venomous-X was about four degrees hotter!

Speaking of the VX I wish they kept the mounting of the TRUE-C which is positive locking much like the Megahalems. The adjusting bracket with the VX is nice (adds pressure) but the heatsink is not positively locked down (easily twisted) and this does not go over well with many TIMs - especially Coollaboratory and the reflowable types. Once those are broken they need to be re-done.

 

[aamsel]

Best bet is to refresh the surfaces and mount it and check temperatures. They should be about the same as before. The only difference is if you used TIM such as Coollaboratory Liquid Metal. That stuff loves copper.

Followed your advice, refreshed, looks virtually as new now.
Very glad I did not lap, same temps as before.
Thanks!

...Speaking of the VX I wish they kept the mounting of the TRUE-C which is positive locking much like the Megahalems. The adjusting bracket with the VX is nice (adds pressure) but the heatsink is not positively locked down (easily twisted) and this does not go over well with many TIMs - especially Coollaboratory and the reflowable types. Once those are broken they need to be re-done.

So, is the TRUE-C still your favorite?
If I hadn't had the twisting issues with the Ultra's I
would agree. With the Megahalems mounting system
seen as being excellent, what are the seen negatives
of that cooler?

 

[Rubycon]

I'd have to say Megahalems Shadow is my fav air cooler. TRUE rev. C comes close and the Venomous-X with the positive mount the TRUE rev. C uses would probably be ideal. The cooler must not be allowed to move under any circumstances!

 

[faxon]

"Ruin" may be a little strong here...

However if lapped don't expect to see as good of a result as the original. Now if you happen to have a CPU with a slightly different IHS it is indeed possible that temps could actually improve.

I've had several TRUE, a Venomous-X, and a Megahalem professionally faced (milled) down to a flatness range of 1/10000" verified with a CMM inspection machine. On my chips all were no better - often a degree or two WORSE than stock (un milled) products! The Venomous-X was about four degrees hotter!

Speaking of the VX I wish they kept the mounting of the TRUE-C which is positive locking much like the Megahalems. The adjusting bracket with the VX is nice (adds pressure) but the heatsink is not positively locked down (easily twisted) and this does not go over well with many TIMs - especially Coollaboratory and the reflowable types. Once those are broken they need to be re-done.

after lapping i saw a 2 degree improvement in cooling performance on my megahalems. the problem was my CPU and heatsink were both rounded, facing toward each other, so it was only making contact in the center. i didnt want to lap the CPU, so i lapped the cooler instead. ended up just fine, its not like you're going to see a huge difference lapping the megahalems one way or another, its already pretty smooth as is. the stock surface on the TRUE is a different matter, that thing is rough enough to use as a grinding block softer metals LOL. im glad the ven X comes with a properly milled surface now

 

[aamsel]

I'd have to say Megahalems Shadow is my fav air cooler.

Anyplace that these can be purchased online right now that you know of?

Andrew

 

[Rubycon]

Anyplace that these can be purchased online right now that you know of?

Andrew

Only one place in UK.

Standard B Megahalems is much more common. Cooling performance is the same for both.

 

[aigomorla]

These are NOT flat for a reason per engineers. Lapping these is ruining them!

umm that is also what Thermalright said, and what happened after you lapped them?

OK, what ruby is talking about is a BOW.
The guys in air, learned from us in water in what a BOW does.

My friend scott over AT XS did a very nice intro on what a BOW is:

So I thought maybe water was not flowing all through the nozzle, but perhaps out between the top and mid plate and bypassing the base altogether. I know, sounds crazy so in an attempt to PREVENT this, I took an o-ring from an EK barb and placed it between the top and middle plate like so: (sorry for the crap focus, I need to pwn my camera)

Well doing this had unexpected results! The base was not bowed like the apogee base, it is more bowed from the center, not the outside. You see, the o-ring forces the mid plate center to push the center of the base out. The cool thing is that now I'm experimenting with different size o-rings. For those of you that will rush out and try this, the EK o-ring is as BIG as I would go, a tiny bit smaller in thickness would be better IMO. The EK oring makes the base a bit 'dome-ish'. Does it perform better than stock? Yes it does, not as dramatically as the Apogee, but it does. It is a bit trickier to mount too as it has more of a point than a bowed apogee, at least with the EK barb o-ring.

Perhaps these photos will show what I mean, again sorry for the blur.

Here we have a FLAT base just removed (MP-05 LE):

Even with the blur you will note how the paste is pretty even (don't mind there corners where there is no paste).

Next up is the bowed Fuzion look at the paste, note the center where there is very LIGHT paste, this is where most contact is being made (most pressure I should say)

Next up are 3 bowed Apogee mounts, note the STRIP you see, this is the point of most pressure:

A very important aspect of using a bowed based block are the mounts. YOu will have a larger deviation from mount to mount as you are not mating 2 flat surfaces... the Apogee's bow is a bit more forgiving in this regard - at least so far as I haven't had much time to play with different thicknesses of o-rings and the fuzion yet. Bottom line is if you are after best possible temps - use a bowed base and mount several times so you can get a baseline of where your mounts are, you will get a feel for good and bad.

Now, your telling me that the mega's have a bow in a way that the mass from the sink wont interfere with it?

Ruby common, dont make me spank you on that statement.

A bow on a waterblock is more controlled because we dont have an addition 500grams on the pivit point.

The mounts on a waterblock is far more superior then any air sink.

Were talking about a waterblock trick they brought on air.

Do you have any links where people lapped there mega and got worse numbers?
Because if your telling me that lapping a mega will ruin it, i want to ask HOW.

A waterblock has a set jet path, with set cooling areas.

A heatpipe doesnt, unless the heat pipe itself is exposed.

A waterblock concentrates its cooling head on 1 spot, vs a heat sink.

You can get a MORE EVEN MOUNTING PRESSURE on a waterblock for a good BOW mounting then you can on an air sink.


Im not telling ya to lap it right now, but i am asking WHY not.
And dont tell me because vendor says its not correct, because Thermalright tried that bullshit on us, and look what happened.

 

[faxon]

aigo is definitely right. the stock mega definitely cooled the bottom 2 cores (physically lower) a bit better than the top 2 before i lapped it. i saw the biggest improvement on the top 2 cores after lapping, though there WAS also an improvement in temps on the bottom 2. on the flip side, im not touching my apogee with anything unless it gets a little tarnished, in which case im gonna just use a bit of coke to take it off since i dont want to screw with the surface. this thing is cooling my CPU wonderfuly evenly all in all. one of my cores is significantly hotter than the rest, but it was on my megahalems as well, so taking that into consideration i would say that aigo's reasoning about the weight of the cooler playing a big role when the mount is bowed at the center is spot on. currently "idling" @ 15% cpu, my tepms are 44c 41c 39c 39c, with milkyway@home work units landing on the first 2 cores when they need to use the CPU. ambient is about 75f

 

[Rubycon]

You are comparing water blocks to a particular CPU cooler.
Please I've had people test hundreds if not thousands of CPU for me on various samplings. Not ALL of them are conclusive however the majority did in fact show NO change or got WORSE! Also these were engineers doing the testing and the "lapping" was accomplished in a high precision machine shop and accurate to 1/10000 inch. NO lapping with sandpaper can do that. It also cost thousands of dollars to test all this (expletive).

In a nutshell I think I will stick with stock surfaces.

Additionally comparing CPU diode readings between cores is also bogus as the numbers themselves are hardly relative. An actual gradient can exist but it's usually amplified and not that high.

If you need lower temperatures concentrate on achieving cooler "where the rubber meets the road" surfaces. Unless something is obviously warped it's a whole lot of work for such little gain. To each is own I suppose. (Some people love Krell while others prefer Mark Levinson.)

 

[faxon]

Lol yea I'm definitely not doing anything near that in my testing. One thing that's for sure though, once you get in the door with WC, the price for maintenance is worth it . I spent like $500 WCing 2 comps cause i got a lot of stuff on clearance or sale. Keep in mind this is also for 24/7 stability beyond 4ghz in DC apps, and with block bracket upgrades i can carry the parts for years, so great bang for the buck. Keeps my stuff quiet as well which I value

 

[SZLiao214]

I have always had more luck lapping the cpu instead of the heat sink.

On my q6600 i saw a 2C drop at idle and a 3-4 drop under load. The difference didn't stop it from being unstable, or allow me to oc more but it was nice to see .

 

[aigomorla]

You are comparing water blocks to a particular CPU cooler.
Please I've had people test hundreds if not thousands of CPU for me on various samplings. Not ALL of them are conclusive however the majority did in fact show NO change or got WORSE! Also these were engineers doing the testing and the "lapping" was accomplished in a high precision machine shop and accurate to 1/10000 inch. NO lapping with sandpaper can do that. It also cost thousands of dollars to test all this (expletive).

mmm ruby..

so lets recap this.

In order to get a better mounting pressure on a bow, like my friend scott said, the pressure needs to be uniform.

Your applying more presure in the middle where the die actually is.
That is a bow.

And yes your right, however the MEGA itself isnt 1/100000 inch flat itself so what does your statement prove?

Lastly ruby, i have yet to have an air sink which i lapped get WORSE after a good lap job.

Ruby u didnt supply any credible source. In your statement.
Do you have Proof in a form of a logged and real test where if u lap a mega it will get worse?

Becuase Faxconn said his got better, as well as others i know who have lapped it.

 

[videogames101]

Do NOT lap, grind or otherwise change the profile of these heatsinks! These are NOT flat for a reason per engineers. Lapping these is ruining them! An aggressive scrubbing pad will refresh the surface. Sanding may be possible with a sanding sponge but do NOT take off too much material or otherwise change its profile.

Much testing goes into this design and thus is the reason why the surface is not flat. Believe me if they wanted them flat they would be flat!

Thats assuming you haven't lapped the CPU as well.

 

[Rubycon]

I'm just sharing my results. These were not published and the person that helped me asked not to reveal his name/institution since they are a defence contractor.

Your mileage will vary!

For all the work that this takes it is not worth it especially if you get no improvement or worse!

In our testing both the CPU IHS as well as the heatsink block was machined to perfect tolerance. If you put a drop of oil on either and pressed them together you could not remove them by pulling! Only sliding and even then it was quite viscous. (they call this the jo block experiment)

There was a slight improvement in difference between cores however the proportions were still there.

Finally one must consider the bond BELOW the IHS. The infamous indium based low temperature "solder" that bonds the die to the internal side of the IHS. It may have imperfections as well.

More babble...

The thermal diodes are NOT precision devices as well and tolerances can actually reflect larger than ACTUAL temperatures.

It's an enthusiast thing. For real performance gains in cooling you must cool down whatever's touching that IHS. Whether you use water, chilled water, refrigeration, expendable cooling (dry ice, cryogenic fluid, etc.) is up to you.

mmm ruby..

so lets recap this.

In order to get a better mounting pressure on a bow, like my friend scott said, the pressure needs to be uniform.

Your applying more presure in the middle where the die actually is.
That is a bow.

And yes your right, however the MEGA itself isnt 1/100000 inch flat itself so what does your statement prove?

Lastly ruby, i have yet to have an air sink which i lapped get WORSE after a good lap job.

Ruby u didnt supply any credible source. In your statement.
Do you have Proof in a form of a logged and real test where if u lap a mega it will get worse?

Becuase Faxconn said his got better, as well as others i know who have lapped it.