Useful Info on lapping...

[NeoMoses]

I had some questions about conflicting information I had seen around about base finish on heatsinks and waterblocks. I know that theory states that the smoother an interface is, the lower thermal resistance it will have.

I had heard some possible concerns with ultra smooth finishes causing degraded performance over time due to the thermal grease actually 'pumping out.' Although I've never seen this effect personally, I figured it was worth my time to ask the 'pros,' so I sent an email off to Arctic Silver's tech suport. Here's what they had to say.

Keep in mind that this may not be true with all thermal compounds, but since I use ASIII, I asked them.

email quoted with permission.
Bryan,

There is no such thing as a surface that is to smooth. The finer
the finish, the less thermal grease will be needed. With a highly
polished and flat surface, a proper application of thermal grease will appear to be missing when the waterblock is removed from the CPU.

One caution on lapping, it's hard for an end user to keep the surface flat. Lapping a water block with a super flat base like a Swiftech would be counterproductive.

--
Colin Thompson
Arctic Silver, Inc.
Phone: (559)740-0912
Fax: (559)740-0913

--- Original Message ---
Hi,
I am wanting to lap my waterblock for optimum performance for the next 6 months. I have heard conflicting information regarding the optimum sandpaper grit to stop at. I have seen some people
who stop after 400 grit... others go to 2000 grit, a near mirror finish.

The conflicting information I've heard is that if you lap too much, and have too smooth of a surface, you may get a 'pump out' effect of the thermal paste. I would like to get the opinion of
a manufacturer, as I would trust your opinion.

Assuming the base if flat, what is the optimum roughness (rz, ra, or sandpaper grit) for arctic silver 3 and arctic silver 5?

Thanks for your time,
Bryan

One very important thing to note, a smooth base may not be the most flat base. So, if you are lapping, make sure you're doing it on a flat surface. I use a toolroom grade B granite surface plate. It's guaranteed flat to within 0.0001".

Just thought this info might be useful to y'all. Enjoy!

 

[archilochus]

Thanks for the INFO!!!! THE FANMAN

 

[stan03]

thats nice to know.... if only thermalright could make bases like swifty.

 

[dalilman52]

yea...i saw an slk-900a's base once, the base looked like it was dragged down a concrete sidewalk...
________
Lenders mortgage insurance forums

 

[I.M.O.G.]

Thanks Neo... Another myth put to rest.

 

[Shin Akuma]

OK here's my take on this issue... I may be wrong though so don't flame me

Picture shows smooth base and coarse base. Green is heatsink. Blue is TIM. Red is processor die.

With a very smooth base there's a possibility that a thin layer of TIM will "lift" the heatsink preventing direct contact.

With a coarse base the hills will inevitably contact the die.

The problem now is pump-out. As the assembly expands the heatsink will push harder on the die. With the smooth base the tendency is to pump out some TIM as the heatsink moves closer to the die. But with the coarse base pump-out won't happen since there is direct contact and the heatsink can't get any closer to the die anymore.

 

[I.M.O.G.]

Sorry Shin Akuma, after lapping that still isn't how the base looks... Lapping cannot remove the microscopic peaks and valleys (observable under several 1000x magnification) that TIM is intended to fill.

You could never get a base as smooth as you have pictured, and even if you did, there would not be a full layer of TIM between the heatsink and CPU - only the very small regions that are raised would be filled with TIM, the rest would be in direct contact.

Minimizing those peaks and valleys, and allowing maximal direct heatsink and core contact is preferable.

Look here for more info:

http://www.arcticsilver.com/thermal_interface_basics.htm

Note that the link also mentions something about bases that are not flat - concave or convex bases resulting from rotational lapping techniques can result in contact like the last picture in that link.

 

[pvc_and_frost]

i think optimal thermal grease is applied when you wipe it off once or twice so that only the ridges are filled...if you know what i mean...if you dont, ill explain

 

[9mmCensor]

Yeah plus lapping doesn't FILL IN the valleys.

 

[I.M.O.G.]

I don't think its any more complicated than the directions on AS's site explain... A lot of people use that method with their whitewater blocks and they get some of the best temperatures anywhere.