SPCR

I've got my computer tested and running, now that I know it works, I'd like to put the finishing touches on it. I was wondering how much of a difference lapping the heatsink might make? I have the time and am willing to put the effort in, so I would like to know whether it's worth it. I'm using an xp-90 heatsink and was also considering doing it for my thermaltake cl-g0003 heatsink as well ...

Thermaltake heatsinks can definately benefit from lapping, I dunno about the other brands. Some Zalmans also need lapping. It depends on your skill and how well you can lap it. Usually lapping it still won't give it a flat finish and will be slightly curved. Good lapping requires a good equipment setup.

I lapped my Scythe Ninja a few hours ago. The CPU is performing about 5c cooler under load now.

Results vary, I've found even if the heatsinks are the same model. If there is no difference after lapping, then it was already pretty flat to begin with.

josephclemente wrote:I lapped my Scythe Ninja a few hours ago. The CPU is performing about 5c cooler under load now.

Results vary, I've found even if the heatsinks are the same model. If there is no difference after lapping, then it was already pretty flat to begin with.

How well did you think you lapped it? Did you notice whether if the bottom of the heatsink still had a slight curve?

What was the condition of the ninja before lapping? Was it warped as determined with a straight edge or just did not have a smooth finish? - FG

The Ninja had a very nice smooth finish from the factory.

I noticed the warp based on the heatsink compound pattern, after I later removed the heatsink. The compound did not transfer from the CPU to the heatsink base as well as expected. When I was lapping with 3M sandpaper over glass, I could see that the base had very obvious low points in the area where contact did not occur.

I think I did a good job. I can't claim it is now perfectly flat since it was performed by hand, but it is much improved.

Like josephclemente I gained about 5c & for the nominal time it took, it was well worthwhile, but I had a TT Volcano heatsink, which was an initial disappointment.

On a sheet of 1/4" glass it quickly went down to 1200 grade wet & dry, although I didn't bother with any water till the finest grade.

As you seem to be happy to put the time in you will probably end up in front. Good luck!

Interesting that even a well made cooler like the Ninja may benefit from lapping. Another thing new that I haven't thought about earlier was high and low spots on the heat sink face... I had always thought that lapping was mainly to improve the finish if it looked on the rough side, so this is good to know that smoothness and a mirror finish isn't the whole story.


Question: So 1/4" glass is sufficient thickness for lapping? I had read about using glass blocks as a flat surface in the past, but didn't read anything about how thick you needed it...

Any other good surface or material suggestions? Probably not, but I just thought I'd ask.

lapping using sandpaper back and forth still produces a slight curve, but i guess it is good enough!

Going back & forth like cotdt mentioned IMHO could produce a bit of uneveness @ the edges, owing to the initial drag @ the beginning of the stoke? It would be preferable to use a circular motion.

Also the glass is not a block but a sheet, approx 12" square, that I used to use for printing photographic negatives, (in the late 20th Century). It is/was known as "1/4" float". Most decent glass merchants should be able to supply something similar, maybe even an offcut, if you're lucky.

The way I used it was to lay the glass on the worktop, tape a whole sheet of wet & dry to the glass, then just gently & smoothly circle the face of the HS round this surface, carefully holding it by the fins.

Every few motions I would alter my grip on the HS by 90 degrees, so as not to introduce any pattern. I forgot to mention I also used a shim for the HS mounting, which I lapped as above.

BrianE asked about the glass thickness & all that I can say is that regular thin window glass might flex or crack unless the work surface was flat, although there would be ways round this.

Old, or greenhouse glass can be quite "ripley" when held up. Most modern glass is fairly distortion free and consequently adequately "flat" for our purposes. A quick look will confirm this

Hmm ... I'm starting to think that lapping the xp-90 might not be the greatest idea, simply because of the way the weight is distributed on it. I could see myself lapping it and ending up with a 'slant' on the sink in the end ...

I didn't find any needs for lapping my Ninja. I've been running prime95 for a day now, load temps are stuck at 42C with my 3000+ @2250MHz. This is with only two nexuses at low rpm in my P180.

Krispy wrote: Most modern glass is fairly distortion free and consequently adequately "flat" for our purposes.

I've used mirror surface. It has to be "distortion free".
I've lapped 3 heatsinks using ordinary round cosmetic mirror (20cm diameter) and sand paper (of course). It works ok.

BTW. I didn't scratch the mirror.

how much would a machine shop charge to perfectly cut away layer I wonder?

the next time i do a performance system I think i will try asking one to do so.

my current Reserator 1 heatsink is coated in gold, so really it would be bad to lap it. It is mirror polished as well. That doesnt mean though that it is perfectly flat, something that is even more important than smooth.

Good tip about your scythe, makes me think about other systems.

Also, I have seen people try a couple of times with just reapplying thermal grease and they get different temperatures.

Link to an old article about lapping a cpu. The author recongizes that moving a cpu or heatsink in a figure 8 motion on a piece of sandpaper is going to result in a convex surface. A jig is used to apply even pressure across the surface to be leveled. - FG

http://www.benchtest.com/lap.html

frankgehry wrote:Link to an old article about lapping a cpu. The author recongizes that moving a cpu or heatsink in a figure 8 motion on a piece of sandpaper is going to result in a convex surface. A jig is used to apply even pressure across the surface to be leveled. - FG

http://www.benchtest.com/lap.html

A good article, but it sounds like the jig is really only useful for his slotted CPU? I think the issue he was trying to overcome was the fact that you had to hold on to the flexing circuit board, which could create an uneven pattern.

None of my Thermaltake CPU coolers were flat, and I've owned a dozen different thermaltake coolers. After even an ameteurish grinding, I got 5-10C improvement in temps. And thing is, they were all still slightly convex because I am a horrible lapper. But maybe there is a law of diminishing returns for flatness, I don't know because I've never owned any perfectly flat CPU coolers. Some more expensive waterblocks were mirror polish (gold plated) but I don't know if they are really flat. I've also read that mirror polish is not as good as a very slightly rough satin polish, assuming both are flat. So it seems like a slightly rough 1500-2000 grit surface is the optimal. Also, I'm pretty sure gold plating and nickel plating isn't good for performance, so I don't know why some of the waterblocks I've bought were plated with very strange metals, even nonmetals like teflon. Could it be that companies want you to THINK things are flat when they are not?

i would say yes. the plating makes it seem flat.

however.... i have problems with plating and heat conductivity.

when zalman makes their big circle flower 7000 guy, it uses super tightened bolts to compress the metals together. soldering makes for a insulating layer between metals, and is terrible.

i wonder about gold plating and its process. does it oxidize or make strange layers between the gold and the aluminum or copper?

I know the idea should work great but chemistry never works great in real life situations.