About lapping...

[CougTek]

Does it really makes a difference if you use 800 grit paper after you're finished polishing off with the 600 grit paper? I mean, I feel the 800 grit is more for aesthetics than for actually cooling benefits.

Not that the freak wet-sanding his north bridge gave me ideas, but I wanted to prolong the lifespan of a used Thermaltake Volcano 7 heatsink with a quite rough bottom, so I decided to lap it just to see if it would enhance substantially the thermal dissipation level of the heatsink. Using 600 grit paper, the base is fairly flat and shiny, although I can't use it has a mirror. I have 800 grit paper too, but I wonder if it really worthes my time.

 

[ddrueding]

IMO, 600 will give you a better surface than that of the chip anyway. clearly this is "good enough" and nothing thermal paste won't be able to handle.

If you're looking for good operation with a dry-fit...get them both to 1600+ (even at this, it will only be touching at the ridges, but there should be enough of them to do the job).

 

[jtr1962]

I doubt it would make that much difference-maybe a degree or two. This isn't that important for a microprocessor. However, in the case of thermoelectrics where 2 degrees cooler on the hot side means about 1.4 degrees cooler on the cold side, it is worth it (at least to me). I'll go with 600, 800, then 1200. Finally, I polish it to a mirror finish with Noxon. Not just for aesthetics, but if my reflection is crooked I know the surface needs to be touched up a bit. When my reflection is close to perfect (not implying that I look perfect, but rather that the reflection looks the same as one in a mirror ) then I know I'm there (as far as flatness is concerned). Actually, to check flatness something with a grid like graph paper would work better, but unlike my reflection I would only be able to see it at an angle instead of head on.

Another thing-regardless of the apparent perfection of the mating surfaces you still need thermal paste. Air is such a poor conductor of heat (about 30 times worse than standard thermal grease) that even a 1/10000 of an inch gap will cause a significant thermal gradient, and the best mating one can hope to achieve simply sanding and checking is maybe 1/1000 of an inch-ten times worse. I know I can't do any better, and I've been at it quite a while.

 

[Jan Kivar]

I did lap my Celeron's HS (Intel retail one) with 1200 grit. Basically because that was the first piece I found (OK, the first one was 400...). Mostly I sanded off the remainders of Intel's thermal goop/patch, but I didn't get mirror finish. IIRC You'd have to use some kind of polishing detergent, which can take out the "performance gain" from lapping.

BTW, I didn't get any gain from lapping (plus applying AS3); maybe because the Celeron is producing less than 20 W, and it has a heat spreader. Or, the drop was less than ~1°C, which isn't noticeable with mobo sensors...

AFAIK it's best not to sand Intel's heatspreaders; they are slightly convex to have the best contact when the HS is wedged in place.

Cheers,

Jan

 

[CougTek]

Thanks for the replies. Maybe I'll give a 5 to 10 minutes try to 800 grit and then consider it done.

 

[blakerwry]

i've seen some tests performed by people like us that sowed with a thicker thermal paste a 600 grit sand was better than 800. With thinner paste, 800 was better. I have read not to use a polish like buffing/polishing cream.

 

[Buck]

I have read not to use a polish like buffing/polishing cream.

They should be fine as long as you can clean the surface of their residue.

 

[blakerwry]

I'ce seen it in a couple places... the reasoning was that the minute metal particles created by the buffing along with the abrasive used in the buffing cream would basically "clog" the metal and were hard if not impossible to remove unless sanded down.

So I have just stuck with wet sanding on a flat surface.

 

[CougTek]

I guess that Arctic Silver Céramique is a thick thermal paste, isn't it?

I have a big tube of it, enough for my all my personal computers for the next five years probably, so I don't mind using some even for processors that might not need such a high-quality paste.

I also hooked an air duck to the slapped Volcano 7 and I cleaned all the dust that had accumulated in the heatsink and fan. I'm pretty sure the resulting work piece is a decent heatsink. Not something to break overclocking records, but ok for any socket-A processor I guess. The Volcano 7 is rated up to 2800+, but with the lapping, AS Céramique paste and the air duck, I'm sure it would be fine even with a 3200+. Not that I plan to waste that much money on the CPU, but...

 

[Buck]

I find that to be a bit picky. What about the minute particles that 1600 grit sand paper create, and the mineral deposits left by the water that is used? What about the air impurities that settle on the heatsink, CPU, and on the thermal paste during assembly? What about the uneveness of silver in your thermal paste - some areas will have more then others? You are talking about microscopic residue or particles that are of no consequence in this sort of application.

 

[Handruin]

Screw lapping:

Surface grind it. :mrgrn:

 

[blakerwry]

I find that to be a bit picky. What about the minute particles that 1600 grit sand paper create, and the mineral deposits left by the water that is used? What about the air impurities that settle on the heatsink, CPU, and on the thermal paste during assembly? What about the uneveness of silver in your thermal paste - some areas will have more then others? You are talking about microscopic residue or particles that are of no consequence in this sort of application.

What, you don't use distilled water in a clean room environment and a homogenious mix of thermal paste when preparing your HSF for installation?


You're probably right buck, it would make little difference if any.

 

[Buck]

What, you don't use distilled water in a clean room environment and a homogenious mix of thermal paste when preparing your HSF for installation?

:wink: