Lapping and using thermal paste?

[laezyre]

If you spend the time to lap your cpu and heat sink, are you diminishing your return by use thermal paste? It seems that the paste would work better with rough surfaces or voids like the direct heat pipe contacts.

 

[Fullmetal Chocobo]

Originally posted by: laezyre
If you spend the time to lap your cpu and heat sink, are you diminishing your return by use thermal paste? It seems that the paste would work better with rough surfaces or voids like the direct heat pipe contacts.

I think even with lapping there are tiny valleys that are filled in by thermal paste.

 

[WoodButcher]

You use less tim and have better contact.

 

[Billb2]

You're looking at the lapped surface and seeing it is scratch free. Try that with a magnifying class, or even better, a microscope and you will see it is horribly scratched. The more magnification you use the more scratching will be evident. It's those tiny scratches that can be filled in with TIM after lapping.

Since copper conducts heat about 30 times better than TIM, the less TIM you use (ie. the thinner the TIM layer) the better off you are.
Getting the surfaces flat requires less TIM to fill in the voids caused by surface contour. Then getting the surface shiny (ie scratch free in the microscope sence) results in smaller scratches requiring less TIM to fill them in. You can look at it this way: The deeper the scratch, the thicker the TIM has to be fill it in. This scratch ^ requires less TIM than this scratch /\.

I've had water blocks and CPUs professionally lapped a machine shops where "flatness" is measures in wave lengths of light, ie 300-700 nanometers (1/1000 of an inch = 25,400 nm), and use just a "haze" of TIM.

 

[Foxery]

Originally posted by: Billb2
I've had water blocks and CPUs professionally lapped a machine shops where "flatness" is measures in wave lengths of light, ie 300-700 nanometers (1/1000 of an inch = 25,400 nm), and use just a "haze" of TIM.

You made my inner geek giggle with glee.

I take it your sig rig is not up to date, then!

 

[aigomorla]

Originally posted by: Foxery

Originally posted by: Billb2
I've had water blocks and CPUs professionally lapped a machine shops where "flatness" is measures in wave lengths of light, ie 300-700 nanometers (1/1000 of an inch = 25,400 nm), and use just a "haze" of TIM.

You made my inner geek giggle with glee.

I take it your sig rig is not up to date, then!

Billb dont temp me please.

I already spend too much money on this hobby.

LOL....

 

[laezyre]

Wow, the only time I saw something professionally lapped was at a bachelor party

 

[BonzaiDuck]

It's not only an issue of "flatness," but a matter of surface material. The IHS has a thin nickel-plating; ThermalRight coolers, possibly some others, use nickel-plating to cover the copper. Removing the nickel from the IHS doesn't make a dent in the thickness of the IHS, so . . . no risk there.

I think the improvement you see is 4 to 10C degrees. I don't know how this may be different for different thermal pastes, if you use the same paste before and after lapping. Micronized diamond paste is best, though.

 

[laezyre]

I am planning on lapping my vendetta and cpu's nickle plating this weekend. I read one of the linked articles about lapping showing a graph with a large drop in temps. This occurred more dramatically with the removal of the nickle on the IHS. I will post the results.

 

[Billb2]

Originally posted by: Foxery

You made my inner geek giggle with glee.

I take it your sig rig is not up to date, then!

Ah, yes it is. I've been where you all are trying to get to.
DDR memory at DDR665, Athlon FXs at 1.6v, 30+ different latencies to adjust, Iwaki pumps, Thiaphon burmer, etc.
I find getting a locked multi PIIIi500 to 650 on a 100 buss more challenging now.

I need one of those rare PIII/100/1ghz cpus to have some real fun!
Who wuda thunk, a PIII on phase....

 

[Billb2]

Originally posted by: aigomorla
[
Billb dont temp me please.

I already spend too much money on this hobby.

LOL....

If you can find a small shop that thinks you are a nut case and you get them curious about what you're doing they won't charge an arm and a leg. The last batch of stuff I had lapped (three blocks and two CPUs) was only $125, and they threw in polishing one of the waterblocks for free.

 

[CDC Mail Guy]

Originally posted by: laezyre
Wow, the only time I saw something professionally lapped was at a bachelor party

lol I think I was at that party

 

[Syzygies]

The jury is still out for me on whether I believe this site, but they raise an interesting question:

From Benchmark Reviews Best CPU Cooler Performance - Q1 2008, discussing exposed copper heatpipes:

It's recommended that a carbon and aluminum-oxide based TIM be used with all copper surfaces.

I didn't see an explanation there or in the linked article. When we lap, we're exposing copper, so if this is right I shouldn't use my AS5.

 

[laezyre]

I wonder if they are concerned about oxidizing the copper. Copper oxidizes and turns that ugly dark green color we see on statues. Perhaps the oxidized copper has a poorer heat conduction. I'll have to research this.

 

[Syzygies]

Originally posted by: laezyre
I wonder if they are concerned about oxidizing the copper. Copper oxidizes and turns that ugly dark green color we see on statues. Perhaps the oxidized copper has a poorer heat conduction. I'll have to research this.

I see a revised assertion in

Benchmark Reviews Best Thermal Paste Application Methods

As previously mentioned, it's best to avoid oxidizing "white-grease" type TIM's, since they can tarnish the copper surface and bleed out.

My immediate need is to decide whether or not to go with Artic Silver 5 which I have on hand, for lapping in the next few days.

 

[WoodButcher]

If you look around here you will find that AS-5 is one of the best, most widely available, easy to use, often recommended pastes going. It is a carbon based TIM IIRC. The only reason you will hear me recommend AS Ceranique is because people are slobs and can't manage to keep the goop off their boards. There are some pastes better but again, the question of availabiliy.
Get the AS cleaner / prep solutions, much better stuff than iso alcohol IMO.

 

[Syzygies]

Originally posted by: WoodButcher
Get the AS cleaner / prep solutions, much better stuff than iso alcohol IMO.

Thanks. Yup, I had the entire guts of my rig in my carry-on for my flight to CA (An identical case/supply/burner was waiting here), and for the first time in a while I didn't get a secondary search. Go figure.

The AS cleaner / prep went in a ziplock bag with my toothpaste, for screening. I was reminded of the study (I don't believe) that found AS5 and toothpaste interchangeable as TIM.

 

[WoodButcher]

Originally posted by: Syzygies

The AS cleaner / prep went in a ziplock bag with my toothpaste, for screening. I was reminded of the study (I don't believe) that found AS5 and toothpaste interchangeable as TIM.

Maybe the toothpaste will work in a pinch for tim but grey teeth don't cut it!

 

[Syzygies]

Ok, the guts of my machine flew carry-on NY to CA, to install into an identical case, power supply, burner. I lapped my TRUE and Q6600 with 400 then 600 grit. 400 is a high start but just takes more patience, from my research 600 is a reasonable place to stop. I found the TRUE very difficult to lap, but any edge rounding fell outside the profile of the Q6600. The Q6600 itself was very easy to lap.

I then went nuts trying to actually use AS5, but as little as possible. The stuff is sticky, heating it enough to spread better, without frying the chip, could be a good idea, but I didn't. I ended up spreading the thinnest possible film. The experience was unsettling, I came very close to starting over several times while working through subsequent issues.

Too many changes at once, and I had no idea what was going on. Reverting from 8 GB back to 4 GB of memory, I realized that somehow my Linux partitions were no longer booting, but the computer was otherwise overclocking exactly as it did in NY. Reinstalling Ubuntu, I could finally look at core temps while prime testing full load on all cores at 3.2 Ghz, as before.

=> The hottest core is 5 C cooler than I saw before lapping, controlling for ambient temps.

=> I'm seeing a ten degree C spread between my cores.

(I can hear Homer Simpson talking to me, "You say that like it's a bad thing!")

I'm truly baffled. It's possible that the extremely thin TIM application that so unsettled me is exactly what one should do.

I'll ride out the AS5 break-in period, but my low cores and the spread are inconsistent with anything I've ever read. I don't know what to believe.

Simplest explanation is I nailed three of four cores on the lapping, botched the fourth, everything else is just noise.

 

[ICD7]

Here you go, Intel spec on flatness for you lappers

http://download.intel.com/design/processor/datashts/31685201.pdf

page 32. flat .05mm -parallel .203

Somebody on another forum made the observation that large delta temps between cores could be due to the sink making better contact on one side and less on the other. If it's true or not I do not know, just a thought.

 

[godisknugen]

just a thought, maybe the heatsinks are convex for a reason. for example if the convexity is in line with the CPU core (a dual core for example is not a square) you get better contact were the actual heat is produced. second what about tension, if you put heavy loads on the sides of a piece of metal it will eventually change shape...cant the base of a TRUE flatten out over time? if thats true then lapping it would eventually result in a concave base!!!