Absolute Flatness - Stone Grind vs. Lapping

[bobad]

Joe Camel,

I gave up hand lapping. My fingers are just too soft to press down uniformly and consistently. The more I tire, the sloppier my work gets. Last 3 HS I lapped, I used a lap wheel and a small weight bearing down on the HS. It's much slower than hand lapping, but I walk away and drink coffee while it slowly laps away a few 10th's of a thousandth. (~.0001-.0005). When finished, the HS is nothing if not flat. Since my tests showed it didn't help by any appericiable amount, I quit lapping. Well, I DO check for dings and burrs, and pass the HS on some 2000 3M wet-or-dry paper. I do just 5 or 6 quick wipes so I don't wear the corners down. I rely on AS5 to fill in the remaining gaps.

 

[Joe Camel]

EDIT 6/20: imageshack must have a server down...cant even get to the home page...ill repost if they dont come back "soon" ... back up

bobad,

ive been known to cross over into overkill territory...


well, i didnt go this far and not go all the way

Great pics, combine that with a concave heatspreader and we might have a partial explanation for some less than desirable xp-90 cooling results for some.

this one's for you



somehow, through my "amateurish lapping skills" and "uneven pressure during lapping", ive created a perfect "ring" around the top of my CPU heat spreader.....






and now its gone ... and so is the stepping...huh...bummer....3200 Winnie..?

NOTE: this is the pic that shows what "loose" paper can do to the edges








lets test her out...i hope to have stats by later tonight...here is where im starting from...

NOTE: i am NOT as anal about case mods... a quick "dremmel hack" 92mm case fan mod...ugly as sin (in the back) but gets the job done







contact test...








...........








i was SOOO close to getting the CPU and HSF to WRING together, there was a strong attraction, but it was not enough to hold the weight of the CPU.
(if i were to lap the CPU down to pure copper, im sure i could wring them together but im SOOO sick of lapping @ this point!)


** wring:
To measure with a gauge block you simply select a combination of blocks (using the fewest in combination you can, in a process called stacking) which adds up to the length you want, "wring"
the blocks by rubbing the appropriate ends against each other to squeeze the air out from between them, and compare the length to the target.
One interesting feature of gauge blocks is that the surfaces are so smooth and flat that once they are mated in this way they stay together
unless considerable effort is used to part them. The combination of the adhesive action of the ultra-thin film of preservative oil or moisture on the blocks and
the molecular attraction, or bonding, between the very flat and parallel mating surfaces, will actually hold them together.




EDIT: HERE is a link to the entire build and "edited" posts...some good info in the edits (IMO)... POSTS: 65-71

.

 

[Ven0m]

Nice photos and explaination, Joe. Good job.

Guys, if you haven't got graph paper around, I'd recommend to check reflections of more distant lines (increase distance between heatsink and sheet of paper). The greater distance, the more visible curves will be.

Laser pointer test would be probably the best: laser -> heatsink -> distant screen (paper sheet, wall, etc). Then you slide heatsink (or eventually pointer) along a ruler or something like that. If point moves on screen, it means that heatsink base isn't flat. This test is pretty precize, so you need to be careful with sliding and even with pretty flat base, some pointer moves will appear.

It should be pretty easy to calculate flatness, knowing distance between heatsink and screen, angle between laser beam and heatsink surface and range in which point on screen moves.

 

[bobad]

wring

OK guys, when someone uses that term, he has worked in a metrology lab. That means he knows his business about finish, flatness, and parallelism. Attention I would pay to Camel Joe!

Another interesting thing about wringing together 2 lapped pieces of material:

If both pieces are 1" square, the wrung in surface will hold a 14 pound weight. That's because the air pressure pushing them together is ~14 PSI at sea level.

If you can wring a heat spreader and a heat sink together like jo blocks, believe me, AS5 will not improve the contact area. You can't get more intimate contact than with wrung in surfaces unless you get into something tricky at the molecular level.

Joe Camel, did you ever run across my write up on a molten metal interface?

 

[Joe Camel]

OK guys, when someone uses that term, he has worked in a metrology lab. That means he knows his business about finish, flatness, and parallelism. Attention I would pay to Camel Joe!

fancy terms do not a smart man make


i was in 1 metallurgy class in a community collage way back when...

actually, my knowledge/skill has been passed down from my father...


although i dont have any top-notch "Swiss Made" equipment like this:

.... I am "Swiss Made" ; my father is a Swiss mechanical engineer

bobad, i think i saw the thread but dont/cant remember it,
i shall look it up (although i SHOULD do it on NON-work time...)


HERE is a post/pic of copper plates i WAS able to wring together
(didnt post/take pics of them wrung in as i had nothing to prove back then)



PS there is a BIG difference between "Swiss Made" and "French Maid"


.

 

[Sentential]

More importantally did you see a temp drop after all this effort?

 

[Joe Camel]

never installed it stock. so i cant tell ya

go to the F@H thread i linked to in post # 42 @ the bottom for more (non lapping) info / pics

EDIT: EZ link to the "results" post: HERE (note the title)

 

[ls7corvete]

Do you guys start to see different thermal compounds working better as you get to such flat surfaces?

I would think something like nanotherm would work better than a particle based paste like AS5 if you can get near "wring" flatness.

Even though, if I remember correctly, properly applied paste should not yeild measureable differences between brands/types....

 

[Joe Camel]

Do you guys start to see different thermal compounds working better as you get to such flat surfaces?

I would think something like nanotherm would work better than a particle based paste like AS5 if you can get near "wring" flatness.

Even though, if I remember correctly, properly applied paste should not yeild measureable differences between brands/types....

sorry couldnt tell you. this is my first CPU "lap-job" and i only use AS5... (although i do have AS3 and Ceramique around)

as this is a 24/7 FOLDing rig, i was in a "rush" to get it back up and FOLDing!!

if/when i do this again (lap both CPU and HSF) i will get them to wring, and i will test a "no TIM but wrung together" test...although it wont be too easy...

Whenever I read about a thread where someone is lapping a Swiftech, I cringe....

here is a REAL "cringer" for ya / us ALL...lapping the G5 LINK

 

[Pf.Farnsworth]

I have a question. Is a mirror finish any indication of the true quality of the finish or is it just a result of glazing of some sort or any other factors?

 

[subtotal]

a mirror finish means that the tiny ridges in the metal have ben removed, although this does not indicate overall flatness. if you ever lap a heatsink base to the point of polishing it hold the base up to some graph paper, ALL the lines on the graph paper should be straight, if they aren't the base isn't flat.

btw, many people consider hand lapping to a mirror finish pointless, and some believe it even hurts performance

 

[Joe Camel]

I have a question. Is a mirror finish any indication of the true quality of the finish or is it just a result of glazing of some sort or any other factors?

it *can* be.

a perfectly flat object will reflect the light... perfectly.

a shiny object can also reflect the light but will have distortions in the reflected image.

its *easy* to make a surface shiny...its not as easy to make it flat.

can you have a "perfectly" flat surface that does not reflect? ...i would assume so... but by the definition of flat...

huh, anyone else want to jump in here? i just know how to push some metal on some sandpaper...

ok 1 more try:

shiny is about the amount of light reflected off a surface...FLAT has nothing to do with light. but does have to do with how "true" the light is bounced off the surface...and since most objects that are flat are shiny too...

PS Stay in school, dont do drugs and to the forum!!

 

[Skeen]

^ LOL

Keep trying buddy, you're almost there.

 

[Pf.Farnsworth]

hmm, well there seem to be 2 dimentions of flat on a heatsink, one being the flatness in relation from one corner to the opposite, ie the flatness of the square (lets say its a square). Then there is the surface quality. Am I totaly off but imagine a perfectly flat cube, and on the top face there is a scratch. Now being that the corners on the surface at the same elevation save the scratch in the middle I would say its a surface quality issue.

Am I way of or does this sort of sound right:

-shiny objects are smooth but not necessarily flat

-if an object is overall flat and has iregularities on the surface then its flat but not smooth, the imperfections making it not absolutly flat

-a perfectly flat object with perfect surface is absolutly flat

 

[Joe Camel]

well, for what its worth...I like it!

 

[Silver]

More importantally did you see a temp drop after all this effort?

Well, when Joe and I talked about this I expressed to him that just the lapping of my p4 heat spreader had allowed for much better contact and cooled things down a good deal. I had no pattern to speak of when the heatsink was removed. Unfortunately I did not have before and after temps.

Yesterday afternoon, I lapped my sons xp-90 (celly d) as fold temps for the last month (qmds) was running right on 58c. I found that to be higher than my liking and when he had a game crash on him, well got off my fanny and started lapping. Lapping is not something that I have enjoyed doing over the years as it for me takes some time to do. At any rate, when I pulled the heatsink I found the expected (based on prior experience with the xp-90) contact with the edges of the celly D's heatspreader and absolutely no contact with the center which we all know is the core area. Several hours later, all was flat as far as lapping the xp-90 and the heat spreader on the processor were concerned. Assembled and noticed the expected temp drop right off. All other specs exactly as before (1.575vcore, 17 x 200fsb, celly 2.26 D). Folding a QMD over the next few hours showed a loaded temp of 44c. This is a 14c drop in temps. Left machine running all night on the qmd as it is a 24/7 folding machine. Today at 2:56 PM it is still folding stably though temp has dropped an additional 2c. All told, a 16c drop in temps. Far exceeds temp drops experienced over the years with other lap jobs. I attribute this to a generally well designed heat sink with very poor cpu contact. I do not subscribe to the "place a pea sized" amount of as5 and then apply the heatsink theory. I have always placed the small pea sized amount then spread it out to provide a thin layer. If the contact areas are flat then this does not create a problem. If the contacts are not flat then the non spreading of the thermal compound makes perfect sense as at least some heat will be transferred. With the as5 spread out prior to attaching it becomes very, very obvious that NO, NONE, ZILCH contact is being made over the core. If your experience with either the xp-120 or xp-90 or for that matter any heatsink, water block or cooling device is less than expected I would highly recommend pulling the heat sink and looking at the spread pattern. There should be almost no as5 (thermal compound) on the heatsink where the heatspreader makes contact. In the "old" days with amd, you would actually be able to read the core in the imprint made on the heatsink. Not possible in most cases these days with the heatspreader. One might also consider why those experienced (not for novice) oc'ers are experiencing such improved temps by removing the heatspreaders. Area is decreased allowing for better contact on other wise poor bases/heatspreaders. Might try lapping the processor/heatsink base prior to removing the heatspreader and exposing the core. I question any of the processor manufacturers making heatspreaders that do not make contact with the core. The interface of the heatspreader with the cooling mechanism would be highly suspect in my mind. After all, no contact of the core with the heatspreader would most certainly end in immediate death of a manufacters processor. I more suspect the heat spreader of making contact on the out edges. This would allow for stock stability and not much more. By design or by error is of no relavence (sp) as thermal transfer is not ideal. In the case of the xp-90 we know there is a problem, for other manufacturers, I would use the flashlight against a number of known edges to confirm flatness of the base prior to doing any lapping, especially with known flat bases (i.e. swiftech). I would check the imprint after seating a heatsink. I would further (especially in the case of a known high quality manufacturer, read swiftech) take a good long look at the heatspreader on the cpu. I find in the case of the xp-90, both the heatspreader and the heatsink to be of equal blame. Ideal results (from an otherwise well designed heatsink) have been attained by lapping both the heatspreader and heatsink. Both of my experiences are related to Intels and the application of the xp-90. One a Celly D and the other a P4 northwood.

Celly D is now much more repsonsive at the same settings indicating to me I had some thermal issues going on even though it was stable.

P4 was able to overclock an additional 200mhz.

 

[Joe Camel]

thats GREAT news!!

im glad to see that there are some open minded green/blue stars here!

Silver, as you were one of the senior members that originally disagreed with Sent, i, and a few others who found these HSF to not be flat;
my "hat's off to ya" for having an open mind and trying it for yourself!!

the loss of 16C (load) is an incredible gain in performance and well worth the few hours, few dollars and a sore arm!!

congratulations!!


EDIT:

here is a quick pic of the FX55 that was in my phase change (MachII) sig rig....and a good example of the "bad" pattern Silver was speaking of

 

[tommcd64]

I did my xp-120 the easy way early on. For the price I paid for it when they first came out I sent it back and got it replaced. End users shouldn't have to lap heatsinks, especially pricey ones. If they are rough or not flat, ship them back for a replacement.