Lapped my Q6600 :-)

[trazom]

Just want to share my lapping experience. I was a little skeptical about how good it would do to my CPU, I'm quite impressed in the end!

I went to buy some sandpaper sheets (I took 100 (cheap), 240 (cheap) and 600 (waterproof, good brand) grains, they didn't have anything more thin)

Before start, I took the cpu off the computer, remove the thermal paste with a soft cloth, and put back the little plastic protection cap that was provided with the CPU.

I lapped on a working desk, started by 100 paper, moving in a circular way until I can start to see copper on the outside and then in the center. I used one sheet of sandpaper.

Then I continued with 240 paper, until the moment the last non copper area disappeared (used one full sheet)

Then I used 600 paper, I noticed that if I made too big movements, the cpu would become grey, so I started doing very small circular movements, and the result was better: copper almost shiny. I used one full sheet.

Total time took about 40 minutes.

I used a soft clothe to clean the cpu, blow on it to make sure there was no dust left, and put it back in the computer, put some thermal paste (akasa AK-450) and the cooler (after cleaning it a bit)

Here are the results @ 9x333Mhz = 3Ghz on a P5WDG2 WS Pro MB and a Zalman CNPS9700 cooler :

Temperatures are the highes measured by OCCT during a 1 hour session (since it read temps from speedfan, core temps are 15°C less than real):

Before lapping:

System: 36°C
CPU : 71°C
Core 0 : 55°C
Core 1 : 55°C
Core 2 : 53°C
Core 3 : 53°C

After lapping:

System : 35°C
CPU : 63°C
Core 0 : 47°C
Core 1 : 46°C
Core 2 : 43°C
Core 3 : 45°C

That's a 8°C gain! I didn't think I would get more than 1 or 2°C


David.

 

[JAG87]

you do understand that your temps are wrong.

there is no way your cpu temps is 71c while your core temps average 55c if anything it should be the other way around. if you cannot rely on core temp programs (since your temps are obviously wrong), what you need to find out if the actual temps of your cpu is as reported, and then add about 10-15c for core temps. so if you are running 63c, your core temps could be anywhere from 73-78c. and before when you were at 71c, your core temps were between 81-86c (which is kind of hot for 3 Ghz by the way, even with unlapped IHS).

 

[trazom]

Originally posted by: JAG87
you do understand that your temps are wrong.

Well if you read my post above, I mentionned it :

Temperatures are the highest measured by OCCT during a 1 hour session (since it read temps from speedfan, core temps are 15°C less than real):

Now the CPU temperature is measured (approxed) by a MB probe, and it is not off 15°C so it should really be around 71°C. For the 4 cores, you need to add 15°C indeed (68 - 70°C before lapping, 58 - 62°C after)

Core Temp of 70°C @ 3Ghz at full load seems normal for me on a Quad Core


Anyway the point was a gain of 8°C which is very nice.

 

[JAG87]

Originally posted by: trazom

Originally posted by: JAG87
you do understand that your temps are wrong.

Well if you read my post above, I mentionned it :

Temperatures are the highest measured by OCCT during a 1 hour session (since it read temps from speedfan, core temps are 15°C less than real):

Now the CPU temperature is measured (approxed) by a MB probe, and it is not off 15°C so it should really be around 71°C. For the 4 cores, you need to add 15°C indeed (68 - 70°C before lapping, 58 - 62°C after)

Core Temp of 70°C @ 3Ghz at full load seems normal for me on a Quad Core


Anyway the point was a gain of 8°C which is very nice.

oops, my bad then. I missed those brackets

still, 70c tcase temperature is a bit high for 3 ghz. the most Ive seen my qx6700 lapped, at 3.2 ghz is 65c. maybe try re applying the paste? try the blob method, i find that its the best. it spreads into a perfect circle on the IHS.

 

[Diogenes2]

Interesting results .

What HSF are you using ? Did you Lapp it also ?

 

[trazom]

still, 70c tcase temperature is a bit high for 3 ghz. the most Ive seen my qx6700 lapped, at 3.2 ghz is 65c. maybe try re applying the paste? try the blob method, i find that its the best. it spreads into a perfect circle on the IHS.

Well, I don't know, at stock speed with stock intel cooler it was reaching 69°C on load (4x prime95). Might also be related to the MB, it seems to overvolt the CPU by default (I keep default voltage in bios), I know that very slight overvoltage can increase the temp few degrees.


I tried the lapped CPU in another machine with water cooling, but here I don't see a so big improvement.

I go from 39°C Tcase to 35°C when idle (cpu @ 3.2Ghz (8x400))
and from 58°C to 56°C when on load (only 2°C improvement compare to 8 on air cooling). Maybe I'll have to remove the waterblock, clean it and re-apply thermal grease. Or maybe it's because of watercooling and my temps are already "low".

Edit: It might also be because I compare 2 different CPUs, I have 2 Q6600, one lapped, the other not. The first post compare temps of the same CPU before and after lapping, on the water cooled system, I compare temp of one unlapped cpu to the lapped one (different cpu)

I didn't lap the waterblock nor the air cooler (zalman CNPS9700)

 

[aigomorla]

My Laps on my Quadcores:

http://i125.photobucket.com/al...aigomorla/IMG_0569.jpg

http://i125.photobucket.com/al...aigomorla/IMG_0703.jpg


I will LAP all my CPU after 1 month of testing. And yes they make a HUGH improvement on quadcores, and evening the temps out on all 4 cores.

 

[Diogenes2]

I noticed when I changed from CNPS9500 to Ultra-120 that my core temp distribution changed as well as went lower; but one core might be 4 - 5 hotter than others ..

I will probably lap also.. Cooler is better...

 

[trazom]

I will LAP all my CPU after 1 month of testing. And yes they make a HUGH improvement on quadcores, and evening the temps out on all 4 cores.

Yes, on air, the difference is huge, on water cooling it is a bit less. I noticed you have a thermaltake case on your picture, do you have the one with water cooling?

I'll wait a bit to lap the second quad... I kind of hurt my wristle ;-)

 

[PCTC2]

Originally posted by: trazom

I will LAP all my CPU after 1 month of testing. And yes they make a HUGH improvement on quadcores, and evening the temps out on all 4 cores.

Yes, on air, the difference is huge, on water cooling it is a bit less. I noticed you have a thermaltake case on your picture, do you have the one with water cooling?

I'll wait a bit to lap the second quad... I kind of hurt my wrist ;-)

HAHAHA. Aigo with a TT W/C Set-up?!?! He would probably kill himself. He has all custom loops.

Plus, corrected spelling.

I have yet to lap my Ultra120 or my Q6600 B3 b/c I have yet to find a local store with anything above 210 grit. Temps as of now in 23C air-con ambients, load after 7 days straight of F@H as measured by CoreTemp 0.95 and Everest Ultimate Edition:

CPU: 48C
Cores: 63/62/63/60

 

[aigomorla]

Originally posted by: trazom

I will LAP all my CPU after 1 month of testing. And yes they make a HUGH improvement on quadcores, and evening the temps out on all 4 cores.

Yes, on air, the difference is huge, on water cooling it is a bit less. I noticed you have a thermaltake case on your picture, do you have the one with water cooling?

I'll wait a bit to lap the second quad... I kind of hurt my wristle ;-)

dont have the armor anymore. I'll miss that case.

Gave it to my best friend and my third quadcore on loan. It was retrofitted with a Gigabyte P965 and X3210. Water and loop still looks the same, and its front mounted PA120.2 does a good job at cooling the cpu and board only. Her GFX card is on a seperate loop off the rear radiator.


My Q6600 G0 ES is in my Mount Mods U2 Case.

http://i125.photobucket.com/al...aigomorla/IMG_0721.jpg


That is what it took to keep my old Q6600 B3 @ 3.6 stable enough to be pushed to her max 24/7. I WCG crunch for XS, so i push all m yquads to the max.

I have a real reason for owning 4 quads. and they all goto a very good cause.

www.worldcommunitygrid.org

 

[graysky]

8 °C is pretty good for < $20 worth of stuff isn't it... about your temps though. Did you try to minimize your vcore for 9x333? Have a look at this thread for more on this. Scroll down a ways and look at the results I posted comparing two different vcores that differed by only 0.120 V. The temp difference was huge. Also, did I miss it or did you lap your HS as well? Finally, did you take a pic of your lapped chip before you reinstalled it?

Here are my lapping threads:
Lapped the base of my Ultra-120 Extreme

Lapped the IHS on my Q6600

EDIT, I just saw this:

Originally posted by: trazomEdit: It might also be because I compare 2 different CPUs, I have 2 Q6600, one lapped, the other not. The first post compare temps of the same CPU before and after lapping, on the water cooled system, I compare temp of one unlapped cpu to the lapped one (different cpu)

I don't think that's a fair comparison unless both chips are running on the same HS and using the same vcore.

 

[VERTIGGO]

Does everyone think that lapping until mirror like (as opposed to stopping when the copper is smooth) outweighs the disadvantages of having curved surfaces. I mean you can be careful and lap very flat surfaces, but by hand... you're going to end up with some degree of convexity.

 

[Miramonti]

How much value might there be in lapping a e6750/6850 c2d?

 

[lopri]

aigomorla: That's an absolutely beautiful lap job. How long does it take to have a mirror shine like that?

I wonder if there is any service that does lap a CPU for a fee? (if so please provide the links, plz) I'd like to try it out (a lapped CPU, not lapping ) but I neither have time nor skill..

Edit: What the heck is Malboro 'Smooth'?! I didn't know such thing existed.

 

[graysky]

Originally posted by: jjsole
How much value might there be in lapping a e6750/6850 c2d?

You're kidding, right dude? All C2D/C2Q chips (at least all the ones I've seen) suffer from being mild to moderately convex (like a salad bowl). Have a look at my CPU lapping thread; I knocked off about 9 °C from my load temps.

 

[graysky]

Originally posted by: lopri
aigomorla: That's an absolutely beautiful lap job. How long does it take to have a mirror shine like that?

I wonder if there is any service that does lap a CPU for a fee? (if so please provide the links, plz) I'd like to try it out (a lapped CPU, not lapping ) but I neither have time nor skill..

Edit: What the heck is Malboro 'Smooth'?! I didn't know such thing existed.

You may not the time, but I'm sure you have the skill. I did my q6600 and hs only by the info I got through reading a few guides. It's very simple to do. I may have asked you before, but are you water cooling your q6600 @ 8x400 in your sig? What is your vcore?

 

[lopri]

Yes I believe I answered the question before. Throughout entire life as an enthusiast I've never dared to touch anything other than air-cooling. Kinda shame, I guess, but I am not that a handy person and water-cooling has always been intimidating to me. (so does everything else other than air-cooling, lol)

vCore I set in the BIOS is 1.43125V to be exact. I've never measured the exact vDroop on my board using a DMM, but Everest/cpuz reports approx. 1.37~1.39V between idle/load.

 

[aigomorla]

Originally posted by: lopri
aigomorla: That's an absolutely beautiful lap job. How long does it take to have a mirror shine like that?

I wonder if there is any service that does lap a CPU for a fee? (if so please provide the links, plz) I'd like to try it out (a lapped CPU, not lapping ) but I neither have time nor skill..

Edit: What the heck is Malboro 'Smooth'?! I didn't know such thing existed.

this is usually what i do.

i have a large flat glass desk. So the flatness is almost gaurentee'd. I go out and buy another flat piece of glass to stack on top of the desk, so i can make sure its about as dead flat as possible.

Get some vegi oil, as a lube.

Go out to an auto store, not hardware store. Pick up an assorted pack for 8.99

This pack comes with 400, 700, 1000, 1500, 2000

The mirror's start at 1000-2000. The 700 is all you really need if you just want flat.

Draw an X on top of the cpu to gauge how much your shaving. <--- important

Start on the 700, to see how badly unflat it is. If its not so bad, then i continue on the 700 until i see copper. If its really bad, i'll work on the 400 a bit til i see copper, then work up on the 700.

Once i see copper evenly thoughout the CPU, i'll then work on the 1000 til i see a nice clean mirror, and then 1500, even more mirror, Then the 2000 you get the picture i posted on the second one where u see my mobo in the background.


:]


I noticed you wont see much of a difference from 700-2000 in terms of temps, but from a astetic standpoint, Priceless!


Computers are my hobby, so spending the extra 30 min to get that mirror finish is like custom painting a gundam model or some sort. It wouldnt feel right if i left something half finished in a baby of mine. :T

 

[Bill Kunert]

After you've used the 2000 grit you can get an even higher gloss using printer paper as an abrasive. It's abrasive enough to clean contacts on relays and it did improve the mirror of my cpu and heatsink a bit.

 

[jedisoulfly]

Originally posted by: PCTC2

Originally posted by: trazom

I have yet to lap my Ultra120 or my Q6600 B3 b/c I have yet to find a local store with anything above 210 grit.

I have never lapped anything but it seems you should try local auto body shops I'm sure someone there could tell u where to get high grit paper or maybe sell/give you some who knows a quick search and I found this...as someone said before try auto stores

http://www.rightlook.com/index...ch=sand%20paper&Page=2

 

[TVisitor]

I'm an avid pinball person, and polishing and shining to mirror finishes on plastic is one thing I'm big into (I've even flame polished plastic ramps, but I don't think I'll need to do that to a IHS...lol). I wonder if using Novus 2, which is generally used to bring a shine and is a fine-scratch remover - would be useful? That said, I still have some 2000 grit around here somewhere.

 

[firewolfsm]

You only need to go to 6-800 grit, anything past that is useless, and might even raise temps.

 

[aigomorla]

Originally posted by: firewolfsm
You only need to go to 6-800 grit, anything past that is useless, and might even raise temps.

nah... If anything it would save off the .000001C for each flat contact on metal + metal vs metal + tim + metal.

Also you can get away with less tim per application due to less bubbles needing to get filled.


As i said once again, the difference from using 700 -> 2000 is bearly noticible. Its all in astetics. If your getting worse temps, that means you shaved one side too much and you now have a tilt.

 

[trazom]

The mirror's start at 1000-2000. The 700 is all you really need if you just want flat.

I get mirror at 600 already, and if you use polish you can have a very nice result (which I don't recomment because the product will make the temps worst unless you clean very well with isopropyl alcohol)

All depends how you lap, at 600, if I do too big moves, I don't get a mirror finish at all, but I make fast very little circular moves, I get a very nice result