Madshrimps Forum Madness

Madshrimps Forum Madness (https://www.madshrimps.be/vbulletin/)
-   Articles & Howto's (https://www.madshrimps.be/vbulletin/f6/)
-   -   Lapping a P4 IHS: any good? (https://www.madshrimps.be/vbulletin/f6/lapping-p4-ihs-any-good-4893/)

jmke 12th May 2004 22:59

Lapping a P4 IHS: any good?
 
If removing the Integrated Heat Spreader from your P4 sounds too extreme, then maybe lapping it might be an alternative.

Time to find out,

Victim is a P4 2.0 "A".
Here's how he looks unlapped ->

jmke 12th May 2004 23:01

Using P600 sand paper to grind away the concave top of the IHS, and then proceed with P1200 to finish it off.
I used a small mirror as "flat surface", bought it at a local DIY shops for €5, pretty nifty :)

after lapping ->

jmke 12th May 2004 23:03

Although it is no "polished" sample here, I'm wondering if there is any impact on temperatures under load...

jmke 12th May 2004 23:09

CPU: P4 2.0Ghz - 1.54vcore
Env temp: 26.6°C
Cooling: Zalman CNPS7000-CU @ High
Thermal Compound: Arctic Ceramique


Load temps: 15min K7 CPUburn
before: 37°C
after: 36°C


there ya have it, gained 1°C by lapping the IHS; not spectacular. If you have the materials lying around it's worth it (free performance increase!) but don't go out and buy lapping gear because the return on investment is too low.

Gamer 12th May 2004 23:16

1°C is 1°C :)

nice, only thought the temp would drop more.

TerAngreal 12th May 2004 23:45

will do mine in a couple of days as well, good chance temperature drop will be more significant in an overclocked situation

Sidney 13th May 2004 00:29

Intel heat spreader is pretty flat to begin with. The lapping gets rid of the micro irregularity which, otherwise be filled with our AS5 with about 0.06 c/w representing about 6C at 100 watts.

On concave or convex surface lapping will yield much better result.

jmke 13th May 2004 06:31

this IHS was pretty concave :)

TeuS 13th May 2004 07:09

you're forgetting how accurate your results are. your measurements are accurate up to 1°. a difference of 1° between lapped and not lapped doesn't exceed the accuracy of the results.

it's possible the unlapped CPU runs at 36.6°, so it's displayed as 37°. i'ts possible the lapped CPU runs at 36.4 degrees so it's displayed as 36°.

the temperature difference can be very small. as you said, the return on investment is too low. and you're voiding your warranty

jmke 13th May 2004 08:31

Quote:

Originally posted by TeuS
the return on investment is too low.
thank you for repeating what I stated...

Xploited Titan 13th May 2004 09:27

Suggestion (even though I took the idea of someone else): couldn't you create some kind of channels in the IHS (like the Nexos HP-Pro...) and then do a direct-on-IHS watercooling?

I hope you understood me :D

jmke 13th May 2004 09:35

I think you will get better results removing the IHS all together in that case, since creating channels in the IHS is pretty extreme too ;)

jmke 13th May 2004 20:47

this 2.0 "A" is running stable at 2.7ghz using 1.65v vcore and aircooling :)

Sidney 13th May 2004 22:48

My 2A runs at 3Ghz at same voltage. They are very good OCer. Goes higher with 800 mhz board. =-)

jmke 13th May 2004 23:10

hmmmm.. should be doing something wrong then.. let me see if I can get it to run @ higher speed :)
edit: 3ghz = okay , running F@H session no problem so far :)

Sidney 14th May 2004 01:18

You're welcome. :klap:

Tum0r 14th May 2004 01:51

What coler was used?
Was the cooler lapped aswell?

jmke 14th May 2004 08:05

cooling: read 4th post :)
cooler is pretty flat already

Tum0r 14th May 2004 11:59

I got myself a lappings kit form tweakstore.

1 x 400grid (23 x 28cm)
1 x 600grid (23 x 28cm)
1 x 800grid (23 x 28cm)
1 x 1000grid (23 x 28cm)
1 x 1200grid (23 x 28cm)
1 x 1500grid (23 x 28cm)
2 x 2000grid (14 x 28cm)
2 x 2500grid (14 x 28cm)

With this i think u can get 3° lower then stock.

I know its hard to find anything higher then 600 in DIY shops.
U could try "carroserie werken" as they need very fine paper.

Unregistered 14th May 2004 12:40

Even for paintjobs on cars and such, they rarely go higher then a 1200-1400 grain.

Perhaps special-paint painters ;) will go the extra mile by using 2xxx grain sandpaper.

jmke 14th May 2004 13:20

Quote:

Originally posted by Tum0r
With this i think u can get 3° lower then stock.
proof baby. I dont think the difference between 1200grid and 2500grid will account for 2°C temp drop.

Sidney 14th May 2004 21:20

Getting interesting day by day......

1200 already enough to make it mirror like finish; 2500 will make it feel like a baby's *** smooth. Making a 30-45 minutes lapping job into days of rubbing with 2500 grit.

I thought the purpose of John's article was about less extreme. Considering 100% perfect condition; the TIM is 0.05-0.08 C/Wl it would be like running faster than your own shadow.

insulglass 15th May 2004 13:00

Disclaimer: I sell more lapping kits than everyone else combined in the world.

Having said that, there are probably some points to address:

1) JMKE, in my humble opinion, you only did half-a-job. Lapping a CPU will typically only provide 25% of the thermal improvement by lapping. The biggest advantage will come from lapping the heat sink, which is far likely to be manufactured to lower finish tolerances. This is even more exaggerated by AMD processors with die exposure.

2) in my humble opinion, you did not lap through the available grit ranges. While the majority of improvment from lapping does occur at P100-P1200, there IS marginal improvement by lapping to P2000 and beyond.

3) Even though YOUR particular improvement was only 1C, other people may experience more. A drop of 2-4C is average, with many people reporting even more. A 1C drop is actually fairly rare, and I have experience from customer feedback from thousands. There are lots of people who will spend $50 on a new heat sink in order to get a 3C drop. :-)

4) I'd like to see more details about your lapping method. How much lubrication did you use? Did you use a glass backing plate? What was the brand/quality of the sandpaper? How much Ceramique did you use when you reapplied your heat sink? Etc... These and more can be factors that would affect your end result. <edit - I saw where you used a mirror>

5) The return on investment is actually pretty reasonable. I don't know about everyone else, but I sell kits pretty cheap. If done right, even with the nine-sheet premium kit, it should only take about 1-1.5 hours to lap a large heat sink.


Please understand, this post isn't a flame, I'm just pointing out some points that are worthy of consideration. In fact, I'll put my money where my mouth is.... JMKE, PM me and I will send you a free premium kit to take a look at and try out. Fair enough?

BTW... I have sent a couple of premium kits to Madshrimps, one to use in a review and another to use as a giveaway. Should actually be appearing soon.

jmke 15th May 2004 13:11

Hello Insulglass , welcome to our forums :hello:

1) the Zalman CNPS is pretty flat, but could indeed use a lapping, with 1200+ grid paper only though, since it has very few imperfections

2) I couldn't find any smaller grid then 1200 in local DIY hardware shops, 1200 is very fine already, and as you said, beyond 1200 performance increases will be neglilable;

3) It all depends on the original state of the material, if you have a badly lapped heatsink and your P4 IHS is very concave or convex then lapping, even with 600 grid, will give you a higher temp drop (2-5°C range).

4) this goes together with 3). I took pics and noted down data during the whole lapping progress and I'm working on a full "howto" article, including more P4 IHS and heatsinks to evaluate. here's the the info:
* used a mirror
* P600/P1200 silicon carbide waterproof , don't have the brandname lying around unfortunately, packaging has been thrown away
* how much Ceramique, dot in the center of IHS, makes the CPU looks like this afterwards (Another P4 & AS5 in pic)


I will send you a PM right away, thank you for the offer, we will evaluate the EasyPC kits fair and square. If they offer good performance increases over normal "grid's" and drop the CPU load temps considerable we'll make sure everybody knows it :)

Sidney 19th May 2004 23:49

Lapping heatsink does have more return of investment than simply replacing heatsink fan or just the IHS.

Due to manufacturing process in Mass, surface flatness is "always" questionable; hence Intel stock Hsf uses pad to "cushion" automated manufacturing process. Cutting, casting curling processes generate heat which cause the metal surface to "cure" or "settle" in different rate. Lapping an Intel Stock sink alone (both copper or alum) always yield a few Cs' reduction. With a few C's reduction from Intel Stock Sink, you are into the 3rd party high performance sink performance territory.

Based on my hobby in "sword and knife" over many years; a right job in lapping of two surfaces creates a so call "suction" effect that it seems to be super-glued together.

I only use glass block (8"x10") start with 400; 800; 1200 and finish with 1500 and a few drops of detergent and end with sewing machine oil.

My problem with lapping is; you'd almost have to re-do after each installation specially with "real" copper being soft the impression of the IHS is shown on the sink.

Improper lapping could ruin the sink. ;)

Should try with NB and Graphic card sinks; you'd be surprise!!

Depends on how many give away --- we need a workshop in lapping before the LAN Party. :rolleyes:

insulglass 21st May 2004 12:40

Well, I don't mind sponsoring the odd contest or two. Good for the site, good for my business too.

I don't want to give TOO many away, though. :D

jmke 20th June 2004 15:58

TerAngreal is using the EasyLapping kit on his P4 , after a few hours this is the result ->

jmke 20th June 2004 15:58

He will post temperature results very soon :)

here's a close-up:

TerAngreal 20th June 2004 16:26

this is what the victim looked like before i started the 2 hour lapping job :)

DUR0N 20th June 2004 18:22

keep in mind that lapping does not help one bit if you don't straighten out the concave form.
It might look like a mirror, but that doesn't mean the surface is flat.

TerAngreal 20th June 2004 18:29

you'll see i've taken that into account when i post the rest of the pics ...

here's an example which shows the previous curve of the IHS:

Magnum_ 20th June 2004 19:04

Isn't it better to take off that heatspreader, like in the article that was posted here a while ago?

jmke 20th June 2004 19:37

with more recent CPU you run the risk of destroying the core as it has been glued to the IHS.

Magnum_ 20th June 2004 20:12

no **** :/

I didnt think there would be substantial differences between those CPU's... Guess again :p

jmke 20th June 2004 20:15

ouch.. P4 with M0 core stepping?

Magnum_ 20th June 2004 20:23

SL79L costa rica prescott

Luckely I bought it rather cheap second hand (trayed), and I wasnt really satisfied with the temperatures. (that's why I took the risk). Initially, all went well. I've gently put a cutterknife between the heatspreader and the CPU, went all the way around a couple of times, but the damn thing didn't came off. Then, I tried twisting the heatspreader a bit, it didnt move a bit. Then I started carefully using the knife as a lever (hefboom?), and I heard a small crack. Not healthy, so I tested the cpu as it was right away. Dead offcourse ;) Then I took it off with brute force, and yes, the core came off with the spreader. It almost seems soldered to the heatspreader, there's this tin-like substance all around it, I can't even get the damn thing off the heatspreader using a hammer and a screwdriver as a chisel. I'll have to read the manual for my canon A75 if I have a moment, so I can take some clear pictures :)

Anyway, I'm looking for a cheap/fast northwood "C" cpu now ;)

jmke 20th June 2004 20:43

look for a small image of a Flower on your cam to enable Macro mode :)

insulglass 21st June 2004 04:53

Quote:

Originally posted by jmke
TerAngreal is using the EasyLapping kit on his P4 , after a few hours this is the result ->
Dude - that is a SUH-WEET picture. :D

Dave
insulglass@easypckits.com

TerAngreal 21st June 2004 23:07

=)

biCker 21st June 2004 23:38

looks sweet!


All times are GMT +1. The time now is 21:19.

Powered by vBulletin® - Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO