Optically Flat

Jokester · 12 Oct 2006 at 17:40

Someone a while ago posted an image of a criss cross pattern that you could use to check the flatness of a CPU or HS to see if it was flat or not by looking for distortions in the straight lines. Anyone got the image to use or a link to where it was?

Jokester

tom_e · 12 Oct 2006 at 18:06

http://overclockers.com/ theres something like that on there called the poor mans flatness test

Cob · 12 Oct 2006 at 18:13

Jokester · 12 Oct 2006 at 18:14

Ta, just found it as you posted it, cheers guys.

Jokester

Edit: Looking nice and flat

Jokester · 14 Oct 2006 at 21:52

Sorted.

Haven't tested it out yet, but it looks like it might have dropped my temps 5-10C.

Jokester

Defcon5 · 14 Oct 2006 at 22:20

Sorry for sounding dim, but what do you actually do with it?

Jokester · 14 Oct 2006 at 22:23

You look at the surface of the chip at an angle so it reflects what's on the screen. If the surface is flat the reflection will be a perfect image of what's on screen. If it's not flat the lines will appear curved.

Jokester

torrentthief · 14 Oct 2006 at 22:35

lmao, thats pretty crazy, what next, using a plumb-bob to make sure you aligned your board perfectly vertical in your case? hehe

Jokester · 14 Oct 2006 at 22:37

You should have seen the state of it before (Conroe 6600), shockingly bad finish when you get it.

Jokester

Nate--IRL-- · 15 Oct 2006 at 02:28

torrentthief said:
lmao, thats pretty crazy, what next, using a plumb-bob to make sure you aligned your board perfectly vertical in your case? hehe

Oh man thanks for that, got a laugh outta me.

Nate

Participant · 15 Oct 2006 at 10:22

Jokester said:
You should have seen the state of it before (Conroe 6600), shockingly bad finish when you get it.

Jokester

But what did you do with it?

Jokester · 15 Oct 2006 at 10:46

Robbie G said:
But what did you do with it?

I lapped the top of it. Basically you take sand paper of increasingly fine grit and polish the top against a flat surface (I used a sheet of glass). Took about 10 mins of the rough stuff just to get the surface flat the top was so bad. The Conroe heatspreader is concave so it dips towards the centre. This means that in the middle you can't get decent contact and this is where you need it most as that's where the core is situated. Once it's flat it's just a case of using finer sandpapers to remove most of the surface roughness (the microscopic peaks and valleys) as this will aid heat transfer to the heatsink.

Currently appears to have dropped my load temps by a good 15-20C :eek:

.

Jokester

harris1986 · 15 Oct 2006 at 11:31

air or water is this?

Jokester · 15 Oct 2006 at 11:32

Water, my loop has an even bigger heat load as well, but the difference is huge.

Jokester

harris1986 · 15 Oct 2006 at 11:33

whats the temps you are getting now idle/load? and at what clocks volts etc

just wana see how mine's comparing on air!

Jokester · 15 Oct 2006 at 11:37

I've upped the voltage to 1.55V (was about 1.5 before) and currently sitting at 3.6GHz.

Idle temps are 25C (was about 30 before if I remember correctly)
Load temps so far are about 45C though I was hitting about 60C before.

Going to be trying for 3.8GHz Prime stable.

Jokester

"andy" · 15 Oct 2006 at 12:49

wish i still worked at my old work and i could possibly have the flatest cpu ever.

monochromatic lights and optical glass flats coupled with fancy lapping tables for the win

surface flatness within millionths of an inch

harris1986 · 15 Oct 2006 at 13:25

not got contacts there still andy?

FrankJH · 15 Oct 2006 at 16:28

hmmmmmm - Jokester I respect your knowledge and experience but this is laughable - yeah dotn get me wrong I know there will be miniscule peaks and troughs that when smoothed out will aid cooling ( and I am guessing water will benefit more than air) but by 20 degrees!!! I cant see it myself but hey ho

Jokester · 15 Oct 2006 at 16:32

It's not the peaks and troughs in this case though. The IHS on a Conroe isn't even flat, let alone smooth. When I took my water block off it wasn't even making contact in the centre of the block the IHS was so warped.

Jokester