Gotcha
Thought you'd see something raunchy here, didn'tcha?
Well, think again, for I refer to the process called 'lapping'.
What is lapping you ask?
Its the process whereby a metallic surface is progressively ground down until its perfectly flat,
and in most cases polished to a mirror sheen. This is done to enhance heat transfer between
two surfaces that must come into contact with each other, in this case a heatsink and the CPU
heatspreader.
Based on the abysmal performance of the
Thermalright Ultra-120 Extreme HSF which I received
sometime ago, I decided to subject it to this lapping process in an attempt to improve the idle (50C)
and load (62C) temps that I've been seeing of late with this Core 2 Quad Q6600 (B3 stepping),
which was actually
worse than the
Zalman CNPS9500 I was using before.
So earlier today, myself and Saxman, with PCs in tow, journeyed to daniboy79's house, to do the deed.
The following items were used in the process:
- Sandpaper of varying 'grit' (roughness) from 200 up to 1000.
- The HSFs (of course)
- A flat glass surface
- water (you'll see why)
- razor blade
The razor was first used to check for flatness (or lack thereof). It was quite evident that these HSFs
were not flat at all, as the razor blade could actually be
rocked back and forth on the base. Not cool.
The sandpaper (coarse, 200 grit) was set on the glass and the water sprinkled on it so that a 'wet-sand' could be done.
The underside of the HSF base was then evenly and carefully dragged on the sandpaper until all the nickel plating on it
was worn off, and the copper underneath was exposed.
When all the nickel was removed, then the HSF was introduced to the lighter grit 600 sandpaper,
and the wet-sanding process repeated until all the harsh marks left by the previous grit were removed.
After each session, the razor was again employed to check for flatness.
After the 600 grit, the 1000 grit sandpaper was then used to polish that biatch until a mirror like surface was achieved.
So after about 2 hrs of steady sanding, this is what the HSF looked like:
(yes, that is my card
)
The pic on the left shows the lapped HSF (on the left
in the pic
) next to the unlapped one. Note how reflective
the surface is compared to the unlapped HSF on the left. Not too shabby...Of course we didnt stop there.
We had to make SURE that heat transfer would never again be an issue with the quad, so after a very meticulous
sanding of the CPU heatspreader (yes, you read that right) we ended up with this:
Now...how muthaf@#kin L33T is THAT!!?
Of course now you can't tell what CPU this is, but thats really no longer an issue here.
HSF was mounted (with Arctic Silver 5) using two 120mm Thermaltake fans in a 'push-pull' configuration as shown here:
(yeah, cable management could be better, I know)
The REAL issue at hand, is whether all this lapping would have made any difference in idle and load CPU
temps. I'll let the following screenshot speak for itself. Remember, the temps before this endeavour
were
50C idle, and
62C load. Quite hot, even for a 105W quad-core.
Actually the results are quite shocking, to say the least. Idle temps have dropped from 50C to
39C, a whopping
ELEVEN DEGREES Celsius drop. Load temps also show an incredible difference of eleven degrees also, from 62C to
51C.
A VERY impressive result. Now my OC'ing efforts can proceed unimpeeded.
Saxman also achieved some impressive results also, which he will post up in short order.
All in all, a very eventful, but fruitful project, and its not as difficult as I first anticipated. Special big ups to daniboy79 and
his plethora of tools,
and his wonderful work on this (yes, I watched
).
I can safely recommend lapping your HSF and even your Core 2 Duo/Quad if you're not satisfied with your temps. Its even
easier with these Intel CPUs as there are no pins to bend. Make sure and dry it thoroughly afterwards though.
As this project has demonstrated, it seems the recommended air-cooling HSF for this B3 quad-core is this lapped
Thermalright Ultra-120 Extreme. Especially if you plan to overclock it.
