Lapping : does lapping *only* cpu or *only* hsf really help ?

[NoobyDoo]

Never lapped before, so did some reading. Reached this conclusion - may be obvious to others, but took me quite awhile - to really get the maximum benefit from lapping, both the cpu and heatsink need to be lapped. Am I correct ? Came across many posts / articles on the net where users had lapped only one or the other. Not really worth it, right ? Even if one surface is lapped perfectly, but the other is not perfectly flat, there will be gaps between the two. Therefore unless you are willing to lap your cpu, in most cases, it's not really worth lapping your heatsink.

Please correct me if I'm wrong.

Thanks.

PS : If Intel can manufacture 45nm CPUs, why can't they ensure that the heatspreaders are perfectly flat. They do so for the stock hsf, according to frostytech :

The copper is machined perfectly flat, and is smooth. Surface roughness is on the order of ~8 microinches, which is considered excellent.

 

[Capitalizt]

Both need to be lapped to see a difference (at least they did in my case). A perfectly flat CPU won't do you any good if the heatsink has a warped surface, or vice versa.

Just don't ruin all your hard lapping work by putting on too much thermal paste. That's a common mistake. You'll probably need less than ever before. I use same amount shown in the "line method" on AS5's website...then I spread the paste out with a razor blade using a "paint the fence" motion to get a perfectly smooth, PAPER-THIN layer over the entire CPU. When both surfaces are flat, that is all you need.

 

[Markfw]

Well, I checked my new cpu, and couldn;t see the slihghtrst uneven. My HSF was lapped by SVC.com where I got it, and I run Q6600@3500 59c full load.

 

[Navid]

The objective is to minimize the (average) space between the two metal surfaces.

If both the CPU and sink are convex (outward curve), lapping only one helps because you would be reducing the average distance between the two.

If the two are both concave (curved inward), lapping only one still helps for the same reason.

If one is concave and the other is convex, lapping only one could actually make things worse.

 

[NoobyDoo]

OK, thanks all. Guess the lesson is to examine the cpu and hs first, and then decide whether the cpu or hs or both need to be lapped.

 

[Idontcare]

Originally posted by: Navid
The objective is to minimize the (average) space between the two metal surfaces.

If both the CPU and sink are convex (outward curve), lapping only one helps because you would be reducing the average distance between the two.

If the two are both concave (curved inward), lapping only one still helps for the same reason.

If one is concave and the other is convex, lapping only one could actually make things worse.

Well stated.

 

[Idontcare]

Originally posted by: NoobyDoo
PS : If Intel can manufacture 45nm CPUs, why can't they ensure that the heatspreaders are perfectly flat.

Frankly it is because doing so does not impact their product offering.

No matter the temperature that your CPU operates at, provided it is operating below the maximum allowed by Intel's spec sheets then it is still functioning perfectly as intended.

CPU temperature is not the same thing as power consumption. Your CPU is consuming the same wattage whether it has been lapped and operating at 40C or if it is badly concaved/convexed and operating at 70C.

The temperature is caused by the inefficiency of the thermal transfer of the heat from the CPU, not because the CPU uses different amounts of power depending on operating temperature.

(for you pedantic types out there please allow me to ignore Ohm's law here, surely yes a 70C copper wire has higher electrical resistance than a 40C electrical wire and therefore will create ever slightly more waste heat from the same current draw, but let us agree that is second order to the point I am making here)

So from Intel's point of view, you paid for a processor to run at XYZ GHz consuming ABC watts of power. Provided the warpage of the CPU covering is not so bad as to cause such horrific thermal transfer properties as to invoke thermal throttling (resulting in lower performance) then you are getting precisely what you paid for.

 

[NoobyDoo]

just kidding ... or maybe not ...

Just understood why :
a) Intel doesn't bother too much about the heatspreader on CPUs, but
b) The stock heatsink is "machined perfectly flat, and is smooth ... etc".

It's because :
a) The stock hsf is outsourced, and maybe it is easier to ensure a perfectly flat hs base.
b) If the base of the stock hs is perfectly flat, it won't matter that much if the cpu heatspreader is not perfectly flat.

------------------------------------------------------------------------------------

Is it true that base of Thermalright heatsinks are convex by design ? Remember a thread by graysky at silentpcreview where he says :

Yeah I did and yeah a tech from TR told me they are convex by design since most C2D/C2Q chips are convex on the average.

 

[Noubourne]

Originally posted by: NoobyDooIs it true that base of Thermalright heatsinks are convex by design ? Remember a thread by graysky at silentpcreview where he says :

Yeah I did and yeah a tech from TR told me they are convex by design since most C2D/C2Q chips are convex on the average.

No, they are not convex by design. TR pays their engineers, manufacturers and QC team thousands upon thousands of dollars to overlook something as basic as the flatness, shape and finish of the contact area of their heatsinks. It is this ignorance and carelessness that has made their products perform on par with - if not better than - all the other top tier heatsink manufacturers on the planet.

/sarcasm off