HARDOCP ADVISES YOU SHOULD SANDING YOUR CPU!!!!

[bloodbob]

LOL @ HARDOCP

HSF Lapping Guide:
Tweak Factor has a HSF lapping guide posted today. Here is a clip from the guide:

While lapping is simple and hard to screw up, you should still know that you must take all responsibility for any mishaps, including sanding the cooler such that the surface becomes uneven and ineffective!

Hopefully I am mistaken, but I am not sure you will want to follow the advice you see in the guide. Lapping your HSF helps, but it should never be done by holding the sandpaper in your fingers as shown in the review. Always use a completely flat surface to attach the sandpaper to or you will actually make the surface worse by using your fingers or hand. I also advise lapping BOTH mating surfaces ( the HSF and CPU ) together to make sure both surfaces are lapped in unison making a better mating surface between the two surfaces.

 

[AAlcHemY]

It woudn't supprise me that tempfreaks do that

 

[Slides]

Lapping a P4 with the heatspreader shouldn't be too risky, but I wouldn't do that with an Athlon.

 

[Sage]

yeah you're NOT supposed to lap a CPU that has the core exposed... I lapped all three of my Tualatins (2x in a server, 1x in a desktop) and it didnt make much of a difference on any of them.... then I lapped my GF4 and holy crap it made a hell of a difference. P4's, I imagine, would benefit greatly from lapping as well (why dont they just lap the heat spreaders when they make them? it would save a lot of money on cooling.)

 

[Sxotty]

I did my athlon and it made it run cooler, oops I guess but who cares it did no harm.

 

[Natoma]

When I took the stock cooler off my 9800 pro, the face of the core, after cleaning away the default paste, was quite shiny. It looked like a perfect mirror.

I suppose they don't need lapping.

 

[AAlcHemY]

When I took the stock cooler off my 9800 pro, the face of the core, after cleaning away the default paste, was quite shiny. It looked like a perfect mirror.

I suppose they don't need lapping.

Aah, finally ? Just take the heatspreader off ..

 

[Sxotty]

How shiny it is has little to do with how flat it is, it is more related to the rate of change and not the max/min variation. And having a flat, but not shiny surface is preferable because there is a greater surface area that way.

 

[Sage]

update: actually, one of the tualatins in my server was very poorly lapped (i spent like 3 minutes on it) and it runs consistently %10 warmer than the one that I lapped properly. They are sitting side-by-side in the same system with the same heatsink so there shouldn't be much of a variation due to operating conditions.

 

[Nathan]

why dont they just lap the heat spreaders when they make them? it would save a lot of money on cooling

Lapping is a very expensive process. If I worked at Intel, there's no way I'd suggest to lap the heat spreaders. A far better alternative is to monitor the wear in the tooling used to create the spreaders. When you're producing millions of CPUs, the cost of tooling, as a percentage of the total manufacturing cost, is tiny. With lapping, the cost is largely independant of production volumes, and it can be very labour intensive process, so it doesn't apply well to mass production.

 

[Sage]

but if it reduces the core temp by 10 percent or more then it would increase yields of higher-clocked parts

 

[Tagrineth]

but if it reduces the core temp by 10 percent or more then it would increase yields of higher-clocked parts

Which brings us to another question for management:

Do the slight yield improvements bring in enough additional profit to make up for the expense of lapping the cores?

 

[RussSchultz]

I spoke with one of our production guys, and "lapping" the die is standard procedure.

Generally, the wafers start out about 1/8th of an inch thick to retain their rigidity in manufacturing.

Once finished with the fab process, and prior to cutting up into separate die, the bulk of the wafer substrate is ground away. This allows the packages to be as thin as they are and it also allows better heat management.

 

[Nathan]

I spoke with one of our production guys, and "lapping" the die is standard procedure.

That makes sense.

Crappy CPUs already get speed binned, so lapping the heatspreader is a bit of false economy. Also, improving the surface on the heatspreader does not guarantee that the heatsink will be mounted properly. Intel have assumed a minimum level of competency for heatsink installers, which is far below the level of competency that the average "lapper" has, I would imagine. Intel will be trying tp improve their yields by using methods that they have complete control over. Relying on a bored assembly line worker at HP to do a perfect heatsink installation is too risky.