PDA

View Full Version : Is Bowing a Band-aid?


DavidNJ
04-12-2008, 05:25 AM
http://pctuning.tyden.cz/ilustrace3/highlander/WB/grafy/pritlak.jpg

It appears to me that it:

Compensates for excessive thermal grease
Compensates for non-flat surfaces


As the picture shows, it reduces contact the area over the dies with an appropriately thin layer of TIM, and leaves a thick layer of TIM elsewhere, reducing heat transfer in the cooler areas of the die.

If the IHS is 30mm x 30mm, the area in contact under the bowed block is about 10mm x 30mm. If the clamping force is 70#, pressure at the contact area went from 50psi to 150psi. It would require more force than the IHS and pins could handle to achieve this over the full IHS.

This would also account for improvements using a backplate. However, would there be an improvement from a stiffer backplate?

If correct, then would stepping the base of the water block, so it just contacted the IHS over the dies, increase its effectiveness?

And stepping into the air cooled world for a second, it would mean, as I beleive, that in exposed heat pipe HSFs, only the pipe over the dies does the work. If so, a design with a single, flattened 10 or 12mm pipe and press a large flattened area of the pipe against the IHS should be more effective than the designs that use multiple exposed heat pipes or heat pipes in a copper block. That is, a Xigmatek HDT 1283 could be improved by reducing the pressure on the two side heat pipes and that a TRUE (Thermalright Ultra-120 Extreme) could be improved on my several degrees under load.

jimmyz
04-12-2008, 05:55 AM
yes it is. and yes a stepped block should be better. the problem is when the manufacturer affixes the hs to the core sometimes it will bow the hs inward (concave) as it sets, sometimes it is outward (convex). although the vast majority are concave some are not and a rare few are flat.

the bowing of blocks is designed to give those xtreme tweakers another option for enhancing performance. some cpu's would benifit from it some wouldn't.
IMO. you should have gone a little at a time and tested the contact each time, simply cranking it to maximum bow isn't the way to do it.