# Derate heatsink when used passively

I want to reuse my old computer processor heatsinks for LED heatsinking. I think I can calculate their K/W value approximately by dividing the temperature difference vs TDP at their maximum load.

I ignore the CPU packaging thermal resistance and take it as an additional safety margin. This way I get about 60/100 = 0.6K/W for a beefy heatsink.

The problem is, that I want to get rid of the active component and derate the heatsink accordingly. Are there any approximations I can use for derating?

I know I'll have to add a healthy safety margin, but I still need a general figure as to decide what wattage can they sink.

• It's all in the datasheet. Oct 8, 2016 at 23:07
• @IgnacioVazquez-Abrams I tried searching for cpu heatsink data(admittedly after your suggestion), but my google-fu has abandoned me and I couldn't find any specifics. Oct 8, 2016 at 23:28
• @IgnacioVazquez-Abrams, if a heat sink is designed for forced air, it is very unlikely that it will have any data for use under free convection. See details below. Oct 9, 2016 at 2:17

It is a challenging problem to de-rate a heat sink. The problem is that if a heat sink is designed for forced air cooling (using a fan), the spacing between fins is optimized for forced air, and usually is fairly dense, 1.5-2mm apart. For cooling with free convection however, the fin spacing must be at least 1/4" (6mm) or more, for the ambient air to flow between fins with any reasonable velocity, and for thermal boundary layers not to overlap along the fins. Needless to say, you should maintain vertical orientation of cooling surface and vertical orientation of fins to have any positive effect.

In other words, forced-air heat sinks are usually unsuitable for free convection cooling. The best safe estimate would be to consider the entire heat sink as a brick.

CPU heatsink ought to be in the 0.2 to <0.1 'K/W range with silver thermal paste if using same surface area used, otherwise derate according to useage.

Then derate further if ALumClad substrate is not coplanar to within a few microns ... or just measure case rise. I have used 100W LED on a CPU heatsink with a small fan for < 85'C case temp

After following IgnacioVazquez-Abrahams recommendation, I found a way to approximate the thermal resistance of the heatsink.

Some manufacturers of heatsinks specify both the forced-air and free-convection rates of their heatsinks. By searching for similar heatsinks, and checking their free-convection rates I get from 2x to 10x derating. By looking at the geometry of the said heatsinks it is clear that this mainly depends on fin spacing(as already predicted by Ali Chen).

By comparing my heatsink geometry with the data I have, I think that about 3x/4x derating should be appropriate in my case.

Something like this(without the fan) gives me a value of about 1.5-2.5 K/W:

I am going to test this later and report my actual findings.