Blow or suck

[MakerMike]

IMHO it depends on what is in the other side of the fins.  Blowing air from the outside directly onto fins will give it the greatest cooling, if the outside intake is on the other side sucking would coll it the best.

[Hamster]

In this situation the way the heat sinks are orientated, i don't think the direction will matter.

However, if the other end will be in a compartment with other electronics, i would "SUCK" the air out to prevent the hot air heating the other components.

or blowing the hot air AWAY from the other components.

if the heat-sink is in free air, and the exhaust will not effect anything, and there would be no heat source from the fan inlet, is see no benefit of it sucking or blowing.

i follow this up with i have ZERO fluid dynamics education, except that, if a open bottle of bear falls over, its going to get the ground wet.

[Orpheus]

It's probably worth noting that the biggest fan in the world couldn't generate 15 psi of negative pressure inside a case (and if it did lower pressure = lower airmass = less thermal transfer, so at 10-14.7 psi your heatsink would increasingly be in a thermos bottle) but a huge fan could generate 100s/1000s psi positive pressure -- higher thermal transfer but you'd probably want  to precool that air before introducing it to the heatsink (compression concentrates intrinsic heat to a higher temp)

IOW, the asymptote works strongly in your favor in the regime of typical fans

[smoothtalker]

I would say blowing is better. Ive no way to verify. When your food is hot you blow it instead of sucking it. Right??

[AlfBaz]

Just a quick note on air flow quantity.

Whilst looking at data sheets it seems that increasing air flow at some point starts to have insignificant effect on increasing heat transfer away from the heat sink. Curves have a reverse exponential characteristic.

Knowing squat about the physics, it would appear to me that the radiation of heat by the metal is probably a constant and once you have sufficient air flow to immediately remove any heated air, increasing the flow of ambient temperature air will have little effect

[IanB]

Whilst looking at data sheets it seems that increasing air flow at some point starts to have insignificant effect on increasing heat transfer away from the heat sink. Curves have a reverse exponential characteristic.

Knowing squat about the physics, it would appear to me that the radiation of heat by the metal is probably a constant and once you have sufficient air flow to immediately remove any heated air, increasing the flow of ambient temperature air will have little effect

The important factor at play here is actually conduction rather than radiation. Radiation can be a significant mode of heat removal from heat sinks in still air, but it becomes much less important with forced cooling.

Heat is transferred from the heat sink surface into the moving air stream by conduction across a "boundary layer" of air touching the metal surface. The coefficient of heat transfer here (the "film coefficient") increases without limit as the air flow increases. (Heat transfer coefficients are like conductances, they are the inverse of resistances.)

Now the cooling of a device requires the heat to follow a path of several thermal resistances in series. There is from the die to the case, from the case to the heatsink, through the heatsink from mounting point to cooling surfaces, and from the heatsink surface to the surrounding air.

As with electrical resistances, when there are several thermal resistances in series the larger resistances become limiting. So once the heat transfer coefficient between the heatsink surface and the surrounding air becomes good enough then other thermal resistances in the system start to dominate and put an upper bound on the maximum heat flow possible.

[ciccio]

Years ago I designed a range of very high power audio amplifiers, intended to be rack mounted.
The air flow was from front panel to rear panel (air was flowing inside the rack) because we did not want to have an hot front panel.
We tested, with the same fan, both solutions (blowing-in and sucking-out) , and the result was that blowing-in was better, for the following  reasons:
1) the cooling channel was cooler (about 5 degrees lower), because cool air is denser than hot air, and what cools is air mass, not air volume.
2) the fan temperature was lower, increasing fan's operating life
3) a fan spinning in hot air seem to increase her speed, resulting in higher noise.

The problem was that  blowing-in  required mounting the fan just inside  the front panel, and panel aesthetic was compromised (there was no space for an internal tunnel, and the  fan was almost visible) .
Another problem was that the fan's noise was coming from the front panel,  and this was objectionable.

We opted for the less than optimum solution: an higher speed fan sucking air out from the rear panel (noise was inside the rack, and less audible, and panel design was simpler and more elegant).

But the industrial version of these amplifiers, where aesthetics was less important, was designed the correct way.

[AlfBaz]

As with electrical resistances, when there are several thermal resistances in series the larger resistances become limiting. So once the heat transfer coefficient between the heatsink surface and the surrounding air becomes good enough then other thermal resistances in the system start to dominate and put an upper bound on the maximum heat flow possible.

Now that makes perfect sense, thank you Ian
@ciccio Thanks for sharing some good points to consider

[madires]

I'd say it depends on the design of the air flow though the box. If the box has some venting holes in the front or in the side panels mount the heatsink with the fan blowing out of the box. It doesn't matter much if the fan is located at the end or the beginning of the heatsink tube unless there's a considerable difference of the temperature which might stress a fan at the end of the headsink more. The air stream will also cool stuff whithin the stream from the venting holes towards the heatsink. But the air stream won't be distributed equaly. If you put the box into an industrial environment and prefer a closed box and a dust filter add a second fan behind the dust filter to suck air in. When you need more reliabilty put a fan at each side of the heatsink or stack two.