I'm attaching a fan to the RPi's original case. I wanted to use a laptop cpu fan. Should I suck the hot air from RPi? or blow cold air inside the case? Btw, there's a heatsink on the Soc and LAN & USB chip.

  • 8
    Just in case there's any confusion: If you want to do this, that's fine, but unless there's some peculiar reason why you need to do this based on observed temperature of the SoC's internal sensor, beware it probably won't make any real difference to the Pi itself.
    – goldilocks
    Jul 10, 2016 at 19:34

6 Answers 6


It doesn't matter. Either way you are moving air one place to the other. Cooling is about volume of air, not the direction of travel (in the sense of "push into" vs. "pull out of", ie. the direction the fan faces).

In theory, pulling air out may create negative pressure inside the case which, in theory, may be advantageous to cooling, however, in reality this is not going to happen unless your fan is a jet engine, which will probably have very detrimental side effects on the case and things within.

A more pragmatic concern on desktop cases is accumulation of dust inside because of all the moving air. However, by pushing the air in instead of out you can keep the interior clean by including a dust filter on the outside of the fan -- then you just have to vacuum the dust filter occasionally. Trying to use filters the other way around, unless you have some kind of special case, is impossible; if you are blowing air out you will end up sucking it in from some arbitrary crack, gap, or hole and dust will coat components in between there and the output fan.

With a pi sized case dust accumulation won't add up to much hassle, of course.

  • 7
    On the contrary, I have had excellent success cooling my Pi with a small jet engine. :)
    – cat
    Jul 11, 2016 at 1:02
  • 1
    I'm confused ... You want to create Positive pressure (as in pressure higher than outside of the box) to increase convection right ?
    – Antzi
    Jul 11, 2016 at 1:29
  • 3
    You always want to blow air into a case, not suck air out of it. You are relying on a pressure differential to close the flow loop through the case, and the fan can raise pressure more than it can drop it. Jul 11, 2016 at 6:51
  • 1
    @Luaan Agreed. You can also get into trouble with a big case if you have too many fans trying to pull air out of the case and not enough pushing air into it. Pushing sufficient air into the case is critical. Jul 11, 2016 at 17:35
  • 1
    It doesn't matter. Either way you are moving air one place to the other. Cooling is about volume of air, not the direction of travel. Hardly. Not only do you have to deal with basic physics like hot-air rising, but the fact that most *PI cases are small and usually have limited airflow and few holes further complicates the matter. They're definitely not equivalent.
    – Synetech
    Oct 22, 2021 at 2:48

The answer is: You should direct the cold air stream directly onto the heatsink.

There are usually arrows stamped on the fan casing that shows the direction of blades rotation and the direction of air flowing through the fan when it is turned on. Install the fan such that the air flows directly onto the hot surface.

Moreover, in every similar situation (if you have a stream of fluid that you want to use for cooling of a hot surface), make sure you are directing the fastest part of the stream right onto the hottest spot. That would give the best possible result.

Here is why:

There are three general mechanisms of heat transfer:

  1. Heat Conduction - when the energy is transferred by collisions between particles (molecules, atoms);
  2. Heat Convection - when the energy is transferred from one place to another by using the moving fluid (air, water etc) as a transport environment; There are two types of convection - free and forced.
  3. Heat Radiation - when the energy is transferred by emitting or absorption of electromagnetic waves;

When we want to transfer the heat from a hot surface (like the heat sink surface, or the IC chip surface) to the surrounding air with the use of airflow, we are creating a forced convection condition. That is, the air stream 'takes' the molecules of air above the hot surface and move them out into external space.

In general case, the effectiveness of such heat transfer mechanism depends on the speed of the air flow that 'takes' the hot molecules and throws them away, because, the more of hot molecules we can remove from the above of surface, the cooler it will be.

We need to analyze the velocity distribution of the air streams that can be generated in different scenarios of fan placement. For that, let's name the fan sides first: the 'back' of the fan is the side where the air is sucked in. And the 'front' of the fan is the side where the air is blown out. Also, let's assume that there is a heat sink attached to the IC.

Now, let's compare the following two cases.

For a scenario, where fan is attached to the hot surface with its back (so that it is sucking the air onto the hot surface from the sides of the heat sink, perpendicular to the fan axis and blowing out the sucked air into external space through its front), the velocity of the air flow near the hot surface will be slower than the velocity of the output stream that is blown away at the front of the fan.

For a fan that is attached to the hot surface with its front side (so that it is sucking the air through its back, then throws it down directly onto the hot surface and then the air is blown out through the sides of the heat sink), the maximum flow velocity would be right at the hottest spot giving the optimal conditions for the cooling.


I tested this with non-original acrylic case and it seems there are slightly better results with pushing cold air into the box. See this video with graph: https://youtu.be/N6keyV-gOzQ


IBM and HP did a lot of research on this topic in the 60s and 70s. Blowing air in cools the hottest part best, but sends that heat to other parts thereby heating them up and lowering the overall MTBF. Sucking avoids this problem provided the fan is mounted by the hottest parts, but doesn't cool quite as well.

And dust accumulates either way. It always will when you have cold surfaces with an airflow.


As the hot air tends to get stuck inside the case i would guess sucking it out is more effective. Either way, you should make sure there's enough holes besides the fan for air to circulate.

  • 4
    The "getting stuck" problem is about air not circulating to the entire case; depending on case design and fan location, this could occur with negative pressure, positive pressure, or both.
    – Mark
    Jul 11, 2016 at 1:28

Look at the configuration of desktop PC cases - they all have fans that suck air out, and some also have fans that blow air in.

Case fans are always at the back, blowing air out; PSUs blow air out; graphics cards blow air out.

Some rackmount servers have fans in the middle - acting to both blow and suck at the same time!

The trick is good airflow (which is also why some have 2 sets of fans) - if you can seal the rest of the case, and provide a good path for air to be brought in (usually at the front) and blown out (usually at the back) then you get the best cooling for the least amount of air movement. If you can restrict the air intake to a certain area, then you can also put dust filters in place.

You can see this in thin 1U rackmount servers, where the fans are tiny and designed to take air in the front and blow it directly out the back over whatever components are carefully positioned in the flow.

  • "they all have fans that suck air out" -> Very often, but this is a bad design that continues because of the inertia of tradition; see my comments about dust and creating positive pressure by blowing a greater volume of air in than you suck out. As David Schwartz observes there in a comment, this is also more likely to ensure there are no dead spots anywhere. I.e., if you use multiple fans, the biggest one should be the input. Trying to "seal the rest of the case" is a fool's errand.
    – goldilocks
    Jul 11, 2016 at 12:28
  • 1
    Back when I was building PCs, we were advised that we should draw the air out of the top of the case, and push it in the bottom, so we didn't have to fight against heat rising to the top, so that PSU fans would tend to push out, simply because they usually ended up at the top in almost all of the case designs available.
    – SeanC
    Jul 11, 2016 at 13:17
  • @SeanC I think this is why traditionally cases come with one small fan that blows out -- because it's mounted at the top on a small grate, likewise with the PSU (additionally, you wouldn't want a PSU blowing in because that air will be heated by the PSU). Then the cases come with a big open grate at the bottom >_< The best thing to do with something like that would be to add a second fan on that bottom grate (w/ outer dust filter) that has a significantly higher CFM rating than the top one. Of course we've gotten way off topic here WRT the Rpi...one fan is overkill, never mind two.
    – goldilocks
    Jul 11, 2016 at 13:29
  • @goldilocks Obviously not hermetically sealed, but look at the server cases - air flows in the front and out the back in a clear path. Anyway, for the Pi, surely a fan is either: way too big for the case or way too small and thus noisy. Maybe the best solution is a better way to let heat provide the cooling airflow naturally - by allowing it to flow over or through the case under its own convection.
    – gbjbaanb
    Jul 11, 2016 at 15:12
  • 1
    @goldilocks look at the link - its a image of an empty rackmount case. As for grooves on heatsinks - if the fins are horizontal, there's nowhere for dust to accumulate (ie stick out to the sides to air can rise between them). My case at home has a dust filter at the lower front, and the usual gfx, psu and case fans blowing out the back - there is very little dust inside but the dust filter gets covered in gunk, so I think the important aspect is the filter, without that you're going to be sucking dust in regardless of the in/out fan configuration.
    – gbjbaanb
    Jul 11, 2016 at 15:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.