# How to design a cooling chimney for a computer server?

I'm building a server type computer consisting of integrated motherboard, CPU, RAM and hard disk. There will also be a standard power supply unit. This is to be a Steam Punked computer, so...

I want to know the feasibility of passively cooling the server using only a chimney effect. So the electronics would be enclosed, with a chimney at the top and air intake(s) at the bottom. All the components will be custom mounted so there is no typical computer case and therefore no particular size constraint. How would I begin to evaluate /design this? I believe that there are relationships between chimney proportions, temperatures and exhaust velocities. Perhaps like this...

Mini computers (ARM /Celeron) exist that have no fans, so perhaps it isn't really a big deal and anything at all will work?

• A few websites that may be of interest: solar chimney, stack effect & natural draft/draught calculator. The main considerations will be the temperature difference between the bottom & top of the chimney, the length of the chimney, air density which will be affected by humidity & an adequate cross-sectional area of the chimney to allow enough air to pass through. – Fred Feb 20 '17 at 1:18
• The critical requirement is getting enough airflow in the right place - for example directly over the heat sink of the CPU. Cooling air circulating over the whole board won't be enough. You could consider a (large and passive) heat-pipe cooler to move the heat from the CPU to a place where the chimney can deal with it. – alephzero Feb 20 '17 at 6:11
• I agree with @alphazero on the idea that you wont get cooling in the areas you need. I'm also somewhat skeptical about the effectiveness of a passive heat sink though. You would want to make sure there is still enough of a temperature difference to promote a lot of natural convection, or else then the entire plan goes out the window. – JMac Feb 21 '17 at 18:55
• @JMac I expect passive cooling to be no problem at all in principle. A computer hardly puts out any heat. Lap tops don't run fans often so passive cooling should easily handle that averaged out. Micro computers have no fans at all, and 10MW transformers are cooled via radiators. It's probably only 75W or so anyway. – Paul Uszak Feb 21 '17 at 22:00

The key technical term you're looking for here is convective cooling. Either you allow the air to rise through the chimney on its own accord (natural convection cooling) or you pump it through with a fan (forced convection cooling). The second is obviously more effective at pulling heat out of the system.

The amount of heat removed will be a factor of the flow rate of air through the system as well as the heat transfer coefficient (HTC) between the air and the CPU. You can get a good estimate by grabbing an HTC function from engineering toolbox or a text book and using Newton's Law of Cooling:

q = hc A dT


where q is heat transfer (in watts), hc is the heat transfer coefficient (W/m^2 K), A is the surface area to be cooled (m^2), and dT is the temperature difference between the air and the surface (in K).

The HTC for air is (for v between 2-20 m/s):

hc = 10.45 - v + 10 v^1/2


where v is your air flow speed in m/s. The faster the air is moving, the higher your HTC (and the more heat you can pull out of the system).

Broad estimates of HTCs:
Free convection: 0.5-1000 W/(m^2 K)
Forced convection: 10-1000 W/(m^2 K)


You should know from the specs of the machine roughly how much heat is produced in watts. Then your air flow will dictate your HTC - and try to maximize it so that you can pull more heat out of the system.

The concept of convective cooling is pretty widely discussed; any general engineering thermodynamics textbook will have more detail ("Thermodynamics: An Engineering Approach" by Cengel or "Fundamentals of Heat and Mass Transfer" by Incropera are good resources). There's quite a lot of stuff about general engineering approaches online, too: where to locate the object to be cooled in your enclosure, what sort of fans to use, etc. These are the sorts of problems that can be analyzed at a really detailed level with computational fluid dynamics. To check the feasibility, I would start with Newton's Law of Cooling, though.

• I recognise HTC, but I also thought that the length of the chimney was important. Doesn't it stop working if the air has time to cool before exiting at the top? So a thin tin chimney might be worse than a thick woolly one? – Paul Uszak Feb 21 '17 at 15:22
• @PaulUszak On the scale you're dealing with, the length of the stack is less important. The main motivator for insulating chimneys and stove pipes is fire safety. – Air Feb 21 '17 at 16:55
• @PaulUszak - sure, the length does matter (as does the geometry and the material) but the dominant factors here are your air flow & heat source. – phyllis diller Feb 21 '17 at 17:07
• @Air Not completely true. The pressure difference will be directly proportional to the average temperature difference inside and outside the chimney, times its height. The taller you make a chimney and the better you isolate it, the more suction power you will get. – HelloGoodbye Jun 25 '19 at 7:18
• – HelloGoodbye Jun 25 '19 at 14:26