[FINISHED] Thermal Analysis of CPU with different Heat Sinks


Status - WIP

Started by @pankajkumar979 and @SairajuPonnoju



The cooling mechanism is very crucial to ensure the rated performance of CPU, at the same time it accounts for the major operating cost of the system. Therefore, an efficient electronics cooling is necessary to provide the continuous operations at low operating costs. Tools of computational Fluid Dynamics are widely being used to optimize the cooling mechanism of the chips. In this study the focus is on finding the optimum heat sink design for electronics cooling.

Input Data

More detailed description and different simulation scenarios are given in the proposed paper [1].

CPU: 3.5mm x 30mm x 30mm
Heat Sink Base Plate: 3.5mm x 54mm x 65mm
Fin height: 40mm
Fin Pitch: 2.5mm
CPU Power rating: 80W
Inlet Flow rate: 30ft^3/min

Fig.1 The plate heat sink in two arrays.


The purpose of this project is to validate the accuracy of the SimScale CHT solver for electronics cooling. Different heat sinks, along with different base plates will be compared in order to find the best heat sink for the CPU cooling. The heat sink temperature difference results are to be compared with proposed experimental result to find out best heat sink designs.

Fig.2 a),b)Temperature distribution for different heat sinks with different base plates.

Key Words

Computational Fluid Dynamics, desktop computer, heat sink, plate and cylindrical Heat Sink, CHT, conjugate heat transfer

Literature & Sources

[1] http://www.ijcim.th.org/past_editions/2010V18N1/Paper-4-p27-36-IJCIM%20Vol%2018%20No1-Jan-Apr-2010.pdf



Hello sir.

I have to design the cad models of different heat sinks to perform the thermal analysis of cpu with different heat sinks?
I want the exact dimensions of fins


Hello @SairajuPonnoju,

some geometric dimensions vary in this example as can be seen in the Paper. Please have a look at it and chose which geometry you want to test on our platform.

I am willing to help you with your simulations :+1:




I have a same problem in my industry but the heat sink in located at PCB of air conditioner want to decrease material requirement of the heat sink at same heat removal rate or it will be better if we increase the heat removal rate.


Hi @mehakkar123!

Have you already posted your query in the Project Support section of the forum? Maybe someone can help you out there!




I have a confusion over fin pitch.
Which one is right?


Hi @pankajkumar979!

The Fin Pitch is what you have described as A in your picture. Hope that clarifies it. Good luck with your simulation! :wink:



I am done with the geometry.


Is that how cpu is to be connected with base plate?


Hi @pankajkumar979,

Yes I believe your configuration is correct as from the sample diagrams there isn’t any additional complexity. The heat output is mostly centred and there doesn’t seem to be any additional methods of heat transfer to the heat sink.

Side note, modern heat sinks typically used in consumer desktops have a wide range of ways to transfer heat, be it heat pipes leading towards the fins or even more complex water blocks in contact with the CPU which will show up completely differently as compared to the reference diagram.

Good luck for your simulation!



Hi @jousefm!

I am trying to run a simulation using CHT code but it’s showing a convergence error, i.e. solution failed to converge after 1000 iterations. Most probably error might be due to insufficient application of all necessary boundary conditions. Cannot figure out the problem, plz help.




Hi @pankajkumar979!

We’re currently looking into it and we’ll get back to you with more details asap. Please understand that escalations can take some time. Getting back to you as soon as we have found a solution.




Hello @pankajkumar979,
There are some recommendations with regarding to your project.

  1. For performing a CHT simulation the solid and fluid regions should be non-overlapping, but the boundaries touching each other. Please perform a boolean operation in the CAD tool to remove the fluid region which overlaps with solid parts.

  2. Make sure to have coinciding faces - the automatic interfaces are created only when the faces in contact are of same area. For instance, in the below image the highlighted face comes in contact with a smaller face of adjacent solid. Hence no interfaces are created since the face areas differ. Hence make sure to split the faces which come in contact.

  3. You might also consider reducing the number of surfaces defined for result control item. The objective might be to find the average temperature value on the face of heat sink. Here may be 3 or 4 faces can be selected to find the values. Since conduction is predominant inside the solids the change in temperature might not be substantial between the faces. On the other hand defining less faces will decrease the computation time.

  4. For a steady-state simulation it might not be necessary to store many instances. Hence the save instances can also be decreased (you could also go with only one result stored at the last instant).

I hope this helps!



Thanks for explaining it clearly in easy steps!

I tried yesterday by making the geometry again. Divided geometry of both insulator and Base plate so that area of cpu matches both, Used boolean operation by subtracting the solids from fluid domain, still getting error.

Error was regarding zero gradient boundary condition imposed on side walls of the geometry. Still i can’t understand, this boundary condition together with drichtlet boundary condition on Top and Bottom walls should work fine.

Any Ideas how could that be?


GAMG: Solving for p_rgh, Initial residual = 0.510620621348, Final residual = 4.30719765053e-06, No Iterations 21
time step continuity errors : sum local = 0.00682235438637, global = -2.18615349433e-05, cumulative = -2.18615349433e-05
Min/max rho:1.17680655709 1.17681793569
Solving for solid region solid_1
[15] Attempt to cast type zeroGradient to type compressible::turbulentTemperatureCoupledBaffleMixed
[15] From function refCast(From&)
[15] in file at line 114.


Hi @pankajkumar979,

I believe the issue with simulation is already sorted out. Just a note from my side -> it is not necessary to define any interfaces for the simulation setup. The mesher automatically detects all the interfaces and this can also be seen at the end of the mesh log.

Consider running the simulation for longer time to obtain steady state solutions.



Hi @sjesu_rajendra,

Is it possible to get some more core hours as i am running out of computing resources for the project because of some mistakes i made?



Hi Pankaj (@pankajkumar979)!

Will contact my colleague @AnnaFless and ask if you can get some. Getting back to you asap :slight_smile:




Hi @sjesu_rajendra

I am getting a slow convergence in almost all cases. Any suggestions or any guide which can help me out with this convergence problem.

Thanks in Advance.



Hi @AsadAli,

Is it possible to get a extension for a few days as i am out of core hours?


@pankajkumar979 - sorry for the delay. I have added more core hours to your account.