Effect of the number of processors on the Residuls of pressure

miladrakhsha · September 1, 2014, 01:31

Hello everyone,

Today I encountered a very strange thing in OpenFoam. I will explain this issue here since I could not find any similar question in this forum.

The case that I am working on has ~80,000 cells. I have a machine with 32 processors. I once used 28 CPUs to solve the problem in parallel. The other time I used 30 processors and that is the "only" difference between these cases. Both cases are decomposed using scotch method.

I have attached the result of Drag coefficient and Residuals for both cases.
The strange thing is that the residual of the pressure is not at all the same for these cases which confused me

. However, results of a physical parameter such as lift coefficient is almost the same for both cases

.

I was wondering if someone can explain the effect of number of CPUs on the residuals.

Thank you
Milad
28CPU.png

Residuals28CPU.png

Lift.png

Residuals.png

wc34071209 · September 2, 2014, 09:09

It seems the residual of pressure is the only difference. I have been wondering how the pressure residual is calculated in OF for a long time. It seems it is quite hard to push the pressure residual down below 1E-3.

miladrakhsha · September 3, 2014, 11:53

Yes the pressure Residual is the only difference between the two graphs I suppose.

Nobody has any comment?

September 1, 2014, 01:31	Effect of the number of processors on the Residuls of pressure	#1
miladrakhsha New Member miladrakhsha Join Date: Aug 2012 Posts: 29 Rep Power: 13	Hello everyone, Today I encountered a very strange thing in OpenFoam. I will explain this issue here since I could not find any similar question in this forum. The case that I am working on has ~80,000 cells. I have a machine with 32 processors. I once used 28 CPUs to solve the problem in parallel. The other time I used 30 processors and that is the "only" difference between these cases. Both cases are decomposed using scotch method. I have attached the result of Drag coefficient and Residuals for both cases. The strange thing is that the residual of the pressure is not at all the same for these cases which confused me. However, results of a physical parameter such as lift coefficient is almost the same for both cases. I was wondering if someone can explain the effect of number of CPUs on the residuals. Thank you Milad 28CPU.png Residuals28CPU.png Lift.png Residuals.png

September 3, 2014, 11:53		#3
miladrakhsha New Member miladrakhsha Join Date: Aug 2012 Posts: 29 Rep Power: 13	Yes the pressure Residual is the only difference between the two graphs I suppose. Nobody has any comment? Last edited by miladrakhsha; September 3, 2014 at 23:42.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Foam::error::PrintStack	almir	OpenFOAM Running, Solving & CFD	91	December 21, 2022 04:50
AMI speed performance	danny123	OpenFOAM	21	October 24, 2020 04:13
Pressure Outlet Guage pressure	Mohsin	FLUENT	36	April 29, 2016 17:16
[blockMesh] --> foam fatal error:	lillo763	OpenFOAM Meshing & Mesh Conversion	0	March 5, 2014 10:27
[Commercial meshers] Trimmed cell and embedded refinement mesh conversion issues	michele	OpenFOAM Meshing & Mesh Conversion	2	July 15, 2005 04:15

September 2, 2014, 09:09		#2
wc34071209 Senior Member Yuehan Join Date: Nov 2012 Posts: 142 Rep Power: 13	It seems the residual of pressure is the only difference. I have been wondering how the pressure residual is calculated in OF for a long time. It seems it is quite hard to push the pressure residual down below 1E-3.