CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Computation Time compared to OpenFOAM (http://www.cfd-online.com/Forums/cfx/24188-computation-time-compared-openfoam.html)

Florian Fruth June 28, 2007 12:05

Computation Time compared to OpenFOAM
 
Hey all, has anyone of you ever made a comparison of the computation times between cfx and openFOAM? I took a forward facing step supersonic example and computed it at first laminar. It seems like openFOAM is 10-20 times faster than cfx. Can one tune cfx somehow (e.g. leave away all unnecessary outputs) that it is faster? Or is the difference too big, such that the difference must be due to a wrong input of myself?

Regards Florian

opaque June 28, 2007 12:47

Re: Computation Time compared to OpenFOAM
 
Dear Florian,

Would you mind sharing the mesh size you compared? Was it 10^4, 10^5 or > 10^6 elements?

Did you compare serial runs, or parallel runs? If your comparisons were in parallel, what was the mesh size (ballpark number is OK) on each the partitions?

Regards,

Opaque

Florian Fruth June 28, 2007 12:58

Re: Computation Time compared to OpenFOAM
 
Sorry for the inaccurate question. So the mesh size is about 5000 elements, the run is in serial mode. The rms is set to be under 1e-05. It is a transient example, where I took a deltaT of 0.002s and computed 10s in reality. With openFOAM it took about 200 seconds to compute the result, whereas with cfx it would have taken about 5-6 hours!? I tried to take the same numerical schemes, but there are big differences and I tried to take approximately the same.

Cheers Florian

opaque June 28, 2007 13:08

Re: Computation Time compared to OpenFOAM
 
Dear Florian,

Thanks for the information.. Do both codes use the same number of coefficient loops per time step?

Comparisons like this are useful, but must not be taken out of context. How will they compare for more complex physics, or larger meshes?

Try running CFX for largers meshes and see how the wall clock time scales.. Then, pick one of those meshes and see how OpenFoam compares.. I would look at meshes larger than 100 000

Keep us posted..

Opaque


Dave June 29, 2007 10:18

Re: Computation Time compared to OpenFOAM
 
What were you using that took 5-6 hours to run a 5000 element case (albeit a transient one)? It sounds like something is very wrong with your set-up, or you're using a computer that was built sometime in the 1980s.


All times are GMT -4. The time now is 01:37.