[kkpal]

Dear all,
I tried to 2 cases in openfoam at the same time, both of them in parallel with 2 processors. I've found that the simulation speed is severely slowed down, compared to running just one case at a time. These two cases surely did not reach the capacity limit of this computer since the computer has 24 processors and 126Gb ram. I suspected that this problem may have dealings with something inside the MPI.
In view of the above finding, I will either have to run my simulation cases, which is quite a lot , with one processors at the same time, or just run one simulation in parallel and then proceed to another. Both of these two choices are not efficient.
Could someone confirm this? And is there a way to resolve this problem and use openfoam more efficiently?

[cutter]

Hi,

this is not unusual. I don't know your mesh size etc. but the performace is probably limited by memory bandwidth and caches sizes. The fields of
large meshes cannot be kept in cache (even at your really nice machine), they need to be fetched from main memory repeatedly. This problem gets even worse when the cache and bus is used by multiple simulations at the same time.

Cutter

[wyldckat]

Quick answer: From my blog post Notes about running OpenFOAM in parallel - I have a note about this topic... this post details how to handle those situations: http://www.cfd-online.com/Forums/har...tml#post356954 - post #9