CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   Parralel processing (https://www.cfd-online.com/Forums/fluent/34406-parralel-processing.html)

SAM August 14, 2004 03:50

Parralel processing
 
Hi

1.I need to know that wat type fo hardware is required for parralel processing for fluent on windows xp

2. Is the conventional LAN is enough for parralel processing. Is not that wat is the solution

Any help will be greatly appreciated.

Thanks in advance

Regards Sijal Ahmed Memon

ccc August 14, 2004 10:23

Re: Parralel processing
 
hi, SAM

I think no special types of hardware are needed.

conventional LAN is engough if the communicaion between nodes is not very large.

good luck


SAM August 14, 2004 14:51

Re: Parralel processing
 
If the problem is highly 3D with unsteady option on , then conventional LAN is ok

Jason August 17, 2004 09:32

Re: Parralel processing
 
It will work, but if you have it split too many times then the LAN communication wastes more time then you gain by using multiple processors. I routinely run 3D, turbulent, and compressible jobs across a LAN, and have run several 3D, unsteady jobs as well. It seems to be a trial-and-error process to figure out the optimal number of parallel processes.

I have noticed that once a converged solution is found (by monitoring residuals, force coefficients, and usually the pressure at a couple points) there is a small mass imbalance near where the partitions intersect the flow boundary conditions (like pressure-far-field or pressure-outlet, but not wall or symmetry), but it is small and does not seem to effect the solution, as far as the forces or the pressures on the wall surfaces, but it does seem to drag the convergence out some if you watch the residuals. This may be another reason to try not to run it on too many processors. You can plot this mass imbalance under the plot-contours window.


All times are GMT -4. The time now is 14:12.