[Beowulf] Re: torus versus (fat) tree topologies

[Chris Sideroff]

Greg Lindahl wrote:
> In short, you don't want to ever make any any generalization about CFD
> as a whole. You need to specify the solution technique and data size
> at a minimum.

   Yes and the problem is compounded even more when you do not have the
source code.  And as I mentioned in another reply it is likely that
fluent's number one goal is not speed but robustness, stability and
general usability.  By including the ability to use all the models that
they do _has_ to degrade the performance as well as having very
different performance characteristics than specific (in-house) CFD codes do.

Chris SIderoff