[gmx-users] Re: Strange statistics when running parallel Gromacs

[Erik Lindahl]

Dallas Warren wrote:
> Paulo,
> 
>>          Please, I would like to know if someone had a similar 
>> problem  and
>> appreciated any suggestion.
> 
> 
> If you look in the archives for this emailing list, last week I posted 
> some data on a system I am simulating.  It has about 42,000 atoms using 
> PME and it is not worthwhile going past 4 nodes.  Some extra performance 
> was achieved by increasing the PME order and the fourier grid spacing, 
> as noted in the manual.  Also, the larger the system, the better that is 
> parallelises.
> 

That sounds about correct. It will get better, though; we're working on 
a rewrite of the entire parallelization...

Cheers,

Erik