[gmx-users] MD simulation on 1 and 4 nodes

[Oliver Beckstein]

On Wed, 26 Mar 2003, Alexander Zuev wrote:

> the results of my run with 10 ps MD simulation on 1 and 4 nodes are
> differed from each other:

As I understand it, MD (or rather the numerical integration) is a chaotic
dynamical system in the sense that small initial differences lead to
quickly diverging trajectories. So it is quite natural to find slightly
different coordinates even after a few steps, or when running on different
is good enough, then _averages_ should be the same.

> 1) from gro-file after mdrun with np=1:
>     2GLU      N   16   1.171   0.770   0.983  0.1916 -0.1159 -0.3137
>     2GLU      H   17   1.143   0.813   0.897 -1.2392 -0.6945 -0.1708
[snip]

> 
> 2) from gro-file after mdrun with np=4:
>     2GLU      N   16   1.141   0.738   0.989  0.1650 -0.2755 -0.3296
>     2GLU      H   17   1.096   0.790   0.917  0.0641  0.4210  0.2319

[snip]

> My version of Gromacs is 3.0.5 with single precision.
> What could be the problem?

Doesn't look like a problem so far (unless your protein is doing very 
gross things on one run and not on the other).

Oliver

-- 
Oliver Beckstein * oliver at bioch.ox.ac.uk
 http://indigo1.biop.ox.ac.uk/oliver/