[gmx-users] Inconsistent results in different clusters and cores

Justin Lemkul jalemkul at vt.edu
Sun May 12 19:41:09 CEST 2013

On 5/12/13 1:34 PM, tarak karmakar wrote:
> Thanks Justin for the Quick and Helpful reply.
>     Yes. If I am right, the chaotic behavior of the simulations is inherent
> and can be assessed statistically by generating several independent
> trajectories and analyzing their similar outcomes. But with the same '.mdp'
> file I am getting TOO much different results, and that's where I worry.
> I'll surely try with the recent version of gromacs. But, for now, can you
> give me a little more informations about problems (bugs) with the 4.5.5
> version, related to my context?

The proposed relationship to bug 1012 is unclear to me.  The issue there was an 
incompatibility between an integrator and thermostat, with obvious differences 
in thermodynamic output.  Assessing your system in this context is not helpful. 
  If you want to assess whether different core counts or hardware produce 
problems, then you need a very simple test case (like a box of water) that shows 
significant differences in averaged observables.  As I said before, maybe your 
ligand parameters are insufficiently accurate (how did you generate them?) or 
.mdp settings are incorrect.  Without such information, there is little point in 
trying to debug anything.



Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list