[gmx-users] Load imbalance and Y size error

[Mark Abraham]

nikhil damle wrote:
> Hi all
> 
>          I am trying to simulate a system of 3168 atoms in vacuum in PBC 
> box (-d=0.9 nm). Since there are no solvent molecules, i am unable to 
> neutralise the system using addition of ions (Hence system has +3 
> charge). As soon as i start the MD run after putting the whole system in 
> box, there is imbalanced force of 80% due to which right in the 
> beginning, dynamic load balancing is turned on. But even after this, 
> there is ~22% loss of performance and after certain steps MD crashes 
> saying :-
> 
> "Fatal error: The Y-size of the box (4.618413) times the triclinic skew 
> factor (0.866051) is smaller than the number of DD cells (2) times the 
> smallest allowed cell size(2.000000)"
> 
> Can anyone explain the origin of this error and how to deal with it ?

Try running with one processor to see if it's your system or the domain 
decomposition algorithm that is the problem. DD doesn't necessarily cope 
well with highly heterogeneous distributions of particles.

Mark