[gmx-users] Simulations becoming slower with time

[Erik Lindahl]

David wrote:

>Hi
>
>I'm running a series of simulations on 4 processors. It's a big system
>(19200 atoms) with PME. Since the queueing system limits me to 1 day
>jobs, I cut the simulation in pieces of 250 ps. The first six
>simulations finish within a day, the seventh however doesn't anymore
>(reproducibly). See below for timings. The amount of operations
>(according to the log files) does not increase, files are equally
>long... Density is constant too. Has anyone noticed something similar? 
>
>Simulation	Run time (s)
>0  		61873.000 
>1  		70679.000 
>2  		76894.000 
>3  		77988.000 
>4  		79592.000 
>5  		83272.000 
>6  		84194.000 
>
>  
>
Hi David,

Could it be water sorting?

Cheers,

Erik