[gmx-users] restarting runs

[Brian Stephenson]

Hi all,

I've been running lipid simulations recently and have been tracking 
equilibration by using g_angle to determine the
frequency of dihedral transitions.  As would be expected, this number 
decreases gradually as equilibration occurs.
I just noticed, however, that every time I restart a simulation the number 
of transitions jumps way up to almost its
initial, unequilibrated value.  What's more, the decay time for the 
transition frequency is roughly the same as the
original decay time.

I've  noticed things like this using GROMACS before--when I restart runs 
using my old output as my new starting point,
the initial behavior seems unequilibrated.

What could be going on here?  At first I thought it might be because I was 
using gen_vel = yes and was reinitializing
velocities, but this same effect occurs even with gen_vel = no.  I'm 
running in single precision--would double help
with this sort of thing?  Do I need to increase the sig figs. on my 
coordinate and velocity output somehow?

By the way, I've been using:

nstxout = 1000
nstvout = 1000
nsfout = 0
nstlog = 1000
nstenergy = 1000
nstxtcout = 0 (I figured I didn't need this)
xtc_precision = 1000

Any suggestions will be much appreciated!

Thanks,
Brian