[gmx-users] Simulations becoming slower with time

[Marcin Wojciechowski]

Yes, I had the same problem with my system, here the description:

1. Size: 53683 atoms (17014 water residues and one protein neutralized
with 20 chlorine ions)

2. Box dimensions: 8.60200   7.97100   7.83500

3. PME, 2 fs step and 500ps simulation, t and p coupled, pbc on.

As I start the simulation everything is fine, then after maybe 5000 steps
the finish time start to increase at constant rate. Now this is
interesting: when you terminate the simulation and start from the point
you have terminated (tpbconv), situation is the same (begining ok, later -
incease finish time). I checked energy charts from previous runs -
simulation is very stable with acceptable fluctuations of all energy
components. 

Could it be somehow related to increasing size of trr file? The big system
with frequent coordinate saving could eat lot of cpu power. I save mine
every 500 steps (every 1 ps).

Cheers, Martin.