[gmx-users] System volume "jumps" on exact continuations

[Elizabeth Ploetz]

1. Once you identify a continuation (with associated run script) that gives the discontinuity, if you run many repeats of the original continuation then does the jump always occur or only sometimes?

The jumping does not always occur. See the linked figure<http://s1243.photobucket.com/user/ploetz/media/repeats_zpst3gbe9du.png.html>. The data is the same for 0-10ns in all three rows. The top and middle rows show two 1-ns exact continuations after the 0-10ns run. This shows that the jumps don't always happen, and are not always the same. The bottom row shows 5 exact continuations all starting from 10ns. It looks like only one of them jumped (yellow). The commands and input files were the same for all of the runs. We did NOT use -notunepme for these runs.

2. Did you do any of the MPI runs with -notunepme ? That would be my first suspect.

We have not used -notunepme in any of our past simulations. We are doing some tests of this now, based upon your suggestion. So far, we don't see jumps when using this flag.

3. Did you try explicitly setting a relatively large rlist and set verlet-buffer-tolerance = -1 (both in the .mdp file)? That would be my second suspect, though you say you've never seen the problem with verlet cutoff-scheme so I guess this suggestion is not related to the underlying issue.

So far we have not seen any jumps using Verlet. However, we have significantly fewer Verlet simulations. For many of our projects, we have not yet switched over to Verlet. It is much slower for us (we have tons of water in our systems).