Restart jobs (still) running slower

I am having similar problems to this guy (, in that simulations started from restart files go much slower than those started from the .data files generated by restart2data.

I have attached the lammps input file and an example (pre-minimised) data file of the coarse-grained liquid toluene system I am working on. The thermostats, bariostats, and neighbor list settings are set in the input file, so Im not sure what other settings are being lost in going from restart to data to explain the speedup. For reference, using read_data instead of read_restart roughly doubles the average value of spcpu for the whole simulation, using the march '12 lammps release.

Any help in understanding how this happens would be greatly appreciated.


BER_premin.dat (429 KB)

BER_MD.lammps (1.74 KB)

Can you post the log files for the 2 runs
that have the timing breakdowns?