MD simulations

We found a serious problem on MD simulations.

JOB 1 MD on protein for 1.5 ns.

JOB 2 MD on the same protein for 1.0 ns and 0.5 ns (using the 1.0 ns final conformation).

Comparing JOB 1 and JOB 2, we found the results from 0 to 1.0 ns are the same, but the results from 1.0 to 1.5 ns are different.

We use the same parameters and random number seed.

Is it supposed so?