Re: [AMBER] Random seed reproducibility

From: Daniel Roe <>
Date: Thu, 29 Mar 2018 09:17:41 -0400


You're using a different GPU in the second case (GTX 1060 vs GTX 1080
in the first case). Try to reproduce the results on one or the other
(i.e. run the same calculation on the same GPU twice) - they should be
the same. If they aren't, that could indicate you have a faulty GPU
(at least from a computational standpoint).


On Thu, Mar 29, 2018 at 4:32 AM, Sergey Samsonov
<> wrote:
> Dear Amber users,
> I have carried out a 1 ns long simulation printing the output file every
> 5000 steps (ntpr=5000), but then I realized I needed the energy values
> every 500 steps. So I took the same random seed written in the original
> md.out file (attached md6-original.out : Note: ig = -1. Setting random
> seed to 454827 based on wallclock time in microseconds. ), and started
> the simulation again just changing the output frequency (attached
> md6-repeated.out). Both simulations started with the same initial
> conditions (ntx=7) obtained from the previous simulation. Unfortunately,
> I still don't get identical results in the outputs and in the
> trajectories (see the output files attached, take a look at NSTEP = 5000).
> What can be the reason for
> this, and how can I get an identical trajectory with energy values
> printed more often?
> Thank you and best regards,
> Sergey
> _______________________________________________
> AMBER mailing list

Daniel R. Roe
Laboratory of Computational Biology
National Institutes of Health, NHLBI
5635 Fishers Ln, Rm T900
Rockville MD, 20852
AMBER mailing list
Received on Thu Mar 29 2018 - 06:30:02 PDT
Custom Search