Re: [AMBER] repeat amber run on cuda

From: Daniel Roe <>
Date: Fri, 5 Jul 2013 14:45:44 -0600


On Fri, Jul 5, 2013 at 2:35 PM, Kanin Wichapong
<> wrote:
> I know that repeat MD runs will not get the same exactly results, however,
> it should be somehow similar, especially when you use the same computer and
> same MD parameters, right? So, I just wonder how can I believe in the

Not necessarily; to get an exact reproduction of your run you also
need to use the same random seed (ig) if you're using a Langevin
thermostat and/or you're generating an initial velocity distribution
(irest=0). If you're not using the same random seed/initial
coformation/intial velocities etc. you should eventually get similar
conformational ensembles between your two runs; the key word is
"eventually", and how fast your runs converge depend on a lot of
different factors. If you set ig to -1 you can see what value ig
actually gets set to during the run in the mdout file. If your mdin
(including random seed), coordinates, and velocities are identical
then you should get the same run on the same GPU (and if you don't
there could be a problem with your GPU - there has been a lot of
discussion regarding this recently on the mailing list)...


Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 201
Salt Lake City, UT 84112-5820
(801) 587-9652
(801) 585-9119 (Fax)
AMBER mailing list
Received on Fri Jul 05 2013 - 14:00:02 PDT
Custom Search