On Nov 26, 2011, at 5:40 PM, Sasha Perkins wrote:
> Hello,
>
> I am a new Amber user and I was wondering if it's at all possible to reduce
> computation time by changing the nrespa variable (or some form of multiple
> time stepping) when running an MD production run of 4ns with constant
> pressure using periodic boundary conditions of 600 small molecules using
> Amber 11. My input file looks as follows, however, when using sander it
> tells me nrespa must be 1 if ntp>0. However, using nrespa=1 will not help
> me further reduce computation time. Is nrespa>1 only used with NVT or no
> periodic boundary conditions?
>
> &cntrl
> imin = 0, irest=1, ntx=5,
> ntb=2, ntp=1, taup=1.0,
> pres0= 1.0,
> ig=-1
> cut = 20.0,
In addition to what Bill said about the multiple image issue, cut==20 vastly increases the computational cost of a PME calculation by driving up the cost of the direct space calculation (which is already typically the most expensive part of a PME calculation even at a cutoff of 8Å in my experience).
You can safely remove this variable and leave it at the default of 8Å, which will make negligible difference to the VDW calculation (since those fall off as 1/r^6). This should significantly increase your computation time.
Furthermore, I don't see any variable that you've set in your input file that doesn't work for pmemd, which is significantly faster than sander for any functionality that they both share. Thus, you will see a *significant* speedup over your current simulation by reducing cut to 8 and switching to pmemd instead of sander (the results are exactly the same for the same input between the two programs) without significantly changing the results of your simulation.
To answer your original question, though, nrespa > 1 with constant pressure simulations (and minimizations) is disabled in both sander and pmemd, so in order to get nrespa > 1 working, you'll need to switch to NVT or some implicit solvent simulation.
> ntc=2, ntf=2,
> temp0=300.0,
> ntt=3, gamma_ln=5,
> nstlim= 2000000, dt =0.002,
> ntpr=100, ntwx= 100, ntwr = 1000,
> nrespa=2
> &end
HTH,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Nov 27 2011 - 02:30:02 PST
