Hi Martin,
> These settings with slightly longer cutoff and coarser PME grid would
> fit my system very well. Is there a reference for using this
> combination, as most recommend a 1Å PME grid spacing?
I don't know of any specific references for this that describe things in
detail. You might be able to go back to the original Darden PME paper and
the P3ME paper and work from there.
I think when things were originally implemented the metrics in terms of
performance were pretty simple. FFT was always faster than direct space
(these were the days of serial and limited cache remember) and thus the
optimum setting would be a very FFT grid and a cut off of zero. Of course
that trashes the VDW so I think a value of between 8 and 9 angstroms was
chosen for the direct space cutoff as the smallest you could go before you
started hurting the VDW and then the 1 angstrom grid for the FFT fell out
from here. On newer hardware it can sometimes be beneficial to use a larger
cutoff since direct space calculations can be faster than the FFT (and scale
better).
One thing you can try is using the CPU code run a series of single step MD
simulations with different grid spacing and cutoff and see what that does to
the Ewald error estimate. My understanding is that as you make the FFT grid
coarser so you need to increase the cutoff size to keep the Ewald error
estimate constant. Others who are experts in the PME approach might be able
to shed more light on this.
All the best
Ross
/\
\/
|\oss Walker
---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
|
http://www.rosswalker.co.uk |
http://www.wmd-lab.org/ |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jun 19 2012 - 10:00:03 PDT
