Re: difference in energy in amber6 and amber7

From: David A. Case <>
Date: Wed 11 Dec 2002 07:40:48 -0800

On Mon, Dec 09, 2002, Prabal Maiti wrote:

> When I use amber7 on to continue my dynamics runs I get different energy
> compared to amber6 runs. Can someone suggest which parameters (default
> etc) needs to be specified to get exactly same energy. The inital energy
> is same in both amber6 and amber7. But during the dynamics energy differs.

See point (3) on p. 4 of the Users' Manual. Basically, with ntt=0 you
should see no differences over short trajectories; for example, the dhfr
and jac benchmarks give identical results for amber6 and amber7.

> Also how accurate is PME. In amebr6 run I had nfft1=45 and in amber7
> nfft1=48 (it does not accept odd number) and the electrostatics energy is
> different by 100 kcal/mol though the change is less than 1%.

I gather this is a different problem(?), since the example you quoted in
your mail had identical electrostatic energies.

Sander 7 has instituted a faster (but equivalent) way of doing the reciprocal
sum. A side effect of this was to disallow odd values for NFFT1 (for
convenience in coding, no fundamental reason.) You could relax this by
defining -DnoRCFFT in your MACHINE file and re-compiling. Mike Crowley may
also know some other work-around (if I remember correctly, NFFT2 and NFFT3
can be odd...)

I agree that a 1% change seems pretty big, but I haven't tried test cases
myself. Take one of the benchmarks or test cases (say dhfr) and experiment
with changing the grid spacing slightly. Again, maybe others will have good


David A. Case                     |  e-mail:
Dept. of Molecular Biology, TPC15 |  fax:          +1-858-784-8896
The Scripps Research Institute    |  phone:        +1-858-784-9768
10550 N. Torrey Pines Rd.         |  home page:                   
La Jolla CA 92037  USA            |
Received on Wed Dec 11 2002 - 07:40:48 PST
Custom Search