Re: NBFIX and cuda

From: Axel Kohlmeyer (
Date: Wed Dec 22 2010 - 10:51:19 CST


On Wed, Dec 22, 2010 at 4:13 AM, Jacopo Sgrignani <> wrote:
> Dear all
> i would like to use amber topologies to run namd with cuda but i got the
> error:
> CUDA-accelerated NAMD does not support NBFIX terms in parameter file.

> I found some messages in the list about the meaning of the NBFIX terms, but
> i was not able to understand two questions:
>  can i use the amber topologies ?

probably not. the CUDA implementation obviously uses some very
fast memory on the GPU to store the per atom type non-bonded
parameters and then generates and mixed terms on the fly.
this collides with NBFIX terms in CHARMM parameter files where
those mixed parameters are explicitly set and override the ones
generated from the mixing rules. obviously amber precomputes the
mixed terms and stores it into the prmtop file and thus is incompatible
with GPU acceleration.

>  how can i do it?

since you already have amber input, why don't you try the GPU
accelerated version of amber's pmemd instead?

> I did not find NBFIX terms in the prmtop file, so i'm a little bit confused.

there are no explicit NBFIX terms in the prmtop file, but you have
to remember that NAMD first and foremost assumes that somebody
is using a CHARMM forcefield.


> Thanks
> Jacopo
> ----------------------------------------------------------------
>  SISSA Webmail
>  Powered by Horde

Dr. Axel Kohlmeyer
Institute for Computational Molecular Science
Temple University, Philadelphia PA, USA.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:53 CST