This won't necessarily help, but what ntc and ntf settings are you using?
Another often overlooked aspect of SHAKE in AMBER (and minor pet peeve
since it's usually miscited) is that 3 point waters (maybe all rigid
waters?) don't actually use the true SHAKE algorithm as other bonds to
hydrogen, they use the SETTLE algorithm. Now I couldn't tell you in
detail what that means but I would guess that a rigid water with additional
bonds/constraints could muck up the math there.
Regards,
Brian
On Wed, Jan 4, 2012 at 5:53 PM, Ben Roberts <ben.roberts.geek.nz> wrote:
>
> On 4/1/2012, at 5:03 p.m., Ben Roberts wrote:
>
> > Hi,
> >
> > This is apparently the cause of the recent troubles I had when running
> pmemd.CUDA on Longhorn. These were covered in an earlier thread, and
> ultimately appeared as bug #172 on the bugzilla.
> >
> > The upshot of the bug hunting (thanks Scott!) is that it's not a real
> bug at all, but a hydrogen connected to more than one oxygen, or else an
> oxygen connected to more than two hydrogens. This is, apparently, a problem
> for SHAKE, that is caught in the CUDA executable but not in the regular
> executable.
> >
> > However, this surprised me, because I thought I had taken good care of
> my topology. And, indeed, when I checked my topology using rdparm, I found
> that the only hydrogens bonded to more than one atom of any kind were those
> in the rigid TIP3P waters, where there's a known H-H bond.
> >
> > There were a number of oxygens with three substituents. This is because
> the system I'm simulating involves a dinickel cluster, and each nickel has
> bonded to it a water molecule, so that the oxygens in these water molecules
> each have two hydrogen substituents plus a nickel. Also, there's a bridging
> hydroxide, in which the oxygen has two nickel substituents plus a hydrogen.
> >
> > These are the only oxygen atoms with three substituents in the system.
> >
> > Now, this presents a problem to me, because it seems to me that SHAKE
> oughtn't to whinge and moan about too many hydrogens in such cases. Scott
> suggested, though, that the nickels could be wrongly identified as
> hydrogens. If so, how could that be? And is it because of a problem in my
> input, for example, my PREPI file that I used to construct topologies?
> >
> > On the other hand, is there perhaps a section in the PRMTOP that lists,
> either explicitly or as part of a section that serves multiple purposes,
> which atoms should be considered hydrogens?
> >
> > And can I get around this problem (even if just as a temporary
> workaround) by just putting the nickel ions in my system into a NOSHAKEMASK?
>
> Actually, no, I can't. If I try running with the following NOSHAKEMASK:
>
> noshakemask = ':768,769,1540,1541,2312,2313'
>
>
> I get this error (from STDOUT):
>
> TACC: Done.
> TACC: Starting up job 121069
> TACC: Setting up parallel environment for MVAPICH ssh-based mpirun.
> TACC: Setup complete. Running job script.
> TACC: starting parallel tasks...
> cudaGetDeviceCount failed unknown error
> TACC: MPI job exited with code: 1
> TACC: Shutting down parallel environment.
> TACC: Shutdown complete. Exiting.
> TACC: Cleaning up after job: 121069
> TACC: Done.
>
> Any thoughts? Should I file a new ticket on the bugzilla?
>
> Cheers,
> Ben
>
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
================================ Current Address =======================
Brian Radak : BioMaPS
Institute for Quantitative Biology
PhD candidate - York Research Group : Rutgers, The State
University of New Jersey
University of Minnesota - Twin Cities : Center for Integrative
Proteomics Room 308
Graduate Program in Chemical Physics : 174 Frelinghuysen Road,
Department of Chemistry : Piscataway, NJ
08854-8066
radak004.umn.edu :
radakb.biomaps.rutgers.edu
====================================================================
Sorry for the multiple e-mail addresses, just use the institute appropriate
address.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jan 05 2012 - 13:30:03 PST
