RE: AMBER: Failure to SCF convergence & parallelization and memory

From: Francesco Pietra <chiendarret.yahoo.com>
Date: Sun, 12 Aug 2007 00:37:33 -0700 (PDT)

Hi Ross:
Thanks. Please read below at your convenience.

--- Ross Walker <ross.rosswalker.co.uk> wrote:

> Hi Francesco.
>
> > While carrying out a AM1 QM-MM md2 procedure (restarting from
> > previous md1.out)
> > I noticed casually from tail -f that after 1000 steps
> > (precisely at step 291)
> > SCF was not reached at the tolerances specified. Procedure
> > continues with
> > warning about energy reliability.
>
> This is a function of the self consistent field procedure not being an exact
> science. Sometimes, particularly if you have a 'weird' structure the
> converger can get stuck and not get the energy or density difference below
> the tolerances. If this only happens once or twice you probably don't need
> to worry about it. You might want to write ntwx every step for the first 300
> steps or so and take a look at the structure of step 291 and see if it looks
> strange at all.
>
> The issue with energies and forces is that the accuracy of the forces is a
> function of how well the SCF gets converged. 10^-8 is typically needed for
> energy conservation.

Yes, in md2.out I see among the QMMM options:

scfconv = 0.100E-07 itrmax = 1000


However if the SCF procedure got close then the error
> won't be big and is unlikely to affect your dynamics. Take a look at DeltaE
> and DeltaP that it gives you for the last step of SCF, if it is of the order
> of 10^-7 for deltaE and 10^-5 for DeltaP or so you are probably okay. Also
> see what the smallest deltaE and deltaP are this tells you what was the
> closest it got to converging.

I am unable to detect where such deltaE deltaP are reported in either md2.out
or mdinfo.


>
> You can up the number of SCF steps with the itrmax option in the qmmm
> namelist but typically it won't help you for small systems that occasionally
> show convergence issues.
>
> Simple option is just to ignore it if deltaE is close to 10^-8 and it only
> occurs once or twice.
>
> > md1.out did not reveal problems in the final summary. Are
> > such problems with
> > SCF convergence reported in the final part of md.out?
>
> No they are not, they are only printed to the output file as they happen. I
> could potentially add a Warning at the end of an output file if convergence
> issues were seen during a run. This might be useful information, I'll look
> into how best to do it.


About parallelization: md2.out, Section 3 ATOMIC COORDINATES AND VELOCITIES
indicates:

QMMM: Running QMMMM calculation in parallel mode on 1 threads

....................

Running AMBER/MPI version on 1 nodes

Does this mean that sander, contrary to the command

$AMBERHOME/exe/sander.MPI -O -i .....

is running serial or is it that refers to only this section of sander? In fact,
the machine I am using has four nodes (as reported by QM calculation with
either NWChem or MPQC). Also, this installation of Amber9 passed all parallel
tests.

Another concern is the Memory Use /Allocated:

I see from md2.out that only a tiny part of the 4GB ECC memory per node are
being used (while the OS Debian Linux amd64 is set to provide all memory, as it
in fact occurs when running QM with NWchem or MPQC). I know that memory is of
less concern for MD than QM, nonetheless I am concerned about using all the
hardware that the machine can provide.

Thanks for advice

francesco


>
> All the best
> Ross
>
> /\
> \/
> |\oss Walker
>
> | HPC Consultant and Staff Scientist |
> | San Diego Supercomputer Center |
> | Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
> | http://www.rosswalker.co.uk | PGP Key available on request |
>
> 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.
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>



      ____________________________________________________________________________________
Park yourself in front of a world of choices in alternative vehicles. Visit the Yahoo! Auto Green Center.
http://autos.yahoo.com/green_center/
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Aug 15 2007 - 06:07:11 PDT
Custom Search