[gmx-users] Problem with OpenMP+MPI

[Justin Lemkul]

On 2/27/13 10:01 AM, jesmin jahan wrote:
> Hi Justin,
>
> Thanks for your reply.
>
> I am able to run it now. I am using the following parameters.
> --------------------------------------------------------------------------------------------------------------
> constraints         =  none
> integrator          =  md
> cutoff-scheme       = Verlet
> pbc                 =  xyz
> verlet-buffer-drift = -1
> dt                  =  0.001
> nsteps              =  0
> rcoulomb            = 1
> rvdw                = 1
> rlist               = 1
> nstgbradii          = 1
> rgbradii            = 1
> implicit_solvent    =  GBSA
> gb_algorithm        =  HCT ;
> sa_algorithm        =  None
> gb_dielectric_offset    = 0.02
>
> optimize_fft             = yes
> energygrps               = protein
>
> --------------------------------------------------------------------------------------------
> I got the following energy values for this run.
> Statistics over 1 steps using 1 frames
>
>     Energies (kJ/mol)
> GB Polarization        LJ (SR)   Coulomb (SR)      Potential    Kinetic En.
>     -1.52255e+05    5.75419e+08   -2.44085e+05    5.75023e+08    6.36555e+11
>     Total Energy    Temperature Pressure (bar)
>      6.37130e+11    1.60603e+10    7.90024e+10
>
>
> ###############################################################################
> On the other hand, when I used gromacs-4.5.3, with the following parameters:
> -----------------------------------------------------------------------------------------------------------------------------------------
> constraints         =  none
> integrator          =  md
> pbc                 =  no
> dt                  =  0.001
> nsteps              =  0
> rcoulomb            = 300
> rvdw                = 300
> rlist               = 300
> nstgbradii          = 300
> rgbradii            = 300
> implicit_solvent    =  GBSA
> gb_algorithm        =  HCT ;
> sa_algorithm        =  None
> gb_dielectric_offset    = 0.02
>
> optimize_fft             = yes
> energygrps               = protein
> --------------------------------------------------------------------------------------
> I got the following energy values:
>
>    Energies (kJ/mol)
> GB Polarization        LJ (SR)   Coulomb (SR)      Potential    Kinetic En.
>     -1.58040e+04    0.00000e+00   -2.42022e+05   -2.57826e+05    0.00000e+00
>     Total Energy    Temperature Pressure (bar)
>     -2.57826e+05    0.00000e+00    0.00000e+00
>
>
> You can see that there is a huge difference (10 times), in these two
> energy values!
>
> How can I get a similar value that I was receiving previously (i.e.,
> -1.58040e+04)? The previous energy was correct, while the current
> value is 10 times more than the previous one.
>

Also note that your temperature is about a thousand times hotter than a nuclear 
explosion, as well :)

You're comparing dissimilar things in the two setups and expecting them to come 
out the same.  With 4.6, you're using PBC and explicit cutoffs, in a box that's 
likely inappropriate as you have massive atomic overlap (which is what gives 
rise to the ridiculous temperature).  In the case of 4.5.3 you have a 
non-periodic system and thus you can't get overlap between images (because there 
are none).  Conclusion: your settings for 4.6 simply don't make sense.

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================