[gmx-users] 3.3beta vs 3.2 dispersion correction term
Sukit Leekumjorn
leekumjo at vt.edu
Wed Jul 27 22:58:17 CEST 2005
Thank you for your reply, David
Sorry about not comparing things one at a time. I took your suggestion
and tried comparing both cutoff (rvdw=1 and rcoulomb=1.8) and pme
(rvdw=0.9 and rcoulomb=0.9) for both versions without corrections; and
it turns out to be almost exactly the same. I then took an additional
step to compare the pme simulation including the correction factor for
both versions. Listed below is the total energy caculated in step 0.
You mentioned that the corrction factor in 3.2.1 version is not correct
and it is fixed in the 3.3B version, I am not sure if this is normal for
the correction factor to be 2 orders of magnitude higher, which greatly
effects the total energy of the system.
3.2.1 cutoff: -1.71612e+6
3.2.1 pme without correction: -1.69195e+6
3.2.1 pme with correction: -1.69664e+6 (Disp Corr = -4.69360e+3)
3.3B cutoff: -1.71385e+6
3.3B pme without correction: -1.69413e+6
3.3B pme with correction: -1.96249e+6 (Disp Corr = -2.68357e+5)
Below are mdp files used:
Cutoff:
title = dppc
cpp = /lib/cpp
integrator = md
nsteps = 100
nstlist = 10
nstxout = 20
nstvout = 20
nstxtcout =
nstlog = 20
dt = 0.002
constraints = all-bonds
nstenergy = 20
ns_type = grid
coulombtype = cut-off
rlist = 1.0
rvdw = 1.0
rcoulomb = 1.8
tcoupl = berendsen
tc_grps = dppc sol
tau_t = 0.1 0.1
ref_t = 323 323
Pcoupl = berendsen
Pcoupltype = anisotropic
tau_p = 2.0
compressibility = 4.6e-5 4.6e-5 4.6e-5 0.0 0.0 0.0
ref_p = 1.0 1.0 1.0 0.0 0.0 0.0
gen_vel = yes
gen_temp = 323
gen_seed = 54321
constraint-algorithm = Lincs
PME:
title = dppc
cpp = /lib/cpp
define =
integrator = md
nsteps = 100
nstlist = 10
nstxout = 20
nstvout = 20
nstxtcout =
nstlog = 20
xtc_grps =
energygrps =
dt = 0.002
constraints = all-bonds
nstenergy = 20
ns_type = grid
pbc = xyz
coulombtype = PME
rlist = 0.9
rvdw = 0.9
rcoulomb = 0.9
vdw-type = Cut-off
DispCorr = EnerPres
fourierspacing = 0.12
pme_order = 8
ewald_rtol = 1e-05
ewald_geometry = 3d
epsilon_surface = 0
optimize_fft = yes
tcoupl = berendsen
tc_grps = dppc sol
tau_t = 0.1 0.1
ref_t = 323 323
Pcoupl = berendsen
Pcoupltype = anisotropic
tau_p = 2.0
compressibility = 4.6e-5 4.6e-5 4.6e-5 0.0 0.0 0.0
ref_p = 1.0 1.0 1.0 0.0 0.0 0.0
gen_vel = yes
gen_temp = 323
gen_seed = 54321
constraint-algorithm = Lincs
unconstrained-start = no
lincs-order = 4
David wrote:
>On Wed, 2005-07-27 at 14:07 -0400, Sukit Leekumjorn wrote:
>
>
>>Dear Gmx Users
>>
>>I have used the dppc system (121,856 atoms) listed in the benchmark
>>section on the Gromacs website to test pme run for version 3.2 and
>>3.3beta.. Based on a short simulation run of 5ps from one node, DispCorr
>>term seem to be a lot different than older version (see below). I used
>>the same .mdp, .top and .gro given in the website. Based on these two
>>runs, the area per headgroup from 3.2 version is about 0.63 nm2 (as
>>reported in many literatures) however the 3.3beta drop down to 0.57nm2.
>> From the Feature/change section in the Gromacs website, I notice the
>>changes in dispersion correction for a large system. I was wandering if
>>the size of the system effect pme calculation in beta version, maybe
>>incorrect DispCorr term which effect the pressure of the system and then
>>the average area per headgroup of lipid?
>>
>>
>
>First you need to be a bit more systematic and change one thing at a
>time. The dispersion correction was wrong in 3.2. Therefore you'll have
>to turn it off to compare codes. Make a tpr in 3.2.1 and use the same
>tpr in 3.3
>
>Second if you want to compare area per headgroup you'll have to check
>many params such as pressure coupling and also have to simulate longer
>
>Third you say you use PME, however you don't say which cut-offs you use
>for Coulomb and VDW. Better to try without PME first.
>
>
>
>>Sukit Leekumjorn
>>
>>3.2 version
>>
>> Step Time Lambda
>> 0 0.00000 0.00000
>>
>>Grid: 39 x 40 x 13 cells
>>
>>Testing x86 processor CPUID...
>>CPU manufactured by Intel.
>>
>>Testing x86 SSE capabilities...
>>CPU and OS support SSE.
>>Using Gromacs SSE single precision assembly innerloops.
>>
>> Rel. Constraint Deviation: Max between atoms RMS
>> Before LINCS 0.012457 2313 2332 0.001510
>> After LINCS 0.000049 734 735 0.000006
>>
>>Long Range LJ corr. to Epot: -4693.6, Pres: -78.1849, Vir: 28161.6
>> Energies (kJ/mol)
>> Angle Proper Dih. Ryckaert-Bell. Improper Dih. LJ-14
>> 8.73166e+04 2.91992e+04 5.71785e+04 4.00762e+03 2.09315e+05
>> Coulomb-14 LJ (SR) Disper. corr. Coulomb (SR) Coulomb (LR)
>> 1.88239e+05 -4.91021e+04 -4.69360e+03 -1.28225e+06 -1.26585e+06
>> Potential Kinetic En. Total Energy Temperature Pressure (bar)
>> -2.02664e+06 3.30030e+05 -1.69661e+06 3.24380e+02 5.36413e+03
>>
>>3.3 beta version
>>
>> Step Time Lambda
>> 0 0.00000 0.00000
>>
>>Grid: 39 x 40 x 13 cells
>>Configuring nonbonded kernels...
>>Testing AMD 3DNow support... not present.
>>Testing ia32 SSE support... present.
>>
>>
>> Rel. Constraint Deviation: Max between atoms RMS
>> Before LINCS 0.012457 2313 2332 0.001510
>> After LINCS 0.000050 734 735 0.000006
>>
>>Long Range LJ corr. to Epot: -268357, Pres: -4470.22, Vir: 268357
>> Energies (kJ/mol)
>> Angle Proper Dih. Ryckaert-Bell. Improper Dih. LJ-14
>> 8.73166e+04 2.91992e+04 5.71785e+04 4.00762e+03 2.09447e+05
>> Coulomb-14 LJ (SR) Disper. corr. Coulomb (SR) Coul. recip.
>> 1.88330e+05 -4.91021e+04 -2.68357e+05 -1.28252e+06 -1.27039e+06
>> Potential Kinetic En. Total Energy Temperature Pressure (bar)
>> -2.29489e+06 3.32431e+05 -1.96246e+06 3.26740e+02 1.04578e+03
>>
>>
>>_______________________________________________
>>gmx-users mailing list
>>gmx-users at gromacs.org
>>http://www.gromacs.org/mailman/listinfo/gmx-users
>>Please don't post (un)subscribe requests to the list. Use the
>>www interface or send it to gmx-users-request at gromacs.org.
>>
>>
More information about the gromacs.org_gmx-users
mailing list