Minimization converges with gromacs 2021.5, but fails with gromacs 2022.2

User discussions
1

GROMACS version: 2021.5 and 2022.2
GROMACS modification: No

I am trying to minimize a system containing membrane/protein/solute.
my minimization parameres are automatically generated by CharmmGui, and are as follows :

define = -DPOSRES -DPOSRES_FC_BB=4000.0 -DPOSRES_FC_SC=2000.0 -DPOSRES_FC_LIPID=1000.0 -DDIHRES -DDIHRES_FC=1000.0
integrator = steep
emtol = 1000.0
nsteps = 5000
nstlist = 10
cutoff-scheme = Verlet
rlist = 1.2
vdwtype = Cut-off
vdw-modifier = Force-switch
rvdw_switch = 1.0
rvdw = 1.2
coulombtype = PME
rcoulomb = 1.2
;
constraints = h-bonds
constraint_algorithm = LINCS

when running the minimization using gromacs version 2021.5, the results are :

Steepest Descents converged to Fmax < 1000 in 1488 steps Potential Energy = -4.5349294e+05 Maximum force = 9.7959656e+02 on atom 4581 Norm of force = 2.4778579e+01

however, using the same mdp and the same data files with gromacs2022.2, the results are :

Energy minimization has stopped, but the forces have not converged to the requested precision Fmax < 1000 (which may not be possible for your system). It stopped because the algorithm tried to make a new step whose size was too small, or there was no change in the energy since last step. Either way, we regard the minimization as converged to within the available machine precision, given your starting configuration and EM parameters.

Double precision normally gives you higher accuracy, but this is often not
needed for preparing to run molecular dynamics.
You might need to increase your constraint accuracy, or turn
off constraints altogether (set constraints = none in mdp file)

writing lowest energy coordinates.

Steepest Descents converged to machine precision in 203 steps,
but did not reach the requested Fmax < 1000.
Potential Energy = -4.0351597e+05
Maximum force = 1.8234898e+04 on atom 1072
Norm of force = 1.9618500e+02

The 2022.2 version would fail to minimize this system even if the restraints were completely removed.
I could not detect any reason for this behavior in the 2022.2 release notes, and I wonder what may cause this issue under the hood and how to overcome it.

2

Could this be due to a bug that was fixed in 2022.3?

Energy minimization would not converge with GPU and without DD
“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”“”

The steepest descent and conjugate gradient minimizers would not converge
when using a GPU for the nonbonded interactions and not using domain
decomposition.

3

Hi,

  1. Thanks for the reply.

  2. I could not pinpoint the root cause of this issue, but I have realized that it could be avoided by running the minimization of the CPU (disabling the GPU), which should be an acceptable solution for most systems.

yours,

Izhar.

4

Then it is the issue I mentioned. I would suggest to upgrade to 2022.4.