[gmx-users] on the ussage of SD1 integrator

[James Starlight]

Dear Gromacs users!


Commonly I'm using langevin's dynamics for  thermostat modelling in my
simulations of the membrane proteins
(sd integrator without t_coupl with tau_t=2.0)

Recently I've found that with sd1 integrator I can obtain better
performance ( up to 50% in comparison to sd integrator). Could I use
that integrator for modelling of stochastic thermostat with the same
tau_t which I've used with SD ?
Also during ussage of that thermostat I obtain notions from grommp like

NOTE 4 [file ./mdps_cuda/md_sd.mdp]:
  With constraints integrator sd1 is less accurate, consider using sd
  instead

What should I do to reduce such degree of inacuracy of SD1?

Thanks for help,

James