[gmx-users] implicit solvent

[Erik Lindahl]

Hi Stefano,

>
>
> I think you are right for what concerns the speed of the calculations, 
> however   don't underestimate the advantage of getting rid of the 
> solvent noise, degrees of freedom and viscosity. Eliminating solvent 
> viscosity could speed up considerably large conformation changes and 
> eliminating solvent noise and degrees of freedom is extremely helpful 
> for scripts that exploit parallelization in a parallel tempering scheme. 

The lack of viscosity is both a feature and a problem; it makes some 
large-scale motions occur 10-15 times faster, but this of course means 
we cannot draw conclusions about timescales of the dynamics anymore. We 
might have to include a stochastic integrator that adds some drag to 
atoms exposed to the (implicit) solvent...

Another reason this is taking a while is that we probably skipping 3.2 
and going directly for 4.0, with better parallelization and a bunch of 
other features...

Cheers,

Erik