[gmx-users] Is Generalized Born implicit solvent model faster than explicit solvent model?

[Justin Lemkul]

On 8/31/16 11:04 PM, Dd H wrote:
> Hi everyone,
> I did two reMD simulations with GB implicit solvent and TIP3P solvent using
> Gromacs4.5.5 respectively (1 cpu per replica for both of them). My results
> show that implicit solvent model is not faster than explicit solvent
> model. I'm confused why the implict solvent model is so slow? Or should I
> use a newer version of Gromacs?

Implicit simulations do not scale well.  At one point, there was a hard limit of 
2 threads in a GB simulation, though this may have been lifted.  DD is of no use 
as there are too few atoms to parallelize efficiently.

Newer versions of GROMACS will not make GB faster, and implicit models are 
actually slated for removal due to lack of feature compatibility and low 
performance (not to mention little developer interest in maintenance).

-Justin

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

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

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