[gmx-users] About running Production simulation in GPU with twinrange cut-off

[Justin Lemkul]

On 8/11/13 3:02 AM, vidhya sankar wrote:
> Thank you Justin
>                              To check Performance in GPU i Need to  run simulation  with same parameters for r coulomb =  rvdw
>
>        rlist        = 0.9        ; short-range neighborlist cutoff (in nm)
> rcoulomb    = 1.4        ; short-range electrostatic cutoff (in nm)
> rvdw        = 1.4
>   But i need to use twin range interaction with rvdw = 1.4; r coulomb = 0.9 in GPU
>
> May i use different  option -nb or -tunepme   option of mdrun
>
> instead -testverlet
>

mdrun -nb gpu should be sufficient to run on the GPU.  The -tunepme option is 
not relevant.

>
> But in one mail  to me you said that the defaut parameters for  Gromos96 53A6,
>
> nstlist     = 5
>   rlist       = 0.9
>   rcoulomb    = 0.9
>   rvdw        = 1.4
> but if i run on GPU regularly i need to change rcoulomb = 1.4 from 0.9 Will it Affect the result ?
> or is it reasonable to change ?
>

When using PME, the value of rcoulomb is a bit more flexible than other 
settings, but you need to carefully validate that such modifications do not 
invalidate the proper use of the force field.  As previous threads have very 
clearly demonstrated, the increase in speed is irrelevant if the results are 
influenced by artifacts.

-Justin

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

Justin A. Lemkul, Ph.D.
Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441

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