[gmx-users] Help on MD performance, GPU has less load than CPU.

[Téletchéa Stéphane]

Le 11/07/2017 à 15:24, Mark Abraham a écrit :
> Guessing wildly, the cost of your simulation is probably at least double
> what the defaults would give, and for that cost, I'd want to know why.

Estimated colleague,

Since this is a wild guess, I'd think to add some guesses myself. I 
remember "some time" back having used a lower tolerance on Ewald for 
amber simulations (around amber 4/5/6 ...) and it was more common at 
this time I presume. This may also be linked to the fact that amber has 
a short cut-off at 8 angstrom for electrostatics ...
Someone apparently "ill" at the time already found this stane in 2009:

http://gromacs.org_gmx-users.maillist.sys.kth.narkive.com/vTjpMdwU/gromacs-preformance-versus-amber

Out of my memroy, I remembered using 10-6 for Ewald tolerance in AMBER, 
and this is mentioned here:

http://ambermd.org/Questions/ewald.html

... apparently linked to DNA simulation as found in JACS 117,4193 (1995)

In short, this value may come in back and forth for "historical" reasons 
(and misuse, of course).

Others may have additional comments :-)

Best,

Stéphane


-- 
Assistant Professor in BioInformatics, UFIP, UMR 6286 CNRS, Team Protein 
Design In Silico
UFR Sciences et Techniques, 2, rue de la Houssinière, Bât. 25, 44322 
Nantes cedex 03, France
Tél : +33 251 125 636 / Fax : +33 251 125 632
http://www.ufip.univ-nantes.fr/ - http://www.steletch.org