[gmx-users] Gromacs on Gigabit Ethernet Dual Opteron Cluster

Erik Lindahl lindahl at csb.stanford.edu
Mon Oct 6 19:01:02 CEST 2003


It depends on the type of interaction, but water-water calculations can 
sometime be as cheap as 50% of normal interactions.

(Because we have special optimized loops for them, and there is only 
one LJ interaction site per molecule).



On Monday, October 6, 2003, at 02:26 AM, Ghermes Chilov wrote:

> Hello Anton,
> Monday, October 6, 2003, 11:59:40 AM, you wrote:
> AF> Ghermes Chilov wrote:
>>> The system (dummified protein in water)is big enough:~ 107.000 atoms,
>>> -shuffle -sort were used.
> AF> If your protein is big w/r the number of waters, it will become 
> limiting
> AF> as soon as it fills a whole processor, i.e. CPU0 will be 
> overloaded with
> AF> protein, and the others are doing water with time to spare...
> No, it's not big: ~ 7-8% of the system... khmm, it becomes sufficient
> for 100%/8%=12 CPU's, isn't it?  Or even for a smaller number, if
> water-water interactions are in principle more easy to calculate.
> Ghermes
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.

More information about the gromacs.org_gmx-users mailing list