Hi
I am running minimizations using sander in AMBER7 and these take forever
compared to doing the same calculation using sander_classic in AMBER6.
(Forever means that instead of 1 minute it takes 30 minutes) I do not
understand what has happened. Here is an example of the input:
----------------------------------------------------------
&cntrl
imin=1, scee=1.2,
ntpr=100, maxcyc=10000,
cut=9.0,
dielc=1,
nsnb=25, ntr=0, drms=0.1,
ntmin=0, ncyc=10, ntb=0,nmropt=0,
&end
--------------------------------------------------------------
The size of the system is a 340 residue protein in vacuum. I run in
parallel using 4 processors. Here is an example of the TIMINGS list from
the output:
-----------------------------------------------------------------
| Check list validity 4.62 ( 0.49% of List )
| Map frac coords 8.77 (0.94% of List )
| Setup grids 0.14 ( 0.02% of List )
| Grid unit cell 351.05 (37.49% of List )
| Grid image cell 18.93 ( 2.02% of List )
| Build the list 551.67 (58.91% of List )
| Other 1.24 ( 0.13% of List )
| List time 936.46 (73.67% of Nonbo)
| Direct Ewald time 281.31 (84.07% of Ewald)
| Finish NB virial 4.31 ( 1.29% of Ewald)
| Ewald MPI wait 0.71 ( 0.21% of Ewald)
| Other 48.30 (14.43% of Ewald)
| Ewald time 334.62 (26.32% of Nonbo)
| Other 0.07 ( 0.01% of Nonbo)
| Nonbond force 1271.15 (90.54% of Force)
| Bond energy 2.92 ( 0.21% of Force)
| Angle energy 13.28 ( 0.95% of Force)
| Dihedral energy 51.85 ( 3.69% of Force)
| FRC Collect time 35.63 ( 2.54% of Force)
| Other 29.15 ( 2.08% of Force)
| Force time 1404.01 (100.0% of Runmd)
|Runmd Time 1404.01 (96.35% of Total)
|Other 53.13 ( 3.65% of Total)
|Total time 1457.17 (100.0% of ALL )
-----------------------------------------------------------------------
Why does the non bonded intractions take so terribly long time when the
cutoff is only 9 A???
I also wonder why the Ewald consumes any time at all. It is a vacuum
simulation and Ewald is supposed to be turned off???
Arvid
Received on Fri Dec 06 2002 - 00:51:10 PST
