AMBER Archive (2009)
Subject: Re: [AMBER] Problems using nmode in NAB
From: Andreas Svrcek-Seiler (svrci_at_tbi.univie.ac.at)
Date: Thu Jun 25 2009 - 08:03:15 CDT
> I was wondering if you can give me a hint about what file should be changed
> on neb to increase memory size?.
> My system is 6000 atoms, I have heard of people here doing 10000 atom
> systems. but I dont know what *.h file I have to change
> to increase the memory :=(
...6000 atoms means 18000 coordinates, which in turn means the
hessian has 324*10**6 entries, each of which takes 8 bytes of memory.
Baseline: for 6000 atoms the first thing you need is about 2.5 GB of
memory for the hessian alone (the hessian is symmetric an one could spare
half of that in principle, but I believe it is not done in the
routines NAB uses (?)).
...just tried it out:
NR minimization of a 6370 atom molecule takes about 6 GB memory.
Calculating the nmodes increases memory-consumption to ~18 GB.
Seems you need a 64 it machine with at least 20 GB of RAM or so.
AMBER mailing list