Dear Varsha,
> I was reading tutorial8 and i have question about
> heating the system. My protein is 260 amino acid long
> and i wanted to heat the system. In tutorial cutoff
> =999.0 and its taking me 10 days to run 10,000 cycles
> on 16 processors. Can you suggest another alternative
> to speed up the system.
If your system is this big then you definately don't want to run without a
cutoff. Remember, the charge-charge interaction scales as N^2 without a
cutoff so for a system 10 times the size of the one in the tutorial you will
need 100 times more cputime. You don't mention if you are running periodic
boundaries or with GB. If you are running PME periodic boundaries set
cut=8.0. If you are running GB then set it to 16.0. Also in the tutorial I
used a large value of RGBMAX that will also slow down the simulation. If you
are running a GB simulation on a large system you should probably set this
back to the default of 25.0. Basically the bigger cut and RGBMAX are the
better but there is a tradeoff between accuracy and cputime that needs to be
made.
Another thing to check is that your simulation runs sufficiently on 16 cpus.
This is especially true if you are using a cluster with a poor interconnect
such as ethernet. Try running 10 steps of MD on 1,2,4,8,16,32 cpus and find
the sweet spot in terms of time.
> I tried to run MD without heating the system but many
> times temp. went about 400 and graph looks wierd.
> Thanks a lot,
Yes, this can often happen with large systems, you need to do controlled
heating so your system can relax slowly and not "blow up".
I hope this helps.
All the best
Ross
/\
\/
|\oss Walker
| Department of Molecular Biology TPC15 |
| The Scripps Research Institute |
| Tel: +1 858 784 8889 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk | PGP Key available on request |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be ready every day, and should not be used for urgent or sensitive issues.
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Jul 26 2005 - 18:53:00 PDT
