AW: First meeting with NAMD

From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Tue Jun 04 2013 - 02:16:18 CDT

Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im Auftrag
von James Starlight
Gesendet: Montag, 3. Juni 2013 16:09
An: namd-l_at_ks.uiuc.edu
Betreff: namd-l: First meeting with NAMD

 

Dear NAMD users!

Hi james,

 

Recently I've tried to launch my first simulation on NAMD :) (previously
I've used Gromacs ).
My questions:

1) Typical gromacs simulation consist of energy minimization + equilibration
in NPT ensemble (with position restraints applied on each atoms of protein)
+ production run.

As I understood minimization should explicitly defined in the conf file

# Minimization
minimize 100
reinitvels $temperature

run 5000000

but what about equilibration stage ? How I can perform short simulation with
the applied porses prior to the production run?

We usually have the following simulation protocol:

1. Minimization for some thousand steps, depending on system size
(check if TOTAL energy converges)

2. Heating up using Langevin thermostat (there are multiple methods and
thermostats available)

3. Constant Pressure (remove vacuum bubbles from solvent using piston
barostat, also other choices available)

4. Heat again as pressure could have changed anything

5. Free simulation <- production run

Each of these steps is represented by a own namd script. Additionally, each
step uses the final coordinates and velocities from the step before, except
minimization which start from the initial structure of course.

2) How I can monitor total performance of the GPU utilized in the
simulation assuming that I use CUDA.

 

Depending on what kind of GPU you got, you can try nvidia-smi to check for
the utilization (I guess only for Tesla). But as others already said, you
should use the CPU:GPU ratio and configuration, that comes with the smallest
time/step. Additionally, for most system sizes I got a nice almost two-fold
speedup by using "twoawayx yes" in my namd script, you should try the
difference. To be sure to get the best performance, it's worth it to do some
benchmarks.

3) I have parameters ( prm and inp files) for some non-standard residue (
GFP chromophore). for this protein I'd like to make model (including protein
covalently bonded to the chromophore and solvent) and perform simulation.
Could you provide me with the example or short tutorial for such task?

See VMD and psfgen. Additionally search for the NAMD Tutorial on the net,
which is a really nice for starting for both NAMD and VMD.

Thanks for help,

James

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:17 CST