Previous Thread
Next Thread
Print Thread
Page 1 of 2 1 2
#16279 11/17/07 02:31 AM
Joined: Sep 2004
Posts: 57
B
Baloo Offline OP
Forum Member
OP Offline
Forum Member
B
Joined: Sep 2004
Posts: 57
Hi,

This might be a silly question to ask. I have made th NAMD dcd file and to find out the number of frames it generated before my NAMD script crashed i used the traj query command. But i get the following output..has any one any clue of why i am getting this..

best


CHARMM> traj query unit 20

READING TRAJECTORY FROM UNIT 20
NUMBER OF COORDINATE SETS IN FILE:********
NUMBER OF PREVIOUS DYNAMICS STEPS:********
FREQUENCY FOR SAVING COORDINATES: ********
NUMBER OF STEPS FOR CREATION RUN: ********

NUMBER OF DEGREES OF FREEDOM: 0
NUMBER OF FIXED ATOMS: 0
THE INTEGRATION TIME STEP (PS):***********
THE FILE DOES NOT CONTAIN CRYSTAL UNIT CELL DATA
THE FILE DOES NOT CONTAIN 4-D DATA

Joined: Sep 2003
Posts: 8,629
Likes: 24
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,629
Likes: 24
There can be trajectory compatibility issues, esp. for different machine types; there's a known issue with CHARMM compiled via gfortran 4.1.x

You may wish to try the 'Search' feature of these forums; set the time range to 2-3 years

Joined: Jul 2004
Posts: 165
B
Forum Member
Offline
Forum Member
B
Joined: Jul 2004
Posts: 165
Hi Baloo ,

Can you solve the problem? How?

Thanks

Joined: Sep 2003
Posts: 4,863
Likes: 10
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 4,863
Likes: 10
Running a simulation is often not the most time consuming part of a computaional biophysics project - the analysis part is often both the most important and the most extensive part; using one program that can do all the aspects of your project may therefore be a good strategy. Only a few projects benfit from the weak parallel scaling of NAMD - mostly you are much better off in terms of sampling if you run 1024 single CPU simulations of 10ns instead of one 10 microsecond simulation, in which case CHARMM is faster than both NAMD and GROMACS. So my advice is simple: Rerun your simuation with CHARMM, and you will have not experience the kind of problems analyzing the trajectory that you now seem to be facing. YOUR time is far more vaulable than the CPU-time.


Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden
Joined: Jul 2004
Posts: 165
B
Forum Member
Offline
Forum Member
B
Joined: Jul 2004
Posts: 165
Thank you very much.

I have been working with charmm for a couple of year. In our MD team, only I myself focus on biomolecular simulation and use the charmm programm. To learn charmm programm alone without expert in the group, it is not easy. Thank you to this forum. Almost the end of the project, I donĀ“t want to spend much time to learn the new program (with no support! any more), and still would like to share the ideas with this active forum. However, I am now working on almost 100,000 atoms. Based on my little experience, I can get 1ns with namd in a week, but a month in charmm (same condition). What should I do then?

Joined: Sep 2003
Posts: 4,863
Likes: 10
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 4,863
Likes: 10
Timings are of course system dependent. I have a benchmark with 67000 atoms (protein in water) that runs 1.2 ns/week on 4 CPUs. Some things that matter for speed, apart from using a good compiler, are to use the BYCBim nonbond list generator for large systems, to use SHAKE FAST, and the table lookup for nonbond interactions (for this you need c34b1 though).


Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden
Joined: Sep 2003
Posts: 8,629
Likes: 24
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,629
Likes: 24
Lennart makes a good point-- it can be fairly important to adjust some of the setup in CHARMM in order to optimize the performance. An example is the CUTNB value; in many cases this can be increased, which makes the calculation run faster (fewer non-bond updates) at the expense of memory usage. The CTOFNB and CTONNB values should be set to their values from the parameter set, usually 12. and 8., respectively.

Joined: Sep 2003
Posts: 8,629
Likes: 24
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,629
Likes: 24
With respect to trajectory compatibility, the machine chipset, operating systems, and the Fortran compiler can affect the binary storage form that can be read by a particular CHARMM version. In some cases, only executables compiled on the same machine with the same compiler can read CHARMM binary trajectories created with that setup. The same can apply to reading NAMD trajectory files; not all CHARMM executables may be able to do so. The solution depends on the details, which have not yet been provided.

Joined: Oct 2007
Posts: 26
M
Forum Member
Offline
Forum Member
M
Joined: Oct 2007
Posts: 26
I just found that post because I'm thinking about using CHARMM for setup and analysis and NAMD for the actual run of the simulation on our new cluster.

I tried to find any recent performance comparison on CHARMM vs. NAMD but this seems to be the most recent:

http://amber.scripps.edu/amber8.bench2.html

They are still using CHARMM c31a2 there...and as I was told CHARMM got a pretty good performance bost with the c32 branch. So I wonder if it is really worth the hassle to have NAMD and CHARMM because the difference might not be as big anymore. Does anybody know how the most current versions of CHARMM and NAMD compare performance wise? The cluster consists of a few AMD Opteron (Barcelona) based Dell PowerEdge R905 (four 2.0GHz Barcelona Quadcores each). Since the connection between the machines is rather slow we would limit ourselves to the 16 cores that are in one machine most of the time.

And then I'm wondering how difficult it is to evaluate NAMD simulations with CHARMM. I thought that both are completely file compatible (besides input files) and that I could just run the analysis like on a CHARMM simulation. But here it sounds as there is more to it. Any comments on that?

Thanks a lot,

Alexander

Joined: Sep 2003
Posts: 8,629
Likes: 24
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,629
Likes: 24
For some types of calculations, esp. with a few compromises, NAMD can be considerably faster.

Working with both programs can be an issue; besides the chipset and compiler issues noted above, there is also a PSF compatibility issue as well. NAMD can read a PSF created with CHARMM and exported in XPLOR format, but CHARMM cannot read a PSF created by NAMD (or VMD). If you wish to work with both programs, the best approach is to create an initial PSF with CHARMM, and export an XPLOR PSF for use with NAMD. Attempting to create a CHARMM PSF that exactly matches a NAMD PSF is not so easy; not only must the number of atoms be the same, but they must appear in the same order in both types of PSF. This is a critical detail that is often overlooked by casual users who wish to analyze their NAMD-created .dcd file using CHARMM. A key test is the ability to evaluate the energy of a stored trajectory frame; if that doesn't work, you probably have a PSF compatibility problem.

Page 1 of 2 1 2

Moderated by  BRBrooks, lennart, rmv 

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.5
(Release build 20201027)
Responsive Width:

PHP: 7.3.31-1~deb10u2 Page Time: 0.010s Queries: 35 (0.005s) Memory: 0.7836 MB (Peak: 0.8788 MB) Data Comp: Off Server Time: 2023-01-27 07:10:32 UTC
Valid HTML 5 and Valid CSS