From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed Aug 02 2006 - 17:25:32 CDT

Michel,
How much memory is it using? If you're already well over 2GB,
then the Opteron will likely be insufficient. VMD 1.8.5 uses a bit
more memory per-atom than older versions because it allows 1.5x more
bond partners per atom, for some special non-biological structures.
This and a few other things can be tuned to save memory if necessary,
but a lot depends on whether you want to load more than one timestep.

If you need/want to be able to load many timesteps of your trajectories,
you're going to want at least 4GB of memory, and maybe more like 8GB to
be really serious about it. Do you guys have any budget for extra RAM?
You should be able to buy 4GB of RAM for something between
$500 to $1,000, based on some prices I see on the net:
  http://www.pricewatch.com/memory/

What does the VMD text console say currently, is it done loading the
structure file already and moving on to bond search, or is it still reading
the structure from disk? If it's reading from disk still (no messages
about bond search or about analyzing the structure) then you're probably
going to need more memory or to run a "lightweighted" VMD build to get
by with 2GB of RAM. 6.5 million atoms requires 78MB per timestep for
just the atom coordinates, so that says that even with 2GB of free
memory (which is never the case on a real machine) you'd only be able
to load _25_ timesteps of your simulation at once.

Most likely, even before you start VMD you've probably only got
something like 1.5GB free at most. Even if I setup a binary plugin
for you, you're going to need to either get more memory or use CatDCD
to chop out unnecessary atoms from your trajectory in order to see
more timesteps in VMD. If you can collect more information on how much
memory VMD is using during the load, and what phase of loading it has
reached, I can give more suggestions.

  John Stone
  vmd_at_ks.uiuc.edu

On Wed, Aug 02, 2006 at 05:04:07PM -0500, L. Michel Espinoza-Fonseca wrote:
> Yeah, is what I thought about the Onyx, even though she's been a good girl
> :).
>
> It'll be great to have such plugin to load this protein much faster! I
> loaded a single snapshot of the molecule into VMD and I've been
> waiting for almost 1.5 hrs! It gives me time to take a break and drink
> some good coffee, but after some time it really annoys me...
>
> BTW, I'm running the latest beta version of VMD 1.8.5 under Fedora 5.
>
> Thanks!
> Michel
>
> 2006/8/2, John Stone <johns_at_ks.uiuc.edu>:
> >
> >Michel,
> > The opteron will be a much faster machine for visualization and
> >analysis since the old Onyx machine has (by today's standards)
> >very slow processors in it. What file format was your structure
> >stored in? If you're working with NAMD, you'll get the best results
> >by loading your PSF file first, and then loading the PDB/DCD second.
> >Doing things in some other order will be slower and eat more memory.
> >
> >While you have VMD open on the Opteron with the big file loaded,
> >see how much memory it is using using the 'top' command (assuming
> >that you're running Unix on this machine). If you can do that and
> >report back the numbers, I can give you some suggestions for getting
> >further performance increases.
> >
> >If it is mainly the file loading time for PSF/PDB that is slow,
> >I could give you a special binary molecule file plugin
> >so you can load this thing MUCH faster during your day-to-day work.
> >I forgot to ask, which version of VMD are you using currently?
> >
> > John Stone
> > vmd_at_ks.uiuc.edu
> >
> >On Wed, Aug 02, 2006 at 04:41:16PM -0500, L. Michel Espinoza-Fonseca wrote:
> >> Hi John,
> >>
> >> Thank you for your reply. I'm actually going to simulate such a big
> >> system with NAMD and further analyze it with VMD. I'm really concerned
> >> about this, because I tried to load a single structure on my
> >> workstation (Opteron, 2Gb RAM) and it took really long time. I have
> >> access to more robust visualization facilities (a SGI Onyx4, 16
> >> processors, 32 Gb RAM), so maybe I could use it for visualization and
> >> analysis. What do you think?
> >>
> >> Concerning the connectivity info, I'll be using a psf file +
> >> trajectory in dcd format. Considering what you told me, I think that
> >> I'll save a lot of time :)
> >>
> >> Thank you for your help!
> >> Michel
> >>
> >> 2006/8/2, John Stone <johns_at_ks.uiuc.edu>:
> >> >
> >> >Michel,
> >> > VMD can load structures that large with no problem if you have enough
> >> >memory, say 2 to 3GB. The main remaining issue after memory is that
> >> >you'll be able to load such a structure MUCH faster if you use a file
> >> >format that contains bond/connectivity information. If there's no
> >> >connectivity info (e.g. lone PDB), then VMD will have to do a bond
> >> >search after the file is loaded, and that can take a long time on
> >> >really really large structures.
> >> >
> >> >With a structure that large, I highly recommend using the
> >> >Points, NewRibbon, and NewCartoon representations since they'll work
> >better
> >> >for things of that size. If I knew more about your structure and what
> >> >you want to do with it once in VMD, I could give you more suggestions on
> >> >getting it loaded and displayed most efficiently. Let me know if you
> >> >need more help.
> >> >
> >> > John Stone
> >> > vmd_at_ks.uiuc.edu
> >> >
> >> >
> >> >On Wed, Aug 02, 2006 at 03:25:41PM -0500, L. Michel Espinoza-Fonseca
> >wrote:
> >> >> Hi John,
> >> >>
> >> >> I was wondering about what's the maximum number of atoms that VMD can
> >> >> handle. Currently, I have a system of about 6.5 million atoms, so I
> >> >> would like to know if I'll be able to visualize it :)
> >> >>
> >> >> Thanks!
> >> >> Michel
> >> >
> >> >--
> >> >NIH Resource for Macromolecular Modeling and Bioinformatics
> >> >Beckman Institute for Advanced Science and Technology
> >> >University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> >> >Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> >> > WWW: http://www.ks.uiuc.edu/~johns/ Fax: 217-244-6078
> >> >
> >
> >--
> >NIH Resource for Macromolecular Modeling and Bioinformatics
> >Beckman Institute for Advanced Science and Technology
> >University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> >Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> > WWW: http://www.ks.uiuc.edu/~johns/ Fax: 217-244-6078
> >

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email: johns_at_ks.uiuc.edu                 Phone: 217-244-3349
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078