Is OpenMM Zephyr the right choice?

OpenMM Zephyr provides a visual application for running GPU-accelerated molecular simulations.
POST REPLY
User avatar
Cole Gleason
Posts: 4
Joined: Tue Apr 13, 2010 9:49 pm

Is OpenMM Zephyr the right choice?

Post by Cole Gleason » Tue Sep 07, 2010 12:10 pm

I'm a high school senior who is currently working on an independant thesis with my professor. He is interested in MD, and we want to create a system students can use at our school to run simulations. He really wants to be able to do explicit water simulations. We really just want to get started with the simulations, as I have already spent a lot of time making a MPI-enabled cluster using some old school computers. I need to know if Zephyr can be run using MPI and whether or not it can do multiple CPUs and GPUs.

User avatar
Joy Ku
Posts: 80
Joined: Tue Oct 02, 2007 5:22 pm

RE: Is OpenMM Zephyr the right choice?

Post by Joy Ku » Tue Sep 07, 2010 12:36 pm

Hi Cole,

Zephyr is now capable of running explicit solvent simulations (including explicit water), and it is extremely easy to use and install. You can access the MPI and multiple CPU capabilities built in to GROMACS, the underlying MD engine for Zephyr, but only through a command-line interface (see Chapter 5 of the user manual and use the standard GROMACS mdrun command. You'll also want toconsult the GROMACS manual available at http://gromacs.org). The friendly user interface that appears when you launch Zephyr does not provide access to multiple CPUs. Multiple GPUs are not currently supported.

Joy


User avatar
Cole Gleason
Posts: 4
Joined: Tue Apr 13, 2010 9:49 pm

RE: Is OpenMM Zephyr the right choice?

Post by Cole Gleason » Tue Sep 07, 2010 4:00 pm

Thanks for the reply! It seems like the MD community is pretty quiet.

So Zephyr is not capable of MPI? Is that a limitation of Zephyr or OpenMM? I already have GROMACS compiled with MPI support, which is fine, but my professor wants to use Zephyr due to its ease of use and his connections at Stanford.

I'm also looking for some device on hardware. Currently I'm running eight old Pentium 4s in a cluster. Would I see better performance if I were to switch to either A) one high performance machine with GPU acceleration or B) use Amazon's E3 service?

User avatar
Joy Ku
Posts: 80
Joined: Tue Oct 02, 2007 5:22 pm

RE: Is OpenMM Zephyr the right choice?

Post by Joy Ku » Wed Sep 08, 2010 2:16 pm

Zephyr helps connect 3 software packages together (OpenMM, GROMACS, and VMD) via a user interface. The computational capabilities are provided by these other packages, not by Zephyr. Since OpenMM does not currently support MPI, that means that Zephyr doesn't have that capability either.

As for the hardware advice and connections with the MD community, I would suggest posting to the OpenMM forum (http://simtk.org/home/openmm, click on Advanced -> Public Forums). That group is much more active.

Joy

User avatar
Joy Ku
Posts: 80
Joined: Tue Oct 02, 2007 5:22 pm

RE: Is OpenMM Zephyr the right choice?

Post by Joy Ku » Wed Sep 08, 2010 2:20 pm

Just to clarify. Zephyr's user interface does not provide access to MPI capabilities through the user interface, though as mentioned earlier, you can use the standard GROMACS's MPI capabilities via a command-line interface. Standard GROMACS is installed when you install Zephyr.

User avatar
Cole Gleason
Posts: 4
Joined: Tue Apr 13, 2010 9:49 pm

RE: Is OpenMM Zephyr the right choice?

Post by Cole Gleason » Wed Sep 08, 2010 3:30 pm

Alright, thanks for the help.

POST REPLY