Current NAMD Feature Summary

Software Setup

  • Free to download and use. (Redistribution prohibited.)
  • Precompiled binaries provided for Linux, Mac, and Windows.
  • Installed at major NSF supercomputer sites.
  • Portable to virtually any platform with ethernet or MPI.
  • C++ source code and CVS access for modification.

Molecule Building

  • VMD used to prepare molecular structure for simulation.
  • Also reads X-PLOR, CHARMM, AMBER, and GROMACS input files.
  • Psfgen tool generates structure and coordinate files for CHARMM force field.
  • Efficient conjugate gradient minimization.
  • Fixed atoms and harmonic restraints.
  • Thermal equilibration via periodic rescaling, reinitialization, or Langevin dynamics.

Basic Simulation

  • Constant temperature via rescaling, coupling, or Langevin dynamics.
  • Constant pressure via Berendsen or Langevin Nose-Hoover methods.
  • Particle mesh Ewald full electrostatics for periodic systems.
  • Symplectic multiple timestep integration.
  • Rigid waters and bonds to hydrogen atoms.

Advanced Simulation

  • Chemical and conformational free energy calculations.
  • Enhanced sampling via replica exchange.
  • Tcl based scripting and steering forces.
  • Analysis implemented as Tcl scripts in VMD.
  • Interactive visual steering interface to VMD.

Scalable Performance

  • Based on the Charm++/Converse parallel runtime system.
  • Spatial data decomposition for limited communication pattern.
  • Message driven execution for latency tolerance on commodity networks.
  • Measurement-based load balancing for scaling to thousands of processors.
  • Largest simulation to date is over 100,000,000 atoms on 300,000 cores.

Historical Version Highlights

NAMD 2.9 New Features (Apr 2012)

  • Improved (temperature/Hamiltonian) replica-exchange implementation
  • Replica-based umbrella sampling via collective variables module
  • Optimized shared-memory single-node and multiple-node CUDA builds
  • CUDA GPU-accelerated generalized Born implicit solvent (GBIS) model
  • CUDA GPU-accelerated energy evaluation and minimization
  • Native CRAY XE/XK uGNI network layer implementation
  • Faster grid forces and lower-accuracy "lite" implementation
  • Hybrid MD with knowledge-based Go forces to drive folding
  • Linear combination of pairwise overlaps (LCPO) SASA for GBIS model
  • Weeks-Chandler-Anderson decomposition for alchemical FEP simulations
  • Collective variables module improvements
  • Updates to CUDA 4.0 and Tcl 8.5.9, plus option to build with FFTW 3
  • Enhanced performance and scalability

NAMD 2.8 New Features (May 2011)

  • Generalized Born implicit solvent model
  • Accelerated molecular dynamics method
  • MARTINI residue-based coarse-grain forcefield
  • Non-uniform grids in grid forces
  • Symmetry and domain restraints
  • Collective variables module improvements
  • Force output and trajectory files
  • Shared-memory single-node and multiple-node builds
  • Measurement-based grain-size adjustment in load balancer
  • Experimental memory-optimized version with parallel I/O
  • Microsoft Windows HPC Server port and released binaries
  • Support for NBFIX parameters in CUDA builds
  • Enhanced performance and scalability

NAMD 2.7 New Features (Oct 2010)

  • Collective variable-based calculations
  • Improved free energy methods for alchemical transformations
  • Grid-based forces and molecular dynamics flexible fitting
  • Additional bonded terms for restraining molecular structure
  • Support for TIP4P water model
  • Support for CHARMM Drude polarizable force field
  • Support for C2 long-range electrostatics splitting function
  • Support for VDW force switching and long-range tail corrections
  • NVIDIA CUDA GPU acceleration of nonbonded force evaluation
  • Direct (non-MPI) support for InfiniBand via OFED ibverbs library
  • Use of mpirun to launch non-MPI (network or ibverbs) binaries
  • Enhanced performance and scalability

NAMD 2.6 New Features (Aug 2006)

  • Ports to Itanium, Altix, and Opteron/Athlon64/EMT64.
  • Port to Mac OS X for Intel processors.
  • Ports to Cray XT3 and IBM BlueGene/L (source code only).
  • Improved serial performance, especially on POWER and PowerPC.
  • Adaptive biasing force free energy calculations.
  • Customizable replica exchange simulations.
  • Tcl-based boundary potentials.
  • Reduced memory usage for unusual simulations.
  • Support for CHARMM 31 stream files and CMAP crossterms.
  • Support for OPLS force field.

NAMD 2.5 New Features (Sep 2003)

  • Improved parallel scaling and serial performance.
  • Trajectory reading and interaction energy analysis.
  • Improved constant pressure simulation and coordinate wrapping.

NAMD 2.4 New Features (Mar 2002)

  • Greatly improved parallel scaling with particle mesh Ewald.
  • Locally enhanced sampling via multiple non-interacting images.
  • Alchemical free energy perturbation for mutation, ligands, etc.
  • GROMACS ASCII topology and coordinate input file compatibility.

NAMD 2.3 New Features (Aug 2001)

  • AMBER file compatibility (parm and coordinate input only).
  • The new psfgen tool for building PSF structure files.
  • Simpler to run on a single workstation. (No more rsh!)
  • New ports to the Compaq AlphaServer SC, Scyld Beowulf, and Mac OS X.
  • Improved serial performance, particularly with PME on Alpha.

NAMD 2.2 New Features (Sep 2000)

  • New ports to the IBM RS/6000 SP and Windows NT.
  • Parallelized particle mesh Ewald FFT and reciprocal space sum.
  • Release binaries contain FFTW (under special license).
  • Much faster minimizer based on conjugate gradient method.
  • Improved load balancer with scaling to over 1024 processors.

NAMD 2.1 New Features (Nov 1999)

  • Tcl scripting language interface and config file parsing.
  • Mollified impulse multiple timestepping method.
  • Faster particle mesh Ewald implementation.
  • Periodic boundaries for non-orthogonal cells.
  • New interactive molecular dynamics interface to VMD.

NAMD 2.0 New Features (Mar 1999)

  • Supports periodic and non-periodic MD simulations
  • Can use DPME for full electrostatics for periodic simulations.
  • Triple timestepping
  • Rigid bonds to hydrogen atoms.
  • Fixed atoms (Atoms which are constrained not to move do not have forces calculated for them).
  • Berendsen and Langevin piston constant pressure methods
  • Steered Molecular Dynamics (SMD)
  • Ability to read CHARMM format parameter files.

NAMD 1.5 Features (Sep 1998)

Many of the statements below do not apply to verisions 2.0 and higher. For example, Charm++/Converse has replaced PVM as the parallel communication library and particle mesh Ewald (PME) has replaced DPMTA as the full electrostatics algorithm of choice. Input file formats and configuration parameters have, however, remained mostly compatible across this transition.

  • Efficient full electrostatics. NAMD incorporates the Distributed Parallel Multipole Tree Algorithm (DPMTA) developed by the Scientific Computing Group at Duke University to provide full electrostatic interactions in O(N) time. To further reduce the computational cost, DPMTA is integrated using a multiple timestep integration scheme which computes full electrostatic interactions only periodically during the simulation.
  • Scalable parallelism, to simulate large systems using many processors. Efficient parallel design uses a spatial decomposition scheme combined with multi-threaded, message-driven execution to achieve load balance and overlap of communication with computation.
  • Modifiable, to enable researchers to experiment with new algorithms and techniques. The design and implementation of NAMD is fully documented in the NAMD Programming Guide. NAMD has an object-oriented design implemented in C++ to help achieve the highest degree of modularity and data abstraction.
  • Portable, to allow NAMD to run across a variety of platforms. For communication, NAMD uses PVM (Parallel Virtual Machine) from Oak Ridge National Laboratory, which has itself been ported to most architectures. Porting NAMD is then simply a matter of having PVM and a reasonable C++ compiler. We have successfully ported NAMD to all of our UNIX machines, which include HP, SGI, Sun, and Linux, both single processor and shared memory multiprocessor.
  • Compatibility with X-PLOR. The input and output files for NAMD are identical to those used by the program X-PLOR. Thus, simulations can easily be migrated between the two packages, allowing the output of NAMD to be analyzed using X-PLOR or any other tool built for these file formats.
  • Implementation of standard molecular dynamics features such as energy minimization, velocity rescaling, spherical harmonic boundary conditions, harmonic atom restraints, and Langevin dynamics.