Motivation for this benchmark

While such information can be useful in guiding researchers as to which programs are best suited for calculations on particular hardware platforms, other factors including flexibility, functionality and availability of a particular program enter into consideration as well. At the National supercomputer sites, researchers are interested in getting the most out of their allocation and getting to their answers in the least amount of time. Thus, consideration of factors like system size, the likely number of nodes to be accessed in a given calculation (often based on queuing priorities and availability of cycles) and overall time to solution are key to choosing the best solution, e.g., simple scaling comparisons when parallel efficiency is low, or overall time to solution is longer than on less scalable codes does not represent an optimal choice.

Our benchmark represents a typical simulation of a bio-macromolecule in water, of the type and size that is most commonly studied at the present time. We have tabulated performance data for three major MD programs: CHARMM, AMBER, and NAMD on a range of representative computer architectures. All input files are provided so that the timings can be reproduced and so that any other programs can be tested for the same MD parameters.

Description of benchmark

• 23,558 atoms
  • protein: 159 residues, 2489 atoms
  • water: 7023 molecules TIP3P, 21,069 atoms
• Cubic periodic box, 62.23 Å dimension
• 9Å nonbond cutoff with 2Å buffer, i.e., list with 11Å cutoff
• 1 fs timestep, 1000 steps
• NVE ensemble (constant energy, constant volume); CHARMM was run with NVT using Nose'-Hoover
• bonds to hydrogen constrained (SHAKE)
• PME electrostatics with 64x64x64 grid
• equilibration temperature was 300K

Program information

• CHARMM version c29b1
• Amber version 7
• NAMD version 2.4 except on the beowulf cluster where the beta release version 2.5 was used. Timings for this benchmark for NAMD 2.5b are also available for some other machines (NAMD Benchmark page ). These indicate that NAMD 2.5b is faster than NAMD 2.4 for single processors, but has about the same performance for larger numbers of processors.