Graphics processing unit is in molecular dynamic simulations

Mathematics. Physics. Mechanics


Semenov S. A.

Moscow Aviation Institute (National Research University), 4, Volokolamskoe shosse, Moscow, А-80, GSP-3, 125993, Russia



Demonstration of graphics processors usage in molecular-dynamic simulations.
Graphics processing unit (GPU) was originally developed for rendering real-time effects and graphical data in computer games. Now manufacturers provide a general purpose development kit. As an example, Brenner's potential is used to simulate carbon systems. To quickly perform the calculations of the equations, the following algorithms of parallel computing are used:
Mesh motion model; parallel integration; define hash function
Methods of scan; parallel potential computing
Results show that the usage of GPU decreases the time of simulation from days to a few minutes with less power consumption. General speed acceleration for optimized algorithms for NVidia Quadro 4000 graphics card was about 300 times in comparison with Core 2 Duo 2.6GHz:
GPU provided ~x10 acceleration,
Cellular flow model ~x10 acceleration,
The separation of the interaction potential of the individual parts ~ x3 acceleration.
185 microseconds are consumed for calculation on hardware Intel Core i3 (4 cores), 2.93GHz, 4GB RAM, nVidia GeForce GTX 480 at each step. In long-term calculation, 3 minutes of machine time were spent to simulate the molecular dynamics of one nanosecond of the "life" of the carbon nanotube.
The present study provides a starting point for further research in the molecular dynamic modeling of hydrocarbons. The software made gives an opportunity to research new undiscovered structures.
The research has proven to be useful in structuring data in molecular dynamics and computer testing on graphics units. This is a notable and promising side-effect of the exploratory studies, at least from an instrumental point of view.


graphics processing unit, videochips nVidia, molecular-dynamic modeling, nanomaterials, buckyball


  1. Zaliznyak V.E. Osnovy vychislitel'noi fiziki. Chast' 2. Vvedenie v metody chastits (Fundamentals of Computational Physics. Part 2. Introduction to the methods of the particles), Moscow – Izhevsk, 2006, ch.2, pp 133-156, (432p).
  2. Plimpton S. Fast parallel algorithms for short-range molecular dynamics, J. Comp. Phys., 1995, vol.117, pp.1-19.
  3. Berendsen H.J.C, D. van der Spoel, R. van Drunen. GROMACS: a message-passing parallel molecular dynamics implementation, Comp. Phys. Commun., 1995, vol.91, pp. 43-56.
  4. Phillips J.C., Braun R., W. Wang W., Gumbart J., Tajkhorshid E., Vill E., Chipot C., Skeel R.D., Kale´ L.,Schulten K. Scalable molecular dynamics with NAMD, J. Comp. Chem., 26 2005, vol.26, pp. 1781–1802.
  5. Kovalev V.L., Yakunchikov A.N. Fuhuo L. Tangential momentum and thermal accommodation coefficients for hydrogen molecules on graphite surface, Acta Astronautica, 2011,vol.69, no.7-8, pp. 744-746.
  6. Brenner D. W., Shenderova O. A., Harrison J.A. A second-generation reactive empirical bond order (REBO) potential energy expression for hydrocarbons, J. Phys.: Condens. Matter, 2002, vol.14, pp.783-2002.
  7. Yin M.T., Cohen M.L. Will Diamond Transform under Megabar Pressures?, Physical Review Letters, 1983, vol. 50, no.25, pp.2006-2009.
  8. Allen Michael P. Introduction to Molecular Dynamics Simulation Computational Soft Matter: From Synthetic Polymers to Proteins, Lecture Notes, John von Neumann Institute for Computing, NIC Series, 2004,vol. 23, pp. 1-28.
  9. Yin M.T., Cohen M.L. Structural theory of graphite and graphitic silicon, Physical Review B., 1983, vol. 29, no.12, pp.6996-6998.
  10. NVIDIA CUDA, available at:, 2012.
  11. NanoHive-1, available at:
  12. Kai Nordlund, “Basics of md,initialization,time step choice,speedup methods,” available at:

Download — informational site MAI

Copyright © 2000-2021 by MAI