You are here: Home / Occam / Benchmarks

Benchmarks

This page lists OCCAM performance on several benchmark problems.


The code description and the parallel implementation of OCCAM MD can be found in the paper: Zhao Y., De Nicola A., Kawakatsu T., Milano G. ,Hybrid Particle-Field Molecular Dynamics Simulations: Parallelization and Benchmarks, Journal of Computational Chemistry2012, doi:10.1002/jcc.22883 - (pdf

 

 home occam home | Project Project | Benchmarks Benchmarks | documentations Documentation

 

  • MONATOMIC FLUID SYSTEMS 
Benchmarks of parallel MD-SCF program with speedup: 

a) MF1 (blue curves with squares) and MF2 (green curves with triangles). For both systems different update frequencies of 100 (empty symbols and dash dot lines) and 300 time steps (filled symbols and solid lines) have been considered.Ideal speedup (red line with circle).

b) Performances of parallel MD-SCF program as steps/s for MF3 system in comparison with GROMACS 4.5.4 (orange curve). Results of OCCAM using 39,304 lattice points (purple curves) and 238,328 lattice points (red curves) are shown.

Particle-field MD simulations, performed with OCCAM MD, have been done using update frequency of 100 time steps (empty symbols and dash dot lines) and 300 time steps (filled symbols and solid lines), respectively. 

SystemNo. of ParticlesNo. of lattice pointsBox size (nm3)
MF1 50000 29x29x29 (24389) 20.0x20.0x20.0
MF2 100000 37x37x37 (50653) 25.2x25.2x25.2
MF3 500000

34x34x34 (39304)

62x62x62 (238328)

43.1x43.1x43.1

bnchk_1 

 

  •  LIPID/WATER and ROD/COIL HOMOPOLYMERS MIXTURE SYSTEMS 
Benchmarks of parallel MD-SCF program with speedup: 

a) Lipid/water SLW1 (blue curves with squares) and SLW2 (green curves with triangles). For both systems different update frequencies of 100 (empty symbols and dash dot lines) and 300 time steps (filled symbols and solid lines) have been considered.Ideal speedup (red line with circle).

b) Rod/Coil Homopolymers mixture SRC1 (yellow curves with circles) and SRC2 (purple curves with triangles). For both systems different update frequencies of 100 (empty symbols and dash dot lines) and 300 time steps (filled symbols and solid lines) have been considered.Ideal speedup (red line with circle).

SystemNo. of ParticlesNo. of lattice pointsBox size (nm3)
SLW1 4096 13x13x11 (1859) 8.18x8.18x6.95
SLW2 24576 24x13x31 (9672) 16.35x8.18x20.85
SRC1 5000

12x12x12 (1728)

23.87x23.87x23.87
SRC2 30000

22x22x22 (10648)

43.36x43.36x43.46

bnchk_2

 

  •  LARGE SYSTEM MONOATOMIC FLUID
Benchmarks of parallel MD-SCF program with speedup: 

We tested the Parallelization Efficiency, CPU time spent for 10000 step employing a large number of cores. Three different systems have been considered for these tests.