Fluid Cube L
This benchmark mirrors the Fluid Cube S benchmark in all aspects except for the particle radius, which has been decreased to 0.0045 and the fluid model, that was doubled in all axis. This adjustment was made to increase the number of particles within the simulation, resulting in a total of 1,366,204 fluid particles and 410,139 boundary particles. The purpose behind this modification is to evaluate how the execution times evolve with increasing number of vCPUs for higher particle densities.
The original input files can be found in our tutorial.
Software versions
The software versions used in this benchmark were the following:
| Component | Version |
|---|---|
| OpenMP | 201511 |
| OpenFAST | v3.5.2 |
| gcc, g++, gfortran | 11.4.0 (Ubuntu 11.4.0-1ubuntu1~22.04) |
| kernel | 6.6.12-linuxkit |
Execution times comparison
Faster execution times are better. Units in minutes.Cost vs Time
The following plot shows how the execution times and costs vary with
the number of cores for c2 and c2d families. The c2d-standard
machines provide the best performance for their cost.