|
Computational performance enhancement for multiprocessor systemsDRAFT Running solver on multiple CPUs enhances the calculation performance and thus (in most cases) reduces significantly the calculation time. Multi CPU (or Multicore) behaviour of the solver is controlled by respective application settings separately. Remark: This function requires a valid license feature MULTICORE.
Complex calculations (such of more then 1 million equations) can be critical in terms of time factor - the calculation can last few seconds only but (for so called "bad conditioned equations") tenths of minutes (on currently typical PC computer hardware). The CPU speed, CPU type but also die memory architecture of the system determine the effective computational power which is available to AnTherm. By employing the MULTICORE-Option the software can access multiple CPUs oder CPU kernels and by that significantly improve the efficiency. By increasing the number of used kernels will add to the computational power as expected. Knowing that it depends on the operating efficiency of many other components of the device it is advisable to determine the optimal number of cores by executing simple measurement.
Above picture shows, that the theoretical multiplication of the computing power will not be reached by simply raising their number of cores! For the computer architecture used for the above example the saturation of computational power is reached already with four cores. This results from the technical structure of the Core i7/920 processor having only for memory paths (one for each pair of cores). In reality saturating the memory bandwidth raises the computing power limit in that case. The number of cores used by the solver shall be set in the application settings.
See also: Solver window, Application settings |
|