Performance Comparison

The software is available for Windows as well as for Linux OS and in both systems the optimisation can either be performed on CPU as well as on GPU (Nvidia Quadro). For the GPU optimisation a separate license is required.

The following comparison of a Jet Engine Bracket optimisation shows the difference between the systems and various combinations of CPU and GPU calculations. The Jet Engine Bracket was input with the same boundary conditions as in the tutorial. The most important factors regarding the performance are the number of Events (load cases) and the Complexity setting. In this case there are 4 Events and a Complexity of 14 was chosen.

Open

The difference between an optimisation on a GPU and a CPU is very dependent on the power and generation of the GPU/CPU.

The best performance is achieved on a Linux server with 3 GPU Quadro RTX6000 Cards. The difference of 1 minute to the 2 GPUs probably lies within other processes on the system. In general for 4 Events, 2 GPUs are very good so that 2 Events can be calculated simultaneously. Especially regarding multiple Events and higher Complexity values the graphics cards make sense in terms of performance which can be seen clearly in the performance comparison. Even the slowest GPU is still faster than the best CPU in this test. In the CPU section it can be observed that the Xeon and the newer generation of i7 CPUs are the fastest. Particularly the last iterations on the highest resolution levels take more time on CPUs in comparison to GPUs. Regarding the CPUs, the generation as well as the speed is of importance for the performance.