Lumerical FDTD simulations run on CPU. Since GPUs are more powerful in terms of processing, can we use GPUs by any means for the simulation? That would immensely reduce the computation run time.
Dear Manoranjan Minz , according to Ansys, Lumerical doesn´t support GPU acceleration and for improving performance they recommend multiprocessor CPU´s because each processor has its own memory bus connection to the RAM.
https://forum.ansys.com/discussion/27476/computer-suitable-for-lumerical-simulation
This said, it doesn´t mean that you cannot use GPU for FDTD calculations, in fact this is possible and there is plenty of information available. The main idea of using this acceleration process is based in the fact that any volume data is stored in the memory as a single row of data, and that vector could be processed with GPU.
Have a look
https://www.lumerical.com/learn/webinar/using-high-performance-computing-to-accelerate-your-fdtd-simulations/
Like others have pointed out. Lumerical doesn't support GPU acceleration. Processors with more cores and RAM will perform better when compared with low configs.
But, Lumerical supports running simulations in cloud-based platforms.
- Badges
- Science topic
- Similar topics
- Academic Journals