Learning Forum

[lin.deng]

We have a license for the ANSYS Academic Lumerical Research (5 tasks), and according to the description on the website, we should have a quantity of 5 for each feature in the "ANSYS Academic Lumerical Research (1 task)" product. Each individual task allows us to have four lum_fdtd_solve, and each of these permits the use of 16 SMs (Streaming Multiprocessors). Does this mean that, in principle, the maximum number of SMs we are allowed to use simultaneously on the GPU is 5 * 4 * 16 = 320 SMs? Of course, the actual GPU we use will have fewer SMs than this number. Additionally, I noticed that "Time monitors limit GPU performance. We suggest using them only for debugging and preliminary simulations." Since I need to use the time monitor to obtain the transmission spectrum, does this mean we cannot use the GPU to assist with the related simulations? I would greatly appreciate it if someone could provide an answer!

[Lito]

Yes, ANSYS Academic Lumerical Research (5 tasks) allows you to run on a GPU up to a maximum of 320 SMs. 

You can use time monitors but this will affect the simulation speed/performance. You might not be getting the GPU performance as compared to running on CPU with time monitors.