Topics related to Lumerical and more

Lumerical AVX usage

    • Thomas Bauer

      Dear Lumerical team (& forum members),

      since there seems to be no information available in any knowledge base or in the forum, we wanted to ask if Lumerical (at least version 2023R2 and later) makes use of AVX-512 instructions or only AVX2 instructions? With the newest generation of workstation CPUs slowly incorporating AVX-512, it would be important to know if any benefits of this advanced instruction set are leveraged, or if processors with (only) AVX2 support are essentially equally powerful for FDTD and RCWA simulations.

      In addition, do we understand it correctly that for academic licenses 32 cores are the limit per license, or actually 32 threads?

      Best regards,


    • Lito Yap
      Ansys Employee

      @Thomas Bauer,

      Ansys Lumerical support AVX instructions though that’s not too critical for our performance at the moment. More important is memory bandwidth. 
          >Information on Hardware Specifications – Ansys Optics 

      1 Solve license = running up to 32 logical processors/threads on your local computer with hyperthreading. Otherwise, without hyperthreading, you will be running on the physical cores of the local machine. 
          >Resource configuration elements and controls – Ansys Optics 
          >Ansys optics solve, accelerator and Ansys HPC license consumption – Ansys Optics 

      Hope this helps.

Viewing 1 reply thread
  • You must be logged in to reply to this topic.