Epyc performance with Empire 8.1

January 2022

EMPIRE-XPU achieves new speed record on AMD Epyc: 50 billion FDTD cells per second on single-system

With over 50 billion FDTD cells per second, EMPIRE XPU sets a new speed record for single-system simulation performance on an AMD Epyc 7763 Dual-CPU server. This represents the first single-system alternative to EMPIREs Multi-PC setup, where comparable simulation performance could be demonstrated last year using four combined AMD Threadripper 3970X workstations. Such “epyc” simulation speeds save valuable time during the design of complex 3D EM models like antenna frontends and multilayer circuits. The attached picture shows the performance of different systems while simulating a multiport LTCC module including losses. It can be seen than EMPIRE XPU also performs excellent on cloud computing solutions (AWS, Azure), where custom Epyc 7003 Series CPUs like the 7V13 are available. Compared to GPU solutions the EMPIRE-XPU cluster solver is superior regarding simulation speed, simulation size and hardware costs.