Feasibility Analysis of Low Cost Graphical Processing Units for Electromagnetic Field Simulations by Finite Difference Time Domain Method

TL;DR

This paper evaluates the feasibility of using a low-cost NVIDIA GT 520 GPU to implement the FDTD method for electromagnetic simulations, comparing its performance with a basic CPU.

Contribution

It demonstrates the potential of affordable GPUs for computational electromagnetics, providing benchmarking data and implementation insights.

Findings

GPU shows promising performance for FDTD simulations

Performance correlates with problem size on CPU and GPU

Low-cost GPU can be a feasible alternative for CEM computations

Abstract

Among several techniques available for solving Computational Electromagnetics (CEM) problems, the Finite Difference Time Domain (FDTD) method is one of the best suited approaches when a parallelized hardware platform is used. In this paper we investigate the feasibility of implementing the FDTD method using the NVIDIA GT 520, a low cost Graphical Processing Unit (GPU), for solving the differential form of Maxwell's equation in time domain. Initially a generalized benchmarking problem of bandwidth test and another benchmarking problem of 'matrix left division is discussed for understanding the correlation between the problem size and the performance on the CPU and the GPU respectively. This is further followed by the discussion of the FDTD method, again implemented on both, the CPU and the GT520 GPU. For both of the above comparisons, the CPU used is Intel E5300, a low cost dual core CPU.