Abstract: How to accelerate the speed of finite-difference time-domain(FDTD)algorithm is one of the most challenging tasks of FDTD research.With dramatic increase of Graphics Processing Unit(GPU)computing power and the proposal of general purpose GPU computing, GPU based computing is an inevitable trend during the past few years.This paper starts with introducing the potential and research status of GPU based FDTD computing; then illustrats the basis idea of GPU computing and compares it with other hardware accelerating approach to prove the low-lost and multi-platform property of GPU-computing. In the end, discusses the enormous potential applications and remaining problems based on current experimental results.