To analyze physical behaviors of a thermal environment, we have to simulate several heat transfer phenomena such as heat conduction, convection, and radiation. Among those phenomena, radiative heat transfer simulation is much time-consuming. In this paper, therefore, one of acceleration techniques developed in the graphics community that exploits a graphics processing unit (GPU) is applied to the basic radiative heat transfer simulation. Implementation of the simulation on GPU makes GPU’s computing power available for the most time-consuming part of the simulation, calculation of form factors between surfaces. This paper improves the computational accuracy of the radiative heat transfer simulation running on GPU, and then examines its performance, in terms of the trade-off between the execution time and computational accuracy. The experimental results clearly show that GPU co-processing can significantly accelerate the form factor calculation. Therefore, the GPU implementation is a promising approach to acceleration of the radiative transfer simulation, especially in the case where the form factor matrix becomes too large to be stored in the main memory.