Potts model was often used to simulate grain growth without necessary evaluation of the effect of lattice scale and simulation temperature. It is found in this paper that such parameters may affect the simulation results markedly. The results show that simulations at zero temperature or on a small scale lattice (say, the number of sites on one edge of the square lattice L=1000) cannot reach the steady-state period of grain growth, while large-scale simulations (say, L=2000) at a much higher simulation temperature can. The steady-state grain size distribution so obtained may be well described by Weibull function other than log-normal or Rayleigh functions.