In this talk, we shall assess the finite ma errors from the overlap fermion. We shall present results on the speed of light from the dispersion relation and hyperfine splitting between the vector and pseudoscalar mesons as a function to ma to reveal
the mLambda_{QCD}a^2 and m^2a^2 errors. We conclude from this study that one should be limited to using ma less than 0.5 in order to keep the systematic ma errors below a few percent level.
