We present a detailed comparison of three different simulations of the epoch of reionization (EoR). The radiative transfer simulation (c2-ray) among them is our benchmark. Radiative transfer codes can produce realistic results, but are computationally expensive. We compare it with two seminumerical techniques: one using the same haloes as c2-ray as its sources (Sem-Num), and one using a conditional Press–Schechter scheme (CPS+GS). These are vastly more computationally efficient than c2-ray, but use more simplistic physical assumptions. We evaluate these simulations in terms of their ability to reproduce the history and morphology of reionization. We find that both Sem-Num and CPS+GS can produce an ionization history and morphology that is very close to c2-ray, with Sem-Num performing slightly better compared to CPS+GS. We also study different redshift-space observables of the 21-cm signal from EoR: the variance, power spectrum and its various angular multipole moments. We find that both seminumerical models perform reasonably well in predicting these observables at length scales relevant for present and future experiments. However, Sem-Num performs slightly better than CPS+GS in producing the reionization history, which is necessary for interpreting the future observations. The CPS+GS scheme, however, has the advantage that it is not restricted by the mass resolution of the dark matter density field.