Chaplot et al. (1998) purport to simulate MgSiO3-perovskite melting. Instead, the paper contains simulations of a thermal instability. One can define thermal instability of a crystal lattice as a mechanical instability induced by temperature. The temperature of thermal instability can be found by gradual increase of temperature until the sudden change of properties appears. This is exactly what was performed by Chaplot et al. The temperatures of thermal instability of the ideal lattice were high compare to experimental. Recognizing the possibility of overheating, Chaplot et al. carried out simulations to determine temperatures of thermal instability of the lattice containing 1% vacancies. The lattice with defects is less stable than the ideal lattice, therefore, the temperature of thermal instability for defect structure was lower than for the ideal lattice. Still, neither the simulation of ideal nor defect structures answered the question: What is the melting temperature of the MgSiO3-perovskite with the accepted model of interatomic interaction? Melting and thermal instability are different phenomena. Temperature of melting (Tmelt) is defined as the temperature where Gibbs free energies of crystal and liquid phases are equal. Morris et al. (1994) showed that calculating Tmelt by calculating Gibbs energies is equivalent to finding Tmelt from two-phase simulations. The question is how much melting temperatures could differ from temperatures of thermal instability calculated by Chaplot et al. for ideal lattice? An exact answer could be provided by the direct comparison of two methods, as was done by Belonoshko (1998) for melting of corundum. In absence of this comparison, we can refer to other studies to obtain an approximate values of overheating in Chaplot et al. calculations. Matsui and Price (1991) and Belonoshko (1994) have studied the melting of perovskite. Matsui and Price (1991) has calculated Tinst, whereas Belonoshko (1994) applied the two-phase method to calculate Tmelt. Interesting, an application of correct method allows to reach good agreement with experiment (Fig. 1). The same model of interatomic interaction was applied in both studies. From the comparison it follows that Tinst is higher than Tmelt by about 1500 K at the pressure of 25 GPa. Because overheating increases with pressure (Belonoshko 1998), the value of 1500 K is likely to be a lower bound for overheating.