The present paper examines the relevance of grid and simulation accuracy of hypersonic CFD in terms of hypersonic sharp double-cone flow. The flow grid and normal grid each adopted 250×100, 500×100, 1000×100, 500×200, 1000×200, 1000×400 and so on grids. When the normal grid was 100, the wall pressure and heat flux distribution obtained from flow grid 500 and 1000 were consistent, indicating that the solution of flow grid convergence was obtained. However, some difference was observed when the separation zone was compared with the experimental data. In increasing the normal grid number and adopting grid 500×200, the position of the separation point, wall pressure and heat flux peak was shown to be consistent with the experiment. When the grid was further encrypted, the calculation using grid 1000×200 and 1000×400 was equal to that using grid 500×200. The simulation of hypersonic sharp double-cone flow also showed that when the separation zone of the simulation was less than the experimental measurement, the wall pressure and heat flux peak moved forward. This is because the backwardness of the intersection of the separation shock and the first shock resulted in the forwardness of the intersection of the first shock and the second shock after interference, making the work region of the induction shock and boundary layer move forward. The key challenge in achieving the correct simulation of the hypersonic sharp double-cone flow is explained as follows: the algorithm can not only capture shock wave strength correctly and give the adverse pressure gradient formed by the interfering shock wave near the wall accurately. It can also prevent the numerical dissipation of the algorithm from affecting the simulation accuracy of the viscous boundary layer to ensure the correct prediction of the size of the separation zone.