Benchmark calculations have been performed for the conventional Canadian deuterium uranium (CANDU) core analysis code RFSP and the Monte Carlo code MCNP-4B using experimental data from the deuterium critical assembly. The benchmark calculation was carried out for the effective multiplication factor (keff), void reactivity, local power peaking factor (LPPF), and power distribution of a uniform core with 1.2 wt% UO2 and two-region cores with PuO2-UO2 fuels. The RFSP calculation was performed with two energy groups, using lattice parameters generated by WIMS-AECL with the ENDF/B-V cross-section library. The RFSP calculation has shown that the root-mean-square (rms) errors of the keff and the void reactivity are within 0.6% k and 0.3% (1/k), respectively. The MCNP simulation was performed using a fully heterogeneous core model that explicitly describes the individual fuel rod and channel. The simulation showed an excellent agreement for the keff against the measurement, while the rms error of the void reactivity was 0.4% (1/k). The LPPF and core power distribution estimated by both codes matched those of the measurements within 4 and 9%, respectively. Conclusively, the physics analysis by the RFSP code in conjunction with the WIMS-AECL produces credible results for the light water-cooled and heavy water-moderated system. In addition, the MCNP-4B code has proved its potential as a computational benchmarking tool for the heavy water-moderated system.