New computational methods in dose assessment and shielding calculations have drastically increased possible accuracy and resolution of the solution, while also increasing both memory demand and running time. In many cases, a trade-off must occur between these two parameters due to limited computational resources. This becomes prominent, particularly in hybrid deterministic-stochastic methods used for automated variance reduction, where the trade-off is additionally sought between the importance-generating deterministic portion and actual Monte Carlo simulations. This technical note examines this trade-off for the FW-CADIS methodology implemented in the MAVRIC (Monaco with Automated Variance Reduction using Importance Calculations) module of SCALE6, applying it to a simplified model of a power reactor. For the purposes of this study, the allowed total CPU time was held constant (12 and 48 h). It was found that improving the accuracy of the deterministic portion (within the single-processor limitation of the program version used) at the cost of reducing the available time for Monte Carlo was beneficial for the overall efficiency. While the analysis is specific to the selected problem, it is expected that the findings in a broader sense are relevant for other similar hybrid shielding methodologies and applications.