A model for a time-sharing operating system is developed in order to assess the effects of dynamic quantum allocation and overhead variability on central processing unit (CPU) efficiency. CPU efficiency is determined by the proportion of time devoted to user-oriented (problem state) tasks within a busy period. Computational results indicate that a dynamic quantum allocation strategy produces significant differences in CPU efficiency compared to a constant quantum. The differences are affected significantly by the variability among allocated quantum values and the demand on the system. Overhead variability also has a pronounced effect. A function that depicts overhead as decreasing with demand produces more stable values of CPU efficiency. The interaction between demand and the amount of overhead is observed to be significant.