In a previous paper [3], the authors presented large sample optimum accelerated test plans to estimate the survival probability at a design stress assuming a linear logistic model, i.e. a linear relationship between stress and the log survival odds at that stress. The optimum plans required testing at two accelerated stresses with a larger allocation of units assigned to the lower of the two stresses. In practice, however, it is often desirable to conduct the accelerated tests at more than two stresses and/or use equal or otherwise prespecified allocation. In this paper we compare the large sample variance of each of ten such non-optimum test plans (and also that of testing exclusively at the design stress) with that of the optimum plan under a variety of conditions.