In this work, the performance of the USP (1010) concept for comparing precision has been investigated. A diagram has been constructed to relate common variance ratios, sample sizes and the corresponding powers. The choice of the upper acceptable limit of variance ratios strongly influences the power. Small upper limits, such as 2.25, are not practical. The proposed upper limit of 4 requires sample sizes of 14 or higher to achieve a power of 80%. If the precision of a method is very good, higher ratios seem to be acceptable, with a significant reduction in measurements. For example, using n = 6 is sufficient to obtain a power of 90%, if the variances are in fact the same and the acceptable variance ratio is 16.