A method by which a popular signal-to-noise ratio (SNR) estimator can be reconfigured to yield improved performance is proposed. The reconfiguration consists of partitioning the symbol interval into a larger (but even) number of subdivisions than the two that have been traditionally suggested for this estimator but still processing the observables in successive pairs. The optimum number of subdivisions is shown to depend on the SNR range in which the true SNR lies; however, we also show that these SNR regions can be significantly widened with very little loss in performance. Most important is the fact that with this reconfiguration, the SNR estimator tracks the Cramer-Rao bound on the variance of the estimator over the entire range of SNR values.