The signal-to-interference ratio (SIR) is a predominant criterion for radio resource management techniques in wireless networks. Many algorithms such as power control, channel allocation, and handoff, etc., require a fast and accurate SIR measurement. In practice, however, the SIR estimation needs a lot of computations. The major challenge for the SIR estimation issue is to find an efficient way to separate the desired signal with interference. We investigate the performance of the SIR estimation techniques for the DS-CDMA cellular system in a time-dispersive (or a frequency-selective) fading channel. The major contribution is to provide a framework incorporating the effect of channel dispersions into the SIR estimator for the CDMA systems. We present new results on the quantitative performance of two easily implemented SIR estimation techniques, taking into account the impact of frequency-selective fading, mobility, different number of interferers, and different initial transmission power levels.