This paper is concerned with coded multi-ary systems over linear channels. Based on a semi-analytical evolution technique, the impact of signaling schemes on the performance of low-cost iterative linear minimum-mean-square-error (LMMSE) detection is studied. It is shown that superposition coded modulation (SCM) maximizes the output signal-to-noise ratio (SNR) of LMMSE detectors. Consequently, SCM may potentially outperform other conventional signaling schemes when LMMSE detectors are used. Numerical examples are provided to verify the theoretical analysis. © 2008 IEEE.