A system and method providing timing alignment of a data mask (DM) signal with respect to a data strobe (DQS) signal for memory devices not designed for adjusting such alignment is provided. Alignment between data signals (DQ) and a DQS signal is first achieved during a first write training procedure where a data delay value is optimized for one of the DQS or DQ signals. Subsequently, using the optimum delay value from the first write training procedure, a second write training procedure is initiated. In the second write training procedure, timing alignment between the DM signal and the DQ signals is achieved by determining an optimal delay value of the DM signal relative to the DQS signal.