To understand ship roll mechanisms and explore effects of forward speed on roll damping, roll free decay process of the DTMB5512 ship model was simulated. The reliability of different roll motion analysis methods was also evaluated according to short-term prediction results. By employing overset grid technology in computational fluid dynamics (CFD) and releasing roll, sway and heave degrees of freedom, roll free decay curves at various speeds and initial heel angles were obtained. The simulated roll amplitudes and periods were in good agreement with experimental results, confirming the reliability of the numerical method. Linear and quadratic roll damping coefficients were derived from fitting of roll free decay curves. Furthermore, roll single significant values were predicted at various sea states based on potential theory. The results indicate that as the Froude number (Fn) increases from 0.138 to 0.410, the roll decay rate significantly accelerates, accompanied by an enhanced wave elevation around the hull. At the same speed, with initial heel angle increasing from 10° to 20°, there shows no obvious differences in linear damping coefficients. For the same initial heel angle, higher speeds corresponds to larger linear damping coefficients. The equivalent damping is derived from combination of linear and quadratic damping at corresponding roll amplitude, which can effectively weaken the interference from roll extinction curve fitting. According to short-term roll prediction results, for speeds corresponding to Fn=0.138 and Fn=0.280, roll damping coefficients obtained from CFD method are both more accurate than that from empirical formula method, especially obvious at higher sea states. The equivalent damping coefficients are more suitable for quantitative analysis of roll motion, increasing with speed.