Interference alignment (IA) has been well recognized as an efficient approach to reduce interference at high signal to noise ratio (SNR). However, it demands global channel state information (CSI) at both transmitters and receivers for precoder design in order to maximize the multiplexing gain. In this paper, we discuss IA with delayed differential CSI feedback for time-correlated multiple input multiple output 8MIMO) block fading channels. We consider the impact of distortion caused by channel estimation errors and quantized CSI feedback delay, and thus, find an optimal feedback interval to minimize this distortion, as well as the sum rate performance. And with the feedback-channel capacity constraint, we further study the relationship between the average sum rate and the feedback interval. Analytical results are verified by simulations.