Intelligent reflecting surface (IRS) has been deemed as a promising technology to reconfigure the signal propagation environments cost-effectively. However, due to the lack of radio-frequency (RF) chains at the IRSs, their channel estimation becomes a significant challenge. In this correspondence, we propose a novel phase hopping scheme for IRS-aided multiple-input multiple-output (MIMO) communication systems, which dispenses with the channel state information (CSI) by randomly changing the IRS's reflections over different time slots within the channel coherence time, thereby creating artificial time diversity to enhance the communication performance. Based on a minor modification on the current cellular protocol, we first design a new and compatible frame structure for the proposed phase hopping scheme. Then, we derive its ergodic capacity and outage probability in closed-form using random matrix theory and gain essential insights into its asymptotic behaviors. Numerical results validate our theoretical analysis and show the superiority of our proposed phase hopping scheme to the benchmark scheme without phase hopping.