The invention discloses a reliability evaluation method and an evaluation device of a high-performance computing cluster system. The reliability evaluation method comprises the following steps: generating a state transition diagram of each subsystem of the high-performance computing cluster system; wherein, the state transition diagrams of a plurality of subsystems are different from each other; traversing the state transition diagrams of each subsystem to calculate the reliability of a high performance computing cluster system. The technical scheme of the invention controls the system risk through modularization, independently evaluates the risk of each subsystem of the high-performance computing cluster system, and realizes the accurate judgment of the global risk through the risk control chain formed by the state transition diagram.