This paper describes a creep–fatigue evaluation method for modified 9Cr-1Mo steel, which has been newly included in the 2012 edition of the Japan Society of Mechanical Engineers code for design and construction of fast reactors (JSME FRs code). In this method, creep and fatigue damages are evaluated on the basis of Miner's rule and the time fraction rule, respectively, and the linear summation rule is employed as the failure criterion. The conservativeness of this method without design factors was investigated using material test results, and it was shown that the time fraction approach can conservatively predict failure life if margins on the initial stress of relaxation and the stress relaxation rate are embedded. In addition, the conservatism of prediction tends to increase with time to failure. Comparison with the modified ductility exhaustion method, which is known to have good failure life predictability in material test results, shows that the time fraction approach predicts failure lives to be shorter in long-term strain hold conditions, where material test data are hardly obtained. These results confirm that the creep–fatigue evaluation method in the JSME FRs code has implicit conservatism in addition to explicit margins in the design procedures such as design factor.