In this study, the creep crack growth (CCG) properties and fracture mechanism of a Cr-Mo-V steel at 566 C in compact tension (CT) specimens were investigated, and the CCG rate was predicted by using the NSW model. The results show that the CCG rate measured by CT specimens is much lower than that predicted by the NSW model under plane-strain state. This means that the NSW model prediction for the CCG rate of the steel is over-conservative. In addition, the CCG rate da/dt versus C measured by the experiments shows the piecewise linear relation on log-log scale instead of a single linear relation predicted by the NSW model. The main reasons for these results are that the actual creep fracture mechanism of the steel and the actual creep crack tip stress field in the CT specimens have not been fully captured in the NSW model. The experimental observation shows that the creep crack propagates in a discontinuous way (step by step) at meso-scale, and the cracks at micro-scale are usually formed by the growth and coalescence of voids on grain boundaries. The NSW model based on the creep ductility exhaustion approach may not correctly describe this creep fracture process. In addition, the opening stress and triaxial stress ahead of crack tips calculated by three-dimensional finite element method is lower than those predicted by the HRR stress field which is used in the NSW model under plane-strain state. The use of the high HRR stress field will cause high CCG rates. The change in the creep fracture mechanism at micro-scale in different ranges of C may cause the piecewise linear relation between the da/dt and C . Therefore, it is necessary to study the actual CCG mechanism in a wide range of C and the actual creep crack tip stress field to establish accurate CCG prediction models. In this study, the creep crack growth (CCG) properties and fracture mechanism of a Cr-Mo-V steel at 566 C in compact tension (CT) specimens were investigated, and the CCG rate was predicted by using the NSW model. The results show that the CCG rate measured by CT specimens is much lower than that predicted by the NSW model under plane-strain state. This means that the NSW model prediction for the CCG rate of the steel is over-conservative. da / dt versus C measured by the experiments shows the piecewise linear relation on log-log scale instead of a single linear relation predicted by the NSW model. The main reasons for these results are that the actual creep fracture mechanism of the steel and the actual The experimental observation shows that the creep crack propagate in a discontinuous way (step by step) at meso-scale, and the cracks at micro-scale are usually form ed by the growth and coalescence of voids on grain boundaries. The NSW model based on the creep ductility exhaustion approach may not correctly describe this creep fracture process. In addition, the opening stress and triaxial stress ahead of crack tips calculated by three-dimensional finite The use of the high HRR stress field will cause high CCL rates. The change in the creep fracture mechanism at micro-scale in different ranges of C may cause the piecewise linear relation between the da / dt and C. Thus, it is necessary to study the actual CCG mechanism in a wide range of C and the actual creep crack tip stress field to establish accurate CCG prediction models .