Display omitted •Specimens containing different artificial defects sizes were tests at RT and ET.•Fatigue life decreases significantly due to the existence of artificial defects.•The life scatter of specimen with similar defect under same condition is small.•A fracture-mechanic based model was proposed to evaluate the existence of a defect. To study the influence of defects, a crack-like artificial defect was introduced at the surface of smooth specimens of a powder metallurgy nickel-based superalloy FGH96. Two kinds of size were set and fatigue tests were conducted at both room temperature (RT) and elevated temperature (ET) with several different stress levels. Results first show that the existence of surface defects significantly reduces fatigue life and the negative effect of temperature is also evident. Meanwhile, specimens with identical conditions exhibit little scatter in life which provides the possibility to evaluate fatigue lives with a relatively deterministic method. To further reveal the failure process, both macro and micro fractography were carefully examined and specimens at RT and ET show evident differences in crack initiation and propagation. Finally, a life prediction method based on the fracture mechanic was proposed. Five stress intensity calculation methods with different strategies for the surface crack in round specimens were applied and their influence on life prediction was further discussed. Given the relative high loading condition, the plasticity modification for different method is also included. Besides, this calculation method was checked for application on other loading conditions. Finally, the final stress intensity value and the parameter sensitivity analysis were also discussed.