The aim of this study is to investigate the use of computational fluid dynamics in predicting the performance and geometry of the optimal design of a steam ejector used in a steam turbine. Many scholars have analysed the steam ejector using the ideal gas model, which lacks accuracy in terms of calculating the flow field of the ejector. This study is reported in a series of two papers. The first part covers the validation of CFX 11.0 results using different equations of state (EOS) on the converging–diverging nozzle flow field carried out with the experimental value. The IAPWS IF97 real gas model works well with the experimental value. The flow field of the ejector was analysed using different EOS after grid-dependent learning. The results show that the performance of the ejector was underestimated under the ideal gas model; the entrainment ratio was 20–40 per cent lower than when using the real gas model. The effect of the optimal geometrical design and operating conditions will be discussed in Part 2.