Natural gas liquefaction systems are based on refrigeration cycles - they consist of the same operations such as heat exchange, compression and expansion, but they have different layouts, components and working fluids. The design of these systems requires a preliminary simulation and evaluation of their performance. However, the thermodynamic models used for this purpose are characterised by different mathematical formulations, ranges of application and levels of accuracy. This may lead to inconsistent results when estimating hydrocarbon properties and assessing the efficiency of a given process. This paper presents a thorough comparison of six equations of state widely used in the academia and industry, including the GERG-2008 model, which has recently been adopted as an ISO standard for natural gases. These models are used to (i) estimate the thermophysical properties of a Danish natural gas, (ii) simulate, and (iii) optimise liquefaction systems. Three case studies are considered: a cascade layout with three pure refrigerants, a single mixed-refrigerant unit, and an expander-based configuration. Significant deviations are found between all property models, and in all case studies. The main discrepancies are related to the prediction of the energy flows (up to 7%) and to the heat exchanger conductances (up to 11%), and they are not systematic errors. The results illustrate the superiority of using the GERG-2008 model for designing gas processes in real applications, with the aim of reducing their energy use. They demonstrate as well that particular caution should be exercised when extrapolating the results of the conventional thermodynamic models to the actual conception of the gas liquefaction chain. (C) 2016 Elsevier Ltd. All rights reserved.