With the increasing use of computational fluid dynamics (CFD)-based simulations in the assessment of thermal hydraulic behaviour of nuclear reactors, there is a need to benchmark the predictions of CFD codes. The case of a frictionless U-tube manometer [Ransom, V., 1992. Oscillating manometer. In: Hewitt, G.F., Delhaye J.M., Zuber N. (Eds.), Multiphase Science and Technology, vol. 6. Hemisphere, New York, USA, p. 591] has been previously used as a test case to benchmark the dispersive and dissipative characteristics of a numerical prediction. In the present paper, this case is extended to cover two new aspects that are often found in nuclear reactor systems: variable density and dissipation. Through a theoretical model, it is shown that the dynamic behaviour of the new system is more complicated than that of a constant-density frictionless U-tube manometer. Experiments of the test case have been carried out to substantiate the essential features of the response. CFD simulations of the system show both dissipative and dispersive errors compared to the theoretical result. The results can therefore serve as a useful benchmark for computer codes used to study the thermal hydraulics of nuclear reactor systems, especially in evaluating the role of numerical damping in systems in which physical damping exists inherently. © 2003 Elsevier B.V. All rights reserved.