The Princeton Ocean Model (POM) is widely used in both the research and real-time marine forecast communities. However, there is a single heavily criticized feature and shortcoming of POM, i.e., the pressure gradient (PG) error associated with sloping topography. To overcome this problem, in this paper we present an eta-coordinate POM (ECPOM). The eta-coordinate system was originally adopted in atmospheric numerical models and has proved to be effective in removing the large errors associated with the horizontal pressure gradient force and advection and diffusion terms along a steeply sloping topography. A familiar idealized seamount problem is used to test the PG error in the ECPOM in order to isolate errors from other sources, as is done in the literature. The model is set up so that, ideally, the ocean should be at rest forever; any development of circulation is regarded as numerically artificial. Results show that in ECPOM, errors are dramatically reduced by two orders of magnitude over a wide range of parameters. As a trade-off, the bottom boundary of the ECPOM is a stepwise instead of a continuous one, as in the sigma-coordinate POM (SCPOM). Modification of the SCPOM to the ECPOM is straightforward and minimal. Compared to the z-coordinate model, the ECPOM is numerically simpler since the eta-coordinate has a surface-followlng character.