Continuous gravity observations of ocean and solid tides are usually done with land-based gravimeters. In this study we analyze a 426-day record of time-varying gravity acquired by an ocean-bottom Scintrex spring gravimeter between August 2005 and November 2006 at the Troll A site located in the North Sea at a depth of 303 m. Sea-bottom pressure changes were also recorded in parallel with a Paroscientific quartz pressure sensor. From these data, we show a comparison of the noise level of the sea-floor gravimeter with respect to two standard land-based relative gravimeters: a Scintrex CG5 and a GWR Superconducting Gravimeter that were recording at the J9 gravimetric observatory of Strasbourg (France). We also compare the analyzed gravity records with the predicted solid and oceanic tides. The oceanic tides recorded by the seafloor barometer are also analyzed and compared to the predicted ones using FES2014b ocean model. Observed diurnal and semi-diurnal components are in good agreement with FES2014b predictions. Smallest constituents reflect some differences that may be attributed to non-linearity occurring at the Troll A site. Using the barotropic TUGO-m dynamic model of sea level response to ECMWF atmospheric pressure and winds forcing, we show a good agreement with the detided ocean-bottom pressure residuals. About 4-hPa of standard deviation of remaining sea-bottom pressure are however not explained by the TUGO-m dynamic model.