Earth-reflected Global Navigation Satellite System (GNSS) have become an attractive tool to be employed in low-Earth-orbit spaceborne missions, for applications such as ocean mesoscale altimetry and scatterometry. For such techniques, several on-board processing strategies have been proposed, either based on the correlation with on-board generated signal or based on 'blind' interferometric processing, which involves the correlation between received direct and reflected signals. This paper will provide a comparison of these two proposed GNSS-R processing techniques, highlighting the possible achieved performance for a typical spaceborne scenario. The performance comparison will be carried out by analyzing the widely used Cramer-Rao Bound, which takes into account the full statistical properties of the reflected signals and provides accurate comparison.