Surface roughness affects the results of nanomechanical tests. The surface roughness values to be measured on a surface of a porous material are dependent on the properties of the naturally occurring pore space. In order to assess the surface roughness of hardened cement paste (HCP) without the actual influence of the usual sample preparation for nanomechanical testing (i.e. grinding and polishing), focussed ion beam nanotomography datasets were utilized for reconstruction of 3D (nanoscale resolution) surface profiles of hardened cement pastes. 'Virtual topographic experiments' were performed and root mean square surface roughness was then calculated for a large number of such 3D surface profiles. The resulting root mean square (between 115 and 494 nm) is considerably higher than some roughness values (as low as 10 nm) reported in the literature. We suggest that thus-analysed root mean square values provide an estimate of a 'hard' lower limit that can be achieved by 'artefact-free' sample preparation of realistic samples of hardened cement paste. To the best of our knowledge, this 'hard' lower limit was quantified for a porous material based on hydraulic cement for the first time. We suggest that the values of RMS below such a limit may indicate sample preparation artefacts. Consequently, for reliable nanomechanical testing of disordered porous materials, such as hardened cement paste, the preparation methods may require further improvement.