Experiments at constant normal stiffness (CNS) are normally carried out to understand underground shear processes of rock joints. However, in many test setups the available space around the joint is limited implying it is not possible to measure the dilatancy directly over the joint. Therefore, the displacement transducers must be in locations where the risk is that additional displacements originating from deficiencies in the test system will be measured causing too low normal loads to be applied. Herein, this issue is investigated in a new 5 MN direct shear test setup. The system normal stiffness was found to be about 11 300 kN/mm derived from normal loading up to 4.5 MN using a steel specimen. The direct shear testing performance under the CNS configuration was evaluated using the steel specimen, which had a joint with a known angle of inclination. The normal load error at 3.9 MN (28 MPa) was 11%, but by application of the effective normal stiffness approach using the system normal stiffness as input the error basically could be eliminated. The results demonstrate the robustness of the setup designed for joint areas up to 400 × 600 mm with normal and shear loads up to 5 MN.