The response of a commercially available diode-based in vivo dosimetry system was studied over a selection of clinically relevant photon beam setups. The dosimetry system consists of a dedicated multichannel electrometer with several diode detectors differing only in their equivalent wall buildup. Each detector is calibrated for a specific nominal beam energy and used clinically with that energy only. To study dosimeter response, a diode taped to the surface of a solid water phantom was irradiated simultaneously with an end-window chamber placed at a depth of dmax inside the same phantom. Photon beams with energies of Co-60, 6 and 18 MV were used. For each beam energy, the response of the diode relative to the given dose as measured by the end-window chamber was evaluated for open and wedged fields (0 degree to 60 degrees) with source-to-surface distances (SSDs) ranging from 75 to 120 cm and collimator settings from 5 x 5 to 40 x 40 cm2. It was found that diode response, i.e., diode reading per cGy of given dose, varies significantly with treatment beam setup. For example, increasing field size for a constant SSD causes a decrease of up to 15% in diode response relative to the given dose for 6 and 18 MV beams, while for Co-60 an increase in response of up to 5% results. Furthermore, increasing SSD for a fixed collimator setting results in decreased diode response (up to 10%) for all beams. The complicated dependence of diode response on beam setup necessitates the use of empirical response curves, similar to those evaluated in this work, to accurately convert clinical dosimeter reading to dose at depth.