Accurate determination of the megavoltage (MV) radiation isocentre of a linear accelerator (linac) is an important task in radiotherapy. The localization of the MV radiation isocentre is crucial for correct calibration of the in-room lasers and the cone-beam CT scanner used for patient positioning prior to treatment. Linac manufacturers offer tools for MV radiation isocentre localization. As a user, there is no access to the documentation for the underlying method and calculation algorithm used in the commercial software. The idea of this work was to evaluate the accuracy of the software tool for MV radiation isocentre calculation as delivered by Elekta using independent software. The image acquisition was based on the scheme designed by the manufacturer. Eight MV images were acquired in each series of a ball-bearing (BB) phantom attached to the treatment couch. The images were recorded at cardinal angles of the gantry using the electronic portal imaging device (EPID). Eight Elekta linacs with three different types of multileaf collimators (MLCs) were included in the test. The influence of MLC orientation, x-ray energy, and phantom modifications were examined. The acquired images were analysed using the Elekta x-ray volume imaging (XVI) software and in-house developed (IHD) MATLAB code. Results from the two different software were compared. A discrepancy in the longitudinal direction of the isocentre localization was found averaging 0.23 mm up to a maximum of 0.75 mm. The MLC orientation or the phantom asymmetry in the longitudinal direction do not appear to cause the discrepancy. The main cause of the differences could not be clearly identified. However, it is our opinion that the commercial software delivered by the linac manufacturer should be improved to reach better stability and precise results in the MV radiation isocentre calculations.