We investigate the effects of different initialisation methods of the GUMICS-4 global magnetohydrodynamic (MHD) simulation to the dynamics in different parts of the Earth's magnetosphere and hence compare five 12 h simulation runs that were initiated by 3 h of synthetic data and followed by 9 h of solar wind measurements using the OMNI data as input. As a reference, we use a simulation run that includes nearly 60 h of OMNI data as input prior to the 9 h interval examined with different initialisations. The selected interval is a high-speed stream event during a 10-day interval (12-22 June 2007). The synthetic initialisations include stepwise, linear and sinusoidal functions of the interplanetary magnetic field with constant density and velocity values. The results show that the solutions converge within 1 h to give a good agreement in both the bow shock and the magnetopause position. However, the different initialisation methods lead to local differences which should be taken into consideration when comparing model results to satellite measurements.