Fast and accurate ray tracing in heterogeneous media is necessary in many seismological analyses. In this paper, we present some comparative results obtained from applying two different pseudo-bending raytracers to an actual 3-D velocity model of north-western Italy. This area is characterized by strong lateral velocity variations and major anomalies, related to the so-called Ivrea Body, lead to velocity perturbations greater than 10%. The ray tracing algorithm implemented in the Simulps12 Code (Thurber [J. Geophys. Res. 88 (1983) 8226]; and Um and Thurber [Bull. Seismol. Soc. Am. 77 (1987) 972]) and a ray tracing scheme based on that of Prothero et al. [Bull. Seismol. Soc. Am. 78 (1988) 1190] were compared, adopting the same velocity interpolation scheme and the same integration parameters. The data set was composed of more than 8700 P-phases and 7500 S-phases and the performances of the two methods were evaluated in terms of travel time difference and ray path geometry. The results showed that the scheme based on that of Prothero et al. [Bull. Seismol. Soc. Am. 78 (1988) 1190] nearly always produced smaller travel time values. In particular, the travel time differences were greater than 0.05 s for 33.2% of the whole S-phases data set whereas, in the case of P-phases, these differences corresponded to 18.9%. Considering rays relevant to hypocenters further than 75 km, we obtained travel time differences greater than 0.05 s for 75.9% of S-phases and 47.1% of P-phases. Furthermore, travel time differences appeared strongly related to the distribution of velocity heterogeneities. Indeed, for all the considered stations, the most significant travel time differences (in some cases greater than 0.15 s) were found to correspond to the azimuthal sector embracing the Ivrea Body. In the case of incomplete azimuthal coverage, these differences had a significant influence on the location procedure. Moreover, the path differences can induce a notable bias in the tomographic procedure. The results of this study had to be taken into account in its companion paper on the reliability of earthquake location procedures (Spallarossa et al. [Phys. Earth Plant. Interiors (2000)]).