The classical time transfer method used to realize International Atomic Time (TAI) is based on the common view technique, with GPS observations collected by C/A code receivers. The resulting clock offsets between the laboratory clock and GPS time are obtained from a fixed procedure defined by the Consultative Committee for Time and Frequency (CCTF). A similar procedure can be applied to the Receiver INdependent EXchange (RINEX) observation files produced by geodetic receivers driven by a stable external frequency. If the link between the receiver clock and the external clock is stable and precisely determined, the geodetic receivers can then be used for time transfer to TAI. In that case, we propose some modifications to the CCTF procedure to adapt it for the links between geodetic receivers, in order to take advantage of the P codes available on L1 and L2. This new procedure forms the ionosphere-free combination of the P1 and P2 codes as given by the 30 s RINEX observation files, the standard of the International GPS Service. The procedure is tested using the Ashtech Z-XII3T geodetic receivers and the results are compared with those obtained with the classical CCTF procedure based on the C/A code by computing the fractional frequency stability (Allan deviation) of the time links. Over short baselines, the two techniques are equivalent, while the new technique provides a factor 2 improvement for a transatlantic time link. For time links between a time receiver and a geodetic receiver, the differential satellite delays (P1-C/A or P2-C/A) must additionally be introduced. We show here that these biases do not, however, alter the long-term (>3 days) stability of the time transfer results. The corrections associated with tidal station displacement are also investigated, and the results indicate that they do not significantly improve the results at the present level of precision.