A method for improving accuracy in the computation of a one-way transfer time between two networked devices. In one aspect, variability in time transfer latency that is caused by cache loading, data structure setup time, and scheduling variability in software is reduced by initiating a first sequence of loading data structures into cache and priming scheduling, and then initiating a second sequence of calibrating the timing of a subsequent synchronization message so that the completion of the first sequence occurs just in time for the reception of the synchronization message. The method is applicable for any network time synchronization protocol, including Network Time Protocol (NTP) and Precision Time Protocol (PTP).