Abstract: Accurate estimation of uncalibrated phase delay (UPD) is an important precondition for ambiguity resolution of precise point positioning (PPP). The commonly used PPP models are mainly divided into the ionosphere-free (IF) combination model and the undifferenced and uncombined model, and the UPD estimation methods used for the two models are different. In this paper, the consistency of UPD estimation based on IF combination model and uncombined model is proved theoretically under the premise of using the same processing strategy. Furthermore, the UPD products of BDS-3 were estimated by using the observation data of 45 International GNSS Service (IGS) stations uniformly distributed around the world, and the time stability and consistency of UPD products obtained by the two estimation models were analyzed. The results show that the UPD of wide-lane and narrow-lane in BDS-3 maintain high stability. The average standard deviation of UPD of wide-lane in ten days is 0.0607, and that of narrow-lane in one day is 0.059 9. For BDS-2 satellites, the UPD results estimated by the two models are highly consistent. However, the UPD results of BDS-3 satellites are related to the characteristics of satellite orbit types and manufacturers. There is a significant difference in UPD between different orbit types and manufacturers of about 0.5 weeks, but there is consistency among BDS satellites of the same orbit type and manufacturer. It can be concluded that the hardware delay of the receiver corresponding to satellites produced by different orbital types and different manufacturers of BDS-3 has a systematic deviation. The estimation of inter-satellite-type bias in PPP further demonstrates this conclusion. Finally, the impact of these biased on PPP and PPP AR was carefully discussed.