Abstract In this paper, a single observation Kalman filter is proposed to speed up the gain matrix computation. The OpenMP technique is introduced to accelerate the covariance matrix update. Lastly, 55 reference stations are selected to estimate real time GPS satellite clocks from day 79 to day 89 in 2017. The estimated clocks are compared with IGS final 30 s clocks. PPP in static and kinematic solutions for 10 stations are carried out. The results show that the differences are within 0.5 ns, the precision in horizontal can be better than 1 cm in static and 1~2 cm in kinematic. The precision in vertical can reach up to 1.5~3 cm in static and range from 3~4 cm in kinematic. Furthermore, the time cost for each epoch is reduced to 4 s from 6 s when 8 slave threads are involved.