With what accuracy do PTB's atomic clocks work?

For the comparison of the atomic clocks constructed at PTB, the rate errors are 1 to 3 billionths of a second per day (for comparison: a quartz wrist watch has a monthly inaccuracy of a few seconds, mechanical watches - no matter how expensive they may be! – are wrong by a multiple of this). In physical terms, for CS2 from which the legal time is derived, the uncertainty has been estimated at 12x10^–15 (This means that over the course of a year, an error of one millionth of a second must be reckoned with relative to an ideal clock.) With the cesium fountain, an even smaller uncertainty, namely 1x10^–15, is achieved. It has also been proved that the seconds of two fountains (of NIST/U.S.A. and of PTB) differ by no more than 1x10^–15.