PTB physicist estimate the relative uncertainty with which the SI second is realized with CSF1 and CSF2 as less than 4 x 10-16 (CSF1) and 2 x 10-16 (CSF2), respectively. In the course of a year, an offset from an "ideal" clock could accumulate to about 13 ( = 4 x 10-16 × 365.25 × 86400 s) or 6 billionth of a second, respectively. Compare this to a quartz wristwatch that usually has a monthly inaccuracy of a few seconds. Mechanical watches - no matter how expensive they may be! - are typically wrong by a multiple of this. In local experiments the agreement among CSF1 and CSF2 was proven many times. It has also been proved that the seconds of fountains at the French metrology institute SYRTE (at Paris Observatory) and PTB do not differ by more than a few 10-16 s.