RT-PCR testing is crucial in the diagnostic of SARS-CoV-2 infection. The use of reliable and comparable PCR assays is a cornerstone to allow use of different PCR assays depending on the local equipment. In this work, we provide a comparison of the Cobas® (Roche) and the RealStar® assay (Altona). Assessment of the two assays was performed prospectively in three reference Parisians hospitals, using 170 clinical samples. They were tested with the Cobas® assay, selected to obtain a distribution of cycle threshold (Ct) as large as possible, and tested with the RealStar assay with three largely available extraction platforms: QIAsymphony (Qiagen), MagNAPure (Roche) and NucliSENS-easyMag (BioMérieux). Overall, the agreement (positive for at least one gene) was 76 %. This rate differed considerably depending on the Cobas Ct values for gene E: below 35 (n = 91), the concordance was 99 %. Regarding the positive Ct values, linear regression analysis showed a coefficient of determination (R2) of 0.88 and the Deming regression line revealed a strong correlation with a slope of 1.023 and an intercept of -3.9. Bland-Altman analysis showed that the mean difference (Cobas® minus RealStar®) was + 3.3 Ct, with a SD of + 2.3 Ct. In this comparison, both RealStar® and Cobas® assays provided comparable qualitative results and a high correlation when both tests were positive. Discrepancies exist after 35 Ct and varied depending on the extraction system used for the RealStar® assay, probably due to a low viral load close to the detection limit of both assays.