Description: The majority of atmosphere-ocean coupled climate models reasonably simulate sea surface temperature (SST) variability of El Niño/Southern Oscillation (ENSO) as a consequence of error compensations between too-weak positive dynamic and negative thermodynamic feedbacks. The central Pacific zonal wind response to ENSO SST anomalies is key dynamic feedback and substantially improved in atmosphere-only (AMIP-type) runs forced by observed SSTs in comparison to coupled runs, but the systematic feedback bias still remains. Here, two physical relationships are examined based on the linear regression analysis to figure out what process weakens the wind feedback: the equatorial Pacific precipitation response to the Niño-3.4 SST anomalies (P-SST relation) and the central Pacific zonal wind response to the equatorial Pacific precipitation anomalies (U-P relation). The P-SST and U-P relations in the coupled historical and AMIP runs of the Coupled Model Intercomparison Project phase 6 (CMIP6) are compared with multiple observational datasets and their seasonal dependences are focused on. As in previous studies, the annual wind feedback is underestimated on average in the CMIP6 AMIP runs although enhanced relative to the corresponding coupled runs. On the one hand, the P-SST relation in the AMIP runs is comparable to the observational levels regardless of seasons, improved from the coupled runs as mean precipitation increases. On the other hand, the U-P relation is too weak in both the coupled and AMIP runs to a similar extent, which appears in boreal winter to early spring and is characterized by too-weak wind and precipitation responses in the southern off-equator.