Chinese loess and stalagmites are two classic terrestrial archives that have been intensively studied to reconstruct the characteristics and dynamics of orbital- and millennial-scale monsoon variability. However, comparability between these two representative records remains contested due to chronological uncertainty and proxy complexity. Here five high-sedimentation-rate loess records on the Chinese Loess Plateau are investigated to assess East Asian monsoon variability at orbital and millennial timescales. By correlating loess grain size (a winter monsoon proxy) to speleothem δ¹⁸O (a summer monsoon proxy), we establish an independent speleothem-based chronology for Chinese loess-paleosol sequences over the past 640 ka. The synchronized loess grain-size records indicate that winter monsoon variability is spatially consistent on glacial-interglacial to millennial timescales, exhibiting distinctive precession- and millennial-scale changes similar to those of speleothem δ¹⁸O records. However, as affected by the complex processes of deposition and weathering, magnetic susceptibility variations are spatially different in terms of both amplitude and frequencies. Cross-spectral results of a loess grain-size stack with benthic oxygen isotope and orbital parameters reveal that ice volume has played a more dominant (70–75%) role than insolation (25–30%) in driving orbital-scale winter monsoon fluctuation. Comparison of high-frequency (<9 ka) components of the loess and speleothem proxies with the North Atlantic and ice-core records reveals that millennial-scale abrupt events are persistent and comparable during glacial periods. However, the magnitudes of abrupt climate events are quite different among these records during interglacials, implying that dynamics of abrupt climate changes are likely dissimilar under different glacial and interglacial boundary conditions. Integration of terrestrial, marine, and ice-core records suggests that future research should focus on quantitatively reconstructing millennial-scale changes in temperature and precipitation independently, and dynamically assessing co-evolution of orbital and millennial climate variability.