Climate models show strong links between Antarctic and global temperature both in future and in glacial climate simulations. Past Antarctic temperatures can be estimated from measurements of water stable isotopes along the EPICA Dome C ice core over the past 800 000 years. Here we focus on the reliability of the relative intensities of glacial and interglacial periods derived from the stable isotope profile. The consistency between stable isotope-derived temperature and other environmental and climatic proxies measured along the EDC ice core is analysed at the orbital scale and compared with estimates of global ice volume. MIS 2, 12 and 16 appear as the strongest glacial maxima, while MIS 5.5 and 11 appear as the warmest interglacial maxima. The links between EDC temperature, global temperature, local and global radiative forcings are analysed. We show: (i) a strong but changing link between EDC temperature and greenhouse gas global radiative forcing in the first and second part of the record; (ii) a large residual signature of obliquity in EDC temperature with a 5 ky lag; (iii) the exceptional character of temperature variations within interglacial periods. Focusing on MIS 5.5, the warmest interglacial of EDC record, we show that orbitally forced coupled climate models only simulate a precession-induced shift of the Antarctic seasonal cycle of temperature. While they do capture annually persistent Greenland warmth, models fail to capture the warming indicated by Antarctic ice core dD. We suggest that the model-data mismatch may result from the lack of feedbacks between ice sheets and climate including both local Antarctic effects due to changes in ice sheet topography and global effects due to meltwater–thermohaline circulation interplays. An MIS 5.5 sensitivity study conducted with interactive Greenland melt indeed induces a slight Antarctic warming. We suggest that interglacial EDC optima are caused by transient heat transport redistribution comparable with glacial north–south seesaw abrupt climatic changes.