The improvement of our quantitative understanding of the processes steering the atmospheric CO2 concentrations in the past is a long-term effort of many disciplines. Ice cores provide a direct atmospheric archive of both the CO2 concentration and its carbon isotopic composition (d13C). Precise and accurate measurements of d13C on ice cores can help attribute atmospheric CO2 changes to fluxes from and to the reservoirs of the global carbon cycle. Ice core d13C measurements performed over the last 20 years already contributed to constrain and quantify the possible processes underlying the glacial/interglacial and the Holocene CO2 dynamics. Methodological hurdles during the gas extraction of the trapped gases and the mass spectrometric measurement have so far limited d13Cs interpretative power in global carbon cycle models. We developed a new sublimation-GC-IRMS system, which combines high-precision d13C analysis (0.06 permil) with a low sample demand (~6 g ice per measurement) to enable high-resolution 13C time series. The applied sublimation extraction technique allows us to quantitatively extract the trapped air in both bubble and clathrate ice. Using a gas-chromatographic separation prior to the measurement excludes troubles from isobaric interferences. To improve our understanding on the Holocene CO2 fluctuations with regard to the Ruddiman hypothesis, new d13C data will be presented and discussed.