There exists a significant pathway in mitigating climate change and achieving carbon neutrality, which is to enhance the efficiency of forestry carbon sink. An objective evaluation of China's forestry carbon sink efficiency and its spatio-temporal evolutionary characteristics plays a crucial role in speeding up the process of national land greening. The three-stage super-efficiency SBM-DEA model was used to estimate forestry carbon sink efficiency in 30 provinces (municipalities and districts) in China from 2008 to 2021. At the same time, in order to find out the change rules and spatio-temporal evolutionary characteristics of forestry carbon sink efficiency, we used the methods of GML index, global Moran index and spatial Markov chain to carry out detailed analysis. The results revealed that: (1) In general, the level of forestry carbon sink efficiency in China was not high, which had great room for improvement. The difference of forestry carbon sink efficiency among different regions was also very obvious. The spatial pattern of forestry carbon sink efficiency in four forest regions was Southwest Forest region>Northeast forest region>Southern forest region>Northern forest region. After eliminating the influence of external environment and random error, the low efficiency regions showed a trend of catching up with the high efficiency regions. (2) The change of forestry carbon sink efficiency showed an upward trend, and the region with the highest GML index was the northeast forest region and the lowest was the northern forest region. Technological progress had a great contribution to the improvement of forestry carbon sink efficiency, and different regions had different dependence on comprehensive efficiency. (3) The efficiency of forestry carbon sink was spatially non-equilibrium, with a "bounding" distribution between high-efficiency regions and low-efficiency regions, and this feature remains stable in time. After 2015, there was a significant spatial positive correlation between forestry carbon sink efficiency, and the spatial distribution had a clustering effect. The results of the spatial Markov chain considering the spatial lag term indicated that the neighborhood type had a significant impact on the state transition of forestry carbon sink efficiency, and there was a strong neighbor effect in the spatial distribution, showing the spatial evolution characteristics of "high and high convergence, low and low convergence, high drive low, low inhibition high". The research is of great practical significance for enhancing the carbon sink capacity of forest ecosystem and constructing the modernization of harmonious coexistence between man and nature.