Tree-ring carbon (C) isotope composition (δ¹³C) varies depending on photosynthetic capacity and stomatal conductance, and it is therefore intrinsically associated with vegetation productivity. Still, very little is known about the relationship between tree-ring δ¹³C values and forest gross primary productivity (GPP). Here, we investigated relationships between tree-ring δ¹³C, tree-ring width index (RWI) and ecosystem-level GPP in a Korean pine (Pinus koraiensis)-dominated old-growth temperate forest in northeastern China. Specifically, we generated chronologies of RWI and δ¹³C from early-wood (δ¹³C_EW) and late-wood (δ¹³C_LW) of Korean pines, and we built a 17-year continuous ecosystem-level GPP series using eddy covariance measurements, in which a GPP model was used for gap filling. RWI showed vague relationships with tree-ring δ¹³C and whole-forest GPP, and it was insensitive to interannual variations in climate at the study site. By contrast, both δ¹³C_EW and δ¹³C_LW showed significant relationships with climate variables and GPP. This suggests that isotopic signals of photosynthetic C uptake were imprinted on the formation of tree rings and that tree-ring δ¹³C of Korean pine is a better proxy for forest GPP than RWI. In addition, tree-ring δ¹³C of Korean pines was positively correlated with GPP, with an especially close relationship between δ¹³C_EW and GPP of the current year. This implies that photosynthetic activity, rather than stomatal conductance, was the main driver of tree-ring δ¹³C signals of Korean pine and it was mainly influenced by the climate conditions of the current year, probably because of a limited use of previously fixed C during early-wood formation. Our findings demonstrate the great potential of tree-ring δ¹³C for reconstructing GPP timeseries at the centennial scale for forest ecosystems, and they could help to constrain parameters in terrestrial C cycle models to improve predictions of C fluxes.