Forests are rich in biodiversity and high in productivity, and understanding the relationship between biodiversity and productivity is critical to the sustainable management of forest ecosystems. Although biodiversity and other factors contribute to forest productivity to varying degrees, it is unclear whether spatial scale (or plot size) influences this relationship. In this study, we analyzed the impacts of biodiversity, trait composition, structural attributes, and soil fertility on forest productivity at three scales in two temperate old-growth forests. We found that structural attributes, including basal area and stand density, were positively correlated with productivity and were consistently the strongest drivers of productivity changes in both forest types. Trait composition had a weak direct positive correlation with productivity. Soil nitrogen had positive effects on productivity, and soil moisture content had negative effects, both of which increased with increasing plot size. Different results for the effect of biodiversity on productivity were found between the two forests. In the mixed broadleaved-Korean pine forest, diversity correlated positively with productivity, and the strength of the relationship increased with increasing plot size. In contrast, in the spruce-fir valley forest, diversity had a direct positive correlation with productivity at 10 m × 10 m, no significant effect at 20 m × 20 m, and a negative correlation at 30 m × 30 m. Our results suggest that the contribution of vegetation attributes and soil fertility to forest productivity varies with scale. Despite the fact that both the niche complementarity hypothesis and the mass ratio hypothesis act simultaneously in forests, they have limited effects on productivity. Stand structural attributes play a key role in elucidating the biodiversity-productivity relationships, and they may serve as a more significant regulator of forest productivity than diversity. At the same time, the role of soil fertility in regulating productivity cannot be ignored.