Chips of black spruce (Picea mariana (Mill.) B.S.P.) logs coming from two provenances, Montmorency (at 47°N) and Chibougamau (at 50°N) sites, were produced by a chipper-canter using three cutting widths (12.7, 19.1, and 25.4 mm). Chip dimensions were assessed by thickness, width, and length (Domtar and Williams classifications). Earlywood density, latewood density, ring density, ring width, rings per mm (R/mm), shear, splitting, modulus of elasticity (MOE), modulus of rupture (MOR) in bending, and basic density (BD) were evaluated on samples obtained within each cutting width area. The results showed that these wood attributes were significantly different between sites or/and cutting widths (12.7 and 25.4 mm). The weighted mean chip thickness (WCT) and chip class distributions (Domtar and Williams) were significantly affected by site and cutting width (CW). WCT increased as cutting width increased and its value was higher at Montmorency than at Chibougamau. The differences in WCT could be associated with the variation of growth ring characteristics, mechanical properties, and BD between sites. Moreover, chip thickness (Domtar) was significantly affected by MOR and BD while chip width and length (Williams) were significantly influenced by R/mm, BD, MOR, and shear. The analyses showed the importance of growth ring characteristics, mechanical properties, and basic density on the chip size variation. Multiple linear regressions showed that CW, MOR, BD, and R/mm were the variables that better described the variation in WCT. Chip dimensions can hence be adjusted as a function of CW but they could vary depending on the provenance of the wood and their specific attributes.