We modeled the technical relationships between volume of Pinus radiata D. Don structural lumber (with stiffness>8 GPa) and log attributes using a stochastic frontier approach. The production models were Cobb–Douglas and Translog, while the log attributes were small end diameter (SED), wood stiffness (STF), and largest branch (LBR); however, the effect of the latter trait was not significant (p>0.05). Economic values of log traits were represented by their values of marginal product (VMP). The coefficients for the Cobb–Douglas frontier were statistically significant and the model met most of the production theoretical properties. VMP derived from the Cobb–Douglas were 2.23 NZ$/cm for SED and 16.88 NZ$/GPa for STF. The Translog frontier coefficients were also significant (p<0.05) and VMP derived from this model were 1.67 NZ$/cm for SED and 9.15 NZ$/GPa for STF. Thus, for the analyzed production stage, changes to SED and STF were relevant for improving log value recovery above MSG8+. Technical efficiency derived from the frontiers allowed to identify and characterize the best logs to produce structural grades with stiffness of 8 GPa or higher.