Density variation, being the timber longitudinally measured, and the precision of spectroscopy of near infrared (NIR) were evaluated to predict density and pulp yield in two Eucalyptus globulus clones. Furthermore, variability in volume and in NIR predictions was evaluated in four E. globulus clones represented in two growth sites. Clones were selected by means of two trials: 10 and 11 year old clones in two sites from Araucanía Region, Chile. From each selected ramet, shavings samples at breast height diameter (BHD) were obtained to NIR predictions. Additionally, of a subgroup of ramets, chips samples from all the timber and six disks BHD, 20 %, 40 %, 60 %, 80 % and 100 % of commercial height were obtained. NIR predictions were contrasted with values obtained for chips and disks by means of standard error of prediction (SEP) and bias. Variance analyses of two factors (clones and sites) and correlation analyses were used. Density showed an increase from the base to the apex in both clones, though with a different variation profile; consequently, clones presented a different mean density, nonetheless their BHD was similar. NIR predictions showed an error inside the rank given for E. globulus, being those from pulp yield especially exact. Concerning volume, density and pulp yield showed low variability, both among clones and response to the site as well.