© 2009 American Registry of Professional Animal Scientists. Carcass data from serially slaughtered steers fed different dietary energy sources during backgrounding were used to formulate equations relating carcass fat thickness (FT) and marbling (MRB) score to hot carcass weight (HCW). Relationships between fat depot accretion and HCW were similar among diets. Pooled data resulted in linear relationships between fat deposition and HCW. These equations and a second system of equations, based on published results, were used to predict final FT and MRB score from unrelated steers. Ultrasound FT and MRB score were collected at d 142 and d 226 of the trial. Carcass traits were predicted from carcass weight gain and ultrasound at d 142 or 226. Percentage differences between predicted and actual values were analyzed with slaughter group, breed, sex, treatment, and 2-way interactions in the model. Ultrasound timing influenced prediction accuracy for FT and MRB score (P < 0.05). Marbling score accuracy increased with scan proximity to slaughter, and FT was within 10% at 158 or 180 d before slaughter. Breed affected MRB score accuracy (P = 0.04) and was within 6% for calves with Angus influence, but was more than 20% for Brahman-influenced calves. Sex influenced MRB score accuracy (P > 0.01) and FT accuracy (P > 0.01). Growing program influenced FT accuracy from ultrasound at d 142 (P > 0.01), underestimating pasture-fed cattle by 58.4%, but did not affect MRB score accuracy from either session. Accuracy of FT and MRB score predictions from growth-based equations is influenced by ADG between ultrasound and end point, breed, and sex, although scans up to 120 d preslaughter may be accurate.