The effects of breed and feeding on the prediction of carcass fat and carcass muscle using rump ‘P8’ subcutaneous fat thickness were investigated in heavy-weight beef carcasses (277 to 512 kg). Hot side weight, eye muscle area and their combination added to fat thickness were evaluated as additional predictors. The predictions of carcass composition from grass-fed steers were validated on data from grain-fed steer carcasses. There were differences between breeds and feeding regimes in the level of tissue proportions at a given fat thickness, but no differences in regression slopes. Fat thickness alone was, therefore, not an accurate indicator of the proportions of side fat or side muscle in heavy-weight carcasses. In the grain-fed steers, the correlations between P8 and tissue proportions were relatively strong, and the addition of hot side weight and eye muscle area, alone or in combination, did not significantly improve the predictions. Conversely, in grass-fed steers where the relationships between carcass composition and fat thickness were weaker, the addition of hot side weight and eye muscle area made a significant contribution to prediction. In the predictions of tissue weights, hot side weight and eye muscle area when added to fat thickness, significantly contributed to the prediction of side muscle weight for both feeding systems. These three variables did not account for the observed differences between breeds or feeding regimes in the proportion or weight of fat, nor in the proportion of muscle; but they effectively eliminated these groups' differences in the estimation of muscle weight.