Feed intake (FI), weight gain (WG), metabolic weight (MW), feed conversion ratio (FCR), residual feed intake calculated by regression (RFI) and feeding standards formulae (RFIS) were recorded on 1481 steers and heifers of tropically adapted and temperate breeds of cattle feedlot finished on a grain based diet for the domestic (liveweight 400 kg), Korean (520 kg) or Japanese (steers only; 600 kg liveweight) markets. Also measured were subcutaneous fat at the rump (P8) and 12/13 rib sites, 12/13 rib eye muscle area and intra-muscular fat (IMF%), time spent eating, eating rate and number of meals per day. Estimated heritabilities of FI, MW, WG, FCR, RFI and RFIS were 0.27, 0.41, 0.23, 0.06, 0.18 and 0.13. RFI and RFIS had very high genetic (0.98) and phenotypic (0.94) correlations, suggesting that they represent very similar traits. RFI had relatively high genetic correlations with rump and rib fat (0.72 and 0.48 adjusted for age; 0.79 and 0.58 adjusted for carcase weight), but lower correlations with IMF% (0.22 and 0.25 adjusted for age and carcase weight, respectively). Selection for lower RFI is therefore possible in feedlot finished cattle, but fatness will also decrease. In this study, selection for reduced fatness was predicted to reduce RFI by more than direct selection. When appropriate, multivariate selection is therefore recommended to achieve increased feed efficiency together with the desired level of fatness, using an index including RFI, on-test weight gain and fat measurements. There were large breed differences for number of meals per day; Brahman cattle ate more frequently than Belmont Red and Santa Gertrudis breeds which ate more often than temperate breed cattle. Within breeds, there was a tendency for more efficient animals to have fewer meals per day.