Our objective was to examine the repeatability of residual feed intake (RFI) across diets with different levels of dietary protein. Mid lactation Holstein cows with initial MY 42 ± 10 kg/d (n = 88, in 3 blocks) were fed either a high protein diet (18% CP; 18P) or a low protein diet (14% CP; 14P) in a crossover design with 2 28-d treatment periods. Diets were similar and both contained at least 10% RDP (DM basis). The 18P diet contained expeller soybean meal, which was replaced by soybean hulls and ground corn in the 14P diet; 18P diet provided 4 percentage units more CP, 2 units less starch and 2 units less NDF than 14P. Cows were milked 2×/d; DMI and MY were recorded daily. Milk composition was measured during 4 consecutive milkings each week and BW was measured 3× weekly. Fixed effects of experiment, parity, diet and period nested within experiment and random effect of cow were included in the model to compare intake and production performance of treatments. RFI value was calculated for each cow on each treatment based on her actual intake, milk energy output, metabolic BW, and body energy change. Cows were ranked as high (>0.5 SD), medium (±0.5 SD) or low (<-0.5 SD) RFI. Compared with the 14P diet, the 18P diet increased DMI by 1.3 kg/d (27.1 vs. 25.8 kg/d; P < 0.01), milk yield by 5.1 kg/d (43.8 vs. 38.7 kg/d; P < 0.01), and BW gain by 0.22 kg/d (0.49 vs. 0.27 kg/d; P < 0.01). Ranking of cows for RFI was relatively repeatable (r = 0.64; P < 0.01). Of all cows, 64% maintained their group ranking across treatments whereas 33% changed ranking by 1 group. Only 2% moved in the ranking from the high to the low RFI group or vice versa. Compared with the previous studies where RFI repeatability was 0.73 across starch levels and 0.44 across forage levels, we presume that nutrient digestibility and protein efficiency are the 2 main sources for RFI variation. In conclusion, RFI was relatively repeatable across 2 diets varying in protein content enough to cause marked changes in production. We suggest this supports the use of RFI as part of a genomic breeding index to enhance feed efficiency.