The aim of this study was to apply the Cornell net carbohydrate and protein system (CNCPS) in subtropical Taiwan. This was done by means of 3 trials, viz, in situ, lactation and metabolic trials, the latter using the urinary purine derivatives (UPD) to estimate the ruminal microbial yield. Dietary treatments were formulated according to different nutrient requirement systems including, (1) a control NRC78 group on NRC (1978) (2) a NRC88 group on NRC (1988), and (3) a CNCPS group on Cornell Net carbohydrate and protein system model. Results from the lactation trial showed that DM intake (DMI) was higher (p < 0.05) in the NRC78 than the other treatment groups. The treatments did not significantly influence milk yield, but milk yield after covariance adjustment for DMI was higher in the CNCPS group (p < 0.05). The FCM, milk fat content and yield were greater in both the NRC78 and the NRC88 group over the CNCPS group (p < 0.05). The treatments did not significantly influence the DMI adjusted FCM. The solid-non-fat and milk protein contents were higher in the CNCPS group (p < 0.05) with or without DMI covariance adjustment. Lactating efficiency was higher in the CNCPS group (p < 0.05) compared to the other groups. The significantly lowest milk urea-N (MUN) with better protein utilization efficiency in the CNCPS group (p < 0.05) suggested that less N would be excreted into the environment. Cows in the CNCPS group excreted significantly more and the NRC88 group significantly less urinary purine derivatives (UPD) implying that more ruminal microbial protein was synthesized in the CNCPS over the NRC88 group. The CNCPS could become the most useful tool in predicting the trends in milk yield, microbial yield and MUN.