This paper investigates the capacity of multiple-antenna broadcast channels with partial channel state information (CSI) at the transmitter. The authors propose a scheme that constructs \( M \) random beams and transmits to users with the highest signal-to-noise-plus-interference ratios (SINRs), which can be determined with minimal feedback from the users. The throughput of this scheme is analyzed asymptotically for fixed \( M \) and increasing \( n \) (number of users), and it is shown that the throughput scales as \( M \log \log n \), which is the same scaling achieved with perfect CSI using dirty paper coding. The paper also explores the scaling of throughput with \( M \) when \( M \) grows to infinity, finding that throughput scales linearly with \( M \) if \( M \) grows no faster than \( \log n \). Additionally, the paper examines the fairness of the scheduling scheme in a heterogeneous network, showing that the probability of transmitting to any user converges to \( \frac{1}{2} \) when \( M \) is large enough, ensuring fairness. The choice of \( M = \alpha \log n \) transmit antennas is recommended for optimal performance in terms of throughput and fairness.This paper investigates the capacity of multiple-antenna broadcast channels with partial channel state information (CSI) at the transmitter. The authors propose a scheme that constructs \( M \) random beams and transmits to users with the highest signal-to-noise-plus-interference ratios (SINRs), which can be determined with minimal feedback from the users. The throughput of this scheme is analyzed asymptotically for fixed \( M \) and increasing \( n \) (number of users), and it is shown that the throughput scales as \( M \log \log n \), which is the same scaling achieved with perfect CSI using dirty paper coding. The paper also explores the scaling of throughput with \( M \) when \( M \) grows to infinity, finding that throughput scales linearly with \( M \) if \( M \) grows no faster than \( \log n \). Additionally, the paper examines the fairness of the scheduling scheme in a heterogeneous network, showing that the probability of transmitting to any user converges to \( \frac{1}{2} \) when \( M \) is large enough, ensuring fairness. The choice of \( M = \alpha \log n \) transmit antennas is recommended for optimal performance in terms of throughput and fairness.