Sharing on-chip network resources efficiently is important in designing a cost-efficient network on-chip (NoC). Concentration is a technique that provides sharing and has been used in many proposed on-chip networks. However, the trade-off in concentration implementation and performance has not been well understood. In this paper, we describe cost-efficient implementations of concentration and show how the use of external concentration in NoC provides a significant reduction in complexity (47% and 36% reduction in area and energy, respectively) compared to previous assumed integrated (high-radix) concentration while degrading overall performance by 7%. To overcome the performance limitation, hybrid implementations of concentrators are proposed to provide a compromise between complexity and performance. However, since concentration results in sharing channels, the resulting architecture results in poor utilization of the channels and achieves lower throughput compared to a conventional 2D mesh topology. We further describe how channel slicing can be used together with concentration to implement virtual concentration, which provides throughput that matches that of 2D mesh topology while providing low-design complexity – reducing area and energy by 66% and 33% respectively.