In a MIMO layered architecture, several codewords/nare transmitted from a multiplicity of antennas. Although the/nspectral efficiency is maximized if the rates of these codewords/nare separately controlled, the feedback rate within the link/nadaptation loop is reduced if they are constrained to be identical./nThis poses a direct tradeoff between performance and/nfeedback overhead. This paper provides analytical expressions/nthat quantify the difference in spectral efficiency between both/napproaches for arbitrary numbers of antennas. Specifically, the/ncharacterization takes place in the realm of the low- and highpower/nregimes via expansions that are shown to have a wide/nrange of validity./nIn addition, the possibility of adjusting the transmit power/nof each codeword individually is considered as an alternative to/nthe separate control of their rates. Power allocation, however,/nturns out to be inferior to rate control within the context of this/nproblem.