In this paper, we determine the maximum achievable "performance" of a wireless CDMA network that employs a conventional matched filter receiver and that operates under optimal link-layer adaptation where each user individually achieves the Shannon capacity. The derived bounds serve as benchmarks against which adaptive CDMA systems can be compared. We focus on two optimization criteria: minimizing the maximum service time and maximizing the sum of users rates (i.e., network throughput). We show that the problem of joint optimization of the transmission powers and rates so as to minimize the maximum service time can be formulated as a generalized geometric program (GGP), which can be transformed into a nonlinear convex problem and solved optimally and efficiently. When the goal is to maximize the sum of the rates, we show that the problem can be approximated as a GGP. Our derivation methodologies are applicable to both ad hoc and cellular networks. Numerical results are provided to show how well variable-rate, variable-power adaptation schemes perform relative to the performance bounds derived in this paper.