This paper presents the system level performance evaluation for energy-detection based cooperative spectrum sensing in cognitive radio networks. Three performance criteria are quantitively analyzed for cooperative spectrum sensing. First, the average error probability is determined given fixed amplifier gains for a fixed number of secondary users by considering all possible channel realizations. Second, the asymptotic error probability is computed in a power constrained cognitive radio network when the number of secondary user approaches infinity. Third, the outage probability is examined when instantaneous error probability is greater than a predefined threshold. In all three calculations, both additive white Gaussian noise (AWGN) and Rayleigh fading assumptions are used to capture the observation and fusion channels. Numerical results indicate that in order to maintain a desired detection performance in low and moderate fusion signal to noise ratio (SNR) regimes, fusion channels need to be as reliable as possible, while local received SNRs can be dynamic and provide spatial diversity. Moreover, it is shown that under AWGN observation channels and Rayleigh fading fusion channels, a diversity order equal to the number of secondary users can be achieved.