Blind equalization system model is established with atmospheric turbulence time-varying channel. Under different scintillation distribution conditions, performances of two kinds of adaptive blind equalization algorithms are compared such as convergence, mean square error (MSE) and stability, and clustering in constellation of wireless optical subcarrier signal are analyzed. With the increasing of turbulence intensity, variable step-size constant modulus algorithm (CMA) algorithm based MSE has faster convergence, smaller MSE but poor stability, and scale factor must be decreased to ensure convergence compared with CMA algorithm. Meanwhile, compared with Gaussian channel, under the same signal-to-noise ratio, iterative step length factor and scale factor under the turbulence channel is smaller, but mean square error is bigger than in Gaussian channel. The simulation results show that two kinds of algorithms can effectively improve the convergence of constellation diagram, and have important significance in improving detection rate of constellation diagram.