The problem of synthesis of optimal controllers for linear stochastic systems with independent control channels is considered. In the considered systems two independent controllers with independent measurement devices are used. The equations for optimal cost function, covariance matrix of the state estimation error and for the controlled system output are derived. It is shown that the application of a well-known separation principle or certainty equivalence principle results in controllers which are not optimal and which can be far from the optimality. It is also shown that improvement of the estimation accuracy by means of manipulation of the controller parameters can significantly improve the control performance as a whole. The numerical examples of calculation of the optimal independent controllers are given to demonstrate the obtained theoretical results and superiorities over controllers based on the separation principle. It is also demonstrated that the optimal controllers have smaller control gains for the control loops with larger observation noises, which can be characterized as cautious properties of the controllers.