<p>The generalized risk parity portfolio allows investors to construct a well-diversified portfolio with an optimal risk-return profile in a pre-set risk dispersion range. It is a special nonconvex quadratically constrained quadratic programming problem. This paper proposes an effective branch and bound algorithm to solve the problem. First, we derive a semidefinite programming (SDP) relaxation with eigenvector based Reformulation-Linearization Technique (RLT) constraints for the problem. Subsequently, we pay more attention to tightening the upper bound. At each iteration, a convex log-barrier model is solved to identify a feasible solution within the same or complementary orthant as the optimal solution of the SDP relaxation. Once the objective function value of the new feasible solution is less than or equal to the current upper bound, we start the successive convex optimization (SCO) algorithm to refine the upper bound. Numerical experiments demonstrate that the proposed algorithm outperforms the Alternating Direction Method of Multipliers (ADMM) algorithm and the state-of-the-art global solver Gurobi in obtaining global optimal solutions.</p>