This experiment provides results in voluntary contribution for a public goods using a payoff function which defines a Pareto optimum inside the strategic contributions space. Two optima are tested. One is defined at 30% of the subjects' endowment, and the other at 70%. The Nash equilibrium of these games remains the non contribution to the public good. Such a reward structure enable to focus on three main items. First, does the over-contribution highlighted in the previous experiments is under-efficient because of subjects' errors or confusion or is-it due to a voluntary decision? Second, does the partial contribution device encourages the social optimum to arise? And third, is there a coordination problem between the subjects' contributive behavior in such a framework? Indeed, all combinations of contribution reaching 30% (70%) of the total endowment of the group fully provides the public good but one of them - when each subject provides 30% (70%) of his /her own endowment - is the Pareto optimum point. The results show that the over but inefficient contribution previously presented is not due to a confusion made by the subjects but is rather due to the subjects' voluntary decision. Next, a public good is easily provided when the optimal point is defined at a low contribution level, even if it is not always fully provided. This last result highlights a coordination problem to reach the social optimum, whatever the Pareto optimum.