We have developed a novel incompressible SPH solver that combines the advantages of previous methods in one model, namely low computational cost per physics update and large time steps. Our method includes a convergence loop which is executed in each physics update step consisting of a prediction and correction iteration. In each convergence iteration, the new particle positions and their densities are predicted and the variations from the reference density are computed. We have derived a formulation that relates the density fluctuation and the pressure, to reduce the density errors and to approach incompressibility. With this method, we gained a speed-up of more than an order of magnitude over the commonly used WCSPH method.