Considering a dense small-cell network with simultaneous wireless information and power transfer (SWIPT), this work jointly designs transmit beamformers at the base stations (BSs) and receive power splitting ratios at the users (UEs). Our objectives is to maximize the minimum UE signal-to-interference-plus-noise-ratio (SINR) under BS transmit power and UE minimum harvested energy constraints. This problem is highly nonconvex, for which semidefinite programming (SDP) relaxation may even fail to locate a feasible solution. We propose an efficient spectral optimization method by expressing the rank-one constraints as a single reverse convex nonsmooth constraint and incorporating it in the optimization objective. The proposed algorithm practically achieves the theoretical bound given by SDP relaxation with almost similar complexity.