We propose novel reconfigurable-intelligent-surface (RIS)-based energy harvesting (EH) systems with linear EH (L-EH) and non-linear EH (NL-EH) models where channels are exposed to Nakagami- m fading. We also compare two RIS-based EH systems in terms of RIS locations, namely RIS-EH and RIS-information processing (RIS-IP). In the RIS-EH, RIS increases the amount of harvested energy, while in the RIS-IP, RIS is utilized to improve the information processing link. Closed-formed expressions of throughput, outage probability, and average harvested power are derived and confirmed via simulation. The L-EH model overestimates system performance, while the NL-EH model provides a realistic interpretation of EH system design.