The Shockley and Queisser limit, a well-known efficiency limit for a solar cell, is based on unrealistic physical assumptions and its maximum limit is seriously overestimated.To understand the power loss mechanisms of record-efficiency cells, a more rigorous approach is necessary. Here, we have established a formalism that can accurately predict absolute performance limits of solar cells in conventional thin film form. In particular, a formulation for a strict evaluation of the saturation current in a nonblackbody solar cell has been developed by taking incident angle, light polarization and texture effects into account. Based on the established method, we have estimated the maximum efficiencies of 13 well-studied solar cell materials [GaAs, InP, CdTe, a-HC(NH 2 ) 2 PbI 3 ] in a 1-µm-thick physical limit. Our calculation shows that over 30% efficiencies can be achieved for absorber layers with sharp absorption edges (GaAs, InP, CdTe, CuInGaSe 2 , Cu 2 ZnGeSe 4 ). Nevertheless, many record-efficiency polycrystalline solar cells, including hybrid perovskites, are limited by open-circuit voltage and fill-factor losses. We show that the maximum conversion efficiencies described here present alternative limits that can predict the power generation of real-world solar cells. *fujiwara@gifu-u.ac.jp sc J kT qV J J −