Add to ORKGIn this paper, a methodology to compare highthroughput turbo decoder architectures, is proposed. The model is based on the area-efficiency estimation of different architectures and design choices. Moreover, it is specifically oriented to the exploration of Parallel-MAP (PMAP) architectures, combined with both the Max-Log-MAP algorithm and the recently proposed Local-SOVA. The main objective is the search for optimal radix-orders, capable to maximize the area-efficiency of the decoder. In this scenario, it is proved that i) radix-orders higher than 4 are expected to drastically reduce the area-efficiency; ii) the optimal choice between radix-2 and radix-4 architectures strongly depends on the area distribution between logic and m...