Different execution models for Prolog have different cost/performance benefits and different performance opportunities. In this paper, we treat two implementations of Prolog, a tailored special purpose WAM processor, the Berkeley VLSI-PLM, and an off-the-shelf general purpose part, the MC68020. This work continues along the same vein as the work reported by Mulder and Tick. Fourteen Prolog benchmarks are compiled to WAM code, and their execution times on the VLSI-PLM and the MC68020 are compared and analyzed. Additional experiments are performed to calibrate the calculated MC68020 execution times with actual execution times of a SUN3/260 and a NCR Tower/32. The paper concludes with a number of observations.