Selecting the most effective tools and languages for teaching parallel and distributed computing presents a challenge for teachers and students. Often a high-level solution seems easier, but it comes with the cost of hiding underlying concepts and problems. In this paper we present a comparison between the Pthreads and OpenMP APIs considering the experiences and performance results of undergraduate and graduate Computer Science students from the University of São Paulo. We show that students were able to achieve better performance results using Pthreads than OpenMP, despite their perception that Pthreads is harder to learn and use.