This article describes an extensive evaluation of introducing a game project to a software architecture course. In this project, university students have to construct and design a type of software architecture, evaluate the architecture, implement an application based on the architecture, and test this implementation. In previous years, the domain of the software architecture project has been a robot controller for navigating a maze. In 2008, the students on the software architecture course chose between the two domains: Khepera robot simulation in Java and XNA game development in C#. Independent of the domain chosen, the students had to go through the same phases, produce the same documents based on the same templates, and follow exactly the same process. This article describes an evaluation where we wanted to investigate if a game development project could successfully be used to teach software architecture. Specifically in the evaluation, the effect of the choice of COTS (Commercial Off-The-Shelf) and domain is compared in relation to popularity of the project type, how the students perceive the project, the complexity of the software architectures produced, the effort put into the project, and the grades achieved for the project and the written examination. The main conclusion is that game development projects can successfully be used to teach software architecture. Further, the results of the evaluation show among other things that students who chose the Game project produced software architecture with higher complexity, and put more effort into the project than the Robot project students. No significant statistical differences were found in final grades awarded to the Game project students vs. Robot project students. However, the Game project students obtained a higher grade in their project than in the written examination, whereas the Robot project students scored higher in the written examination than in their project. Finally compared to the Robot project students, those that chose the Game project had fewer problems with COTS hindering the architecture design and introducing technical challenges. (Contains 8 figures and 10 tables.)