Current infantry training simulators are based on first person shooter gaming products and have been used for many years for individual and small unit training. There is a need for a broader application of simulation-based training systems to train multiple small teams in concert or larger unit operations. Additionally, the systems will need to accurately present the operational area with larger numbers of civilian and opposing forces. This requires a simulation-based trainer to scale from currently tens of users to hundreds of users and entities in the same virtual space at the same time. The biggest limiting factor for this activity has been the inability for the backend simulation architectures of the first person shooters to simultaneously broker the large numbers of entities needed to support the scaled simulation. The U.S. Army Research Laboratory’s Simulation and Training Technology Center (ARL STTC) and the Intel Corporation entered into a Cooperative Research and Development Agreement (CRADA) in February of 2013 to address core simulation scaling issues. The ARL/STTC and Intel Corp. performed a series of five joint scalability experiments over the summer of 2013 to test new prototype architectures that support scaled operations. These scalability experiments were open to the public and included volunteers from industry and academia. The experiments were able to show significant increases in the number of humans who could log into a coherent training simulation and interact with each other while performing a mission. This paper will present the results of one of the events, including the data collected from the distributed simulators which were located at various locations across the continental United States. We will discuss the architecture of the prototype simulator, provide performance findings, the statistical approaches used to analyze this data and provide an interpretation of findings. Finally, we discuss a model developed from the autonomous agent simulator loads and compare it to the performance of the simulators when loaded with large numbers of human users.