DescriptionRelying vast HPC investments on an informed decision making process is important when serving the ever increasing demands for computational power. Providing a quantitative metric to compare and evaluate HPC systems in procurement processes, I define a productivity model with predictive power that focuses on the number of simulation application runs and the total cost of ownership (TCO). As part of TCO, the software development costs determined by efforts to parallelize, port or tune simulation codes for novel HPC setups must be predicted. Since approaches from mainstream software engineering does not directly suite HPC needs, I set up a methodology to estimate software development effort in HPC. Proof of concepts are mostly based on real-world HPC setups at RWTH Aachen University.