This paper presents a novel methodology to optimize high-performance computing hardware selection and allocation for Computational Fluid Dynamics (CFD) simulations. CFD solver performance does not scale perfectly with increasing CPU core count, and scaling behavior varies for different types of hardware and different categories of models. As such, scaling cannot be predicted a priori by looking at hardware specifications or benchmarks; instead, testing is necessary for each model and each hardware type. Our methodology guides this process, simplifying comparison of many alternatives by normalizing metrics of speed and cost. These quantities produce a pareto front allowing easy and intuitive visualization of alternatives. Furthermore, we demonstrate universal scaling behavior across three hardware types and three models with mesh sizes from 3E5 to 5E7 cells.