Benchmarking the performance of algorithms, devices, and programming methodologies is a well-worn topic among developers and research of high-performance computing appliances. Independent and partisan benchmark results are a difficult thing to navigate and understand. Benchmarking has evolved from looking just at clock speed, to encompassing operations per second, and now including total cost of ownership and energy efficiency. Because each methodology is based on a set of assumptions, designers and developers are forced to critique each with a focus on the first two stages of scientific inquiry. Understanding and using reliable device benchmarks for system decisions is essential for scientific and military applications. While individual benchmark results may be insufficient for making device decisions on a standalone basis, it is important to understand the trends and factors that make up the results of benchmarking studies.