EEMBC Standardizes CoreMark Energy Measurement with New Benchmark
EEMBC announced the release of the ULPMark-CoreMark benchmark—the third in the ULPMark series—designed to standardize the measurement and reporting of MCU energy. Pictured above, ULPMark-CM (right) adds more configurability to the analysis of CoreMark (left). Instead of a single instance of the code at one condition, ULPMark-CM provides an API to analyze several operating points.
Under constant demand from customers for lower MCU active power, manufacturers have been using EEMBC's CoreMark benchmark to differentiate their low-power products from their competitors. Until today there were no standard methods or rules governing the consistency of these measurements. This has led to a mixed bag of incomparable results as companies develop their own differing methodologies.
EEMBC's Ultra-Low Power (ULP) Working Group has fixed this problem by developing an official standard for reporting CoreMark energy called "ULPMark-CoreMark", or, ULPMark-CM for short. Building on the success of the ULPMark family of software, the new benchmark uses the same energy monitor and host software framework as ULPMark-CoreProfile and ULPMark-PeripheralProfile.
This new addition rounds out the capabilities of the product line by adding an active-power profile, based on the classic CoreMark benchmark, with a clearly-defined methodology.
The official metric has the same name as the benchmark: the ULPMark-CM, which is proportional to iterations per milli-Joules. This is a crucial metric for understanding how much work an MCU can do for a given amount of energy.
"When comparing datasheets from different companies, the diversity of metrics relating CoreMark performance to energy efficiency complicates cross-product comparisons," said EEMBC President, Peter Torelli. "By creating a single metric and clearly defining the methodology, comparisons become much more meaningful."
The benchmark provides more than just a single score, it generates a set of scores for different configurations that reinforce the relationship between performance and energy. For example, running at a higher frequency improves the CoreMark score (performance) but uses more energy, while running at a lower voltage improves the ULPMark-CM score (energy efficiency) at the cost of performance. By presenting both scores side-by-side, the new benchmark clearly illustrates that tradeoff.