The recent troubles surrounding emissions testing in the automotive industry highlight a simple fact that every electronics designer knows – you can’t always trust the datasheet. Companies – no matter how ethically run – have an interest in presenting their products in the most favorable possible light. Organizations may not go so far as to “fix” test results, but wise engineers will always check the data for themselves.
That’s why we’ve submitted the Apollo ultra-low-power microcontroller for independent benchmark testing by the Embedded Microprocessor Benchmark Consortium (EEMBC). EEMBC is an industry alliance, with more than 50 members ranging from Altera through Marvell to Wind River, whose stated mission is to “develop benchmarks to help system designers select the optimal processors and understand the performance and energy characteristics of their systems”.
EEMBC was originally established to address the inadequacy of simple Dhrystone benchmarking, with the intention of reflecting the requirements of real-world applications. The certification rules are rigorous – the EEMBC Technology Center must execute benchmarks run by the manufacturer, ensuring that scores are repeatable and fairly obtained.
Since its inception in 1997, EEMBC has grown its suite of benchmarks to cover a range of real-world applications: cloud computing and big data; mobile devices; networking; automotive; the Internet of Things (IoT); and, of course, ultra-low power microcontrollers.
The great news for us at Ambiq is that the Apollo microcontroller achieved double the score of the previous best device when measured against EEMBC’s ULPBENCH ultra-low-power benchmarks. That’s a great proof point for our Subthreshold Power Optimized Technology (SPOT), which uses transistors biased in the subthreshold region of operation (ie ultra-low voltage operation) to achieve unmatched energy efficiency.
Aiming to provide “a methodology to reliably and equitably measure MCU energy efficiency”, ULPBENCH takes a “two-pronged” approach, with a suite of tests that comprehensively measure efficiency, and the EEMBC EnergyMonitor™ GUI-based tool for measuring energy. The goal is to benchmark all of the key features of an ultra-low-power MCU: the low active current required when power resources are limited, for devices that rely on technologies such as energy harvesting; the low sleep current that is needed for applications like metering, in which activity is intermittent and sleep-mode energy consumption predominates; and active energy (rather than power) efficiency, measuring how much useful work can be done in a limited time period.
In addition, the benchmarks endeavor to account for the impact of well-designed peripherals and acceleration functions.
The Apollo MCU’s unprecedented benchmark figure of 377 indicates outstanding performance across the board. As well as achieving low dynamic power as low as 34µA/MHz via the use of the SPOT architecture, the device is based on an ARM® Cortex®-M4 processor, allowing it to deliver more performance than competing solutions. As a result it requires less clock cycles to execute a given task than lower-performing devices. An ultra-low sleep mode current of 350nA (with real time clock enabled) serves the needs of sensing, IoT and other applications in which the device spends the majority of its time in sleep mode.
Finally, Apollo includes an ultra-low power interface for off-chip sensors, enhancing the all-important peripheral performance.
Of course, those are the “datasheet facts” – the specs that Ambiq itself presents to the world. And as we have observed, it’s not always easy for the designer to compare such figures between manufacturers. But the device’s ULPBENCH score provides independent verification that these specs are meaningful, and make a tangible difference in real world applications.
Benchmarks like EEMBC ULPBENCH are a valuable addition to the designer’s armoury in making informed decisions about the range of ultra-low-power MCUs on the market. We’re delighted to have become a member of EEMBC , and encourage engineers to make full use of the valuable resource that the benchmark results represent.