MCU benchmark homes in on peripheral power

By Chris Edwards |  No Comments  |  Posted: September 12, 2017
Topics/Categories: Blog - Embedded  |  Tags: , , ,  | Organizations: , ,

The EEMBC benchmarking consortium has released the latest in its collection of test suites for measuring power efficiency in embedded microcontrollers, focusing in this case on peripheral performance.

The initial release of the ULPMark-PeripheralProfile (ULPMark-PP) is accompanied by the first batch of results from members Ambiq Micro and STMicroelectronics. The benchmark recognizes the role that smart peripherals play in optimizing power consumption for embedded systems because they allow the main processor core to be in an energy-saving mode for much of the time they are active. Tests have shown that a good core-efficiency profile does not necessarily translate into a good overall score when peripheral behavior is taken into account. The problem that faces users is determining how well a combination of peripherals will work in a target application.

Peripherals covered by the new benchmark are analog-digital conversion (ADC), SPI communication, PWM control and the real-time clock, which is often the only function to remain active on a sleeping microcontroller and so can contribute heavily to the overall energy consumption of a low duty-cycle IoT node. The benchmark compares performance of activities such as ADC reads that trigger an interrupt and those that simply deliver data to memory using DMA.

Other tests exercise the PWM by changing its speed as it runs, which can be performed in some cases entirely by hardware. Other microcontrollers demand the processor to be woken to reprogram the DMA, which can increase overall energy consumption.

“The initial ULPMark-PP results indicate that there is a huge efficiency difference between microcontrollers,” said EEMBC president Markus Levy.

“Because every application has different needs to optimize power, performance, and cost, providing valuable and relevant microcontroller benchmarks requires ‘peeling the onion’ to understand which MCU is best for each application,” noted Mark Wallis, co-chair of the EEMBC IoT working group and system architect at STMicroelectronics. “In addition to sharing our application expertise with the other outstanding EEMBC Working Group team, EEMBC has adopted ST’s energy-testing solution, currently in the final development stages, which will facilitate benchmarking and ultimately assure validated results.”

Levy added: “These results also highlight the benefits of running at 1.8-2.0 volts versus 3.0 volts. Beyond this, I encourage all embedded system developers to encourage their MCU vendors to publish the results for their devices. A comprehensive table of ULPMark-PP results significantly adds credibility and real-world comparability to the specifications in datasheets.”

The ULPMark-PP uses the same IoTConnect framework as the existing CPU-focused version of ULPMark. Both ULPMark benchmarks are supplemented by the EEMBC IoTMark-BLE that measures the energy used by the full subsystem including the MCU, the radio, and the protocol stack. EEMBC expects to add a security-oriented benchmark to the overall suite in the future. This mimics the operation of an application running TLS-based encryption.

Comments are closed.


Synopsys Cadence Design Systems Siemens EDA
View All Sponsors