Freescale proves Kinetis L-Series MCU has lowest power

In an extremely competitive market fraught with parametric fuzziness, what happened this morning at the opening keynote of day two of the Freescale Technology Forum was both dramatic and unusual: Freescale performed a demo that proved its new L-Series Kinetis MCUs drew less power than the lowest-power devices it could find from Renesas, Texas Instruments and Microchip.

Microcontrollers are everywhere and suppliers delight in finding ways and means to make them simpler to use, more flexible and more power efficient, as well as lower cost, of course. Of all the parameters, however, power efficiency is most debated. Depending on the circumstances and the application, most relatively low-power devices could be made to look to be ‘the best’.

However, the circumstances are not always real-world, said Geoff Lees, VP and general manager of Freescale’s Automotive, Industrial & Multi-Market Solutions while chatting with EDN. For example, he said, some manufacturers talk of low –leakage advantages and long-term power efficiency over multiple years. But, as he pointed out, “Most applications can’t afford the battery that needs to last that long.”

Tadiran comes to my mind, partly because its lithium thionyl chloride batteries are the only guaranteed guaranteed to last 25 years, and partly because they’re right up the road from me on Long Island. OK, twenty five years is a bit extreme, but I digress.

With regard to the demo, Lees  wanted a real-world demo that would cycle every 5 seconds and really make a point of using industry accepted benchmarks and of truly finding the lowest-power device in each competitors’ portfolio.

“We had a PIC24 [from Microchip] that we were going to use, but we found out at the last minute that it wasn’t the lowest-power device: we needed a nanoWatt PIC24.” NanoWatt is Microchip’s nomenclature for its flexible clock management system that helps reduce power consumption.

Demo parameters

Reza Kazerounian, senior vice president and general manager of the Automotive, Industrial & Multi-market Solutions Group, presided over this morning’s demonstration and prefaced the proceedings by saying that the Kinetis devices, “are about to redefine energy efficiency in order to improve the life of battery products. It’s not about raw datasheets, but about doing more with less power.” (Reducing the area under the energy curve.)

The demo comprised an L-series MCU up against a nanoWatt PIC24 from Microchip, an MSP430 from TI and an RL78 device from Renesas. All are low-power devices with similar feature sets.

Each was powered via individually via capacitors, controlled by a K60-based power system, which performed as an energy monitor to measure voltage across each capacitor as well as instantaneous current.

Just for fun, the power source was a hand crank.

The software part of the demo was straightforward. The application was a low-power CoreMark application so it tested the processor core.  Once capacitors were fully charged the demo will start automatically, for 40 coremark iterations. Once each iteration was done, the processors jumped into deep sleep to conserve energy, but only after acknowledging  that had completed that iteration. If there was no ACK in 5 second, it proved the power was gone.

The trigger was pulled, or so to speak, and sure enough the cores dropped like flies. First the Renesas RL78, then the MSP430, leaving it down to the PIC24 vs the L-Series Kinetis. Needless to say, the L-Series won out. But, it was a pretty fair test, using a straightforward procedure and a standard CoreMark benchmark algorithm.

All the while, Freescale was flagging stats such as 1.73 CoreMarks /MHz, low active current, fast processing, deep sleep modes,  12.40 coremarks /mA, “That’s 1.5 times our nearest competition!”

Postscript: More testing to come

Of course, there’s much more that goes around the processor core itself that needs to be considered. In a follow-up chat and email conversation, Lees was particularly excited about the potential for its integrated 90nm TFS (thin film storage) flash, using nanocrystalline domain as the storage medium.

The continued benchmarking of devices is of particular interest to Lees. “When FTF is over, we’re planning to do more testing over longer periods and more widely varying duty cycles with the current line-up,” he wrote. That’ll include some lower-power 8-bitters where he doesn’t expect any Coremark upsets, “as well  a number of other ARM MCU products from different vendors, looking mainly at parts that approach 100uA/MHz active current, rather than the mainstream which are above 150uA/MHz and as high as 800uA/MHz (eg some Stellaris parts).”

Going back to the fairness of the demo test, he added that in converting from Coremark/mA, to Coremark/mW energy efficiency, “we were very fair and treated all the contenders as 3.3V nominal products, however, Kinetis L has an operating range of 1.71-3.6V, while TI and Microchip are 2.2V parts, and only Renesas RL78 operates as low as 1.6V.” He’ll also be exploring the effects of duty cycle and voltage range on useable energy efficiency.

Again, these are focused on the core, but Lees noted that Kinetis L also has  independently operating peripherals while processor, main flash and RAM, as well as DMA and bus can all be in low power states. “The peripherals have an active cross bar matrix to allow cross functionality and I/O operation even while in sleep modes, and so can monitor and perform tasks while in these low power states, further increasing battery life. This functionality of course is beyond the scope of today’s benchmarking activities, which are all targeted at core/software performance.

In truth, stats and demos are interesting, and a good guide, but not the true figure of merit for a given user or application. Ecosystems, IDEs, time to market, users’ experience with a device, a line, word of mouth, a company''s reputation for support, and many other parameter go into a final decision. 

In the meantime, what’s your take on the demo? See the above accompanying video here also.