This year, the ARM Cortex-M series will become the go-to family of cores for most new microcontroller applications.
The price war will be over. With prices as low as 36 cents today, 8-bit part pricing will only beat the Cortex-M0+ pricing in the highest volume applications with the least computational demands. Moore’s Law will drive the current prices down to within a few cents of the cheapest 8-bit parts. The limiting factor is packaging costs, which are the same regardless of word size.
Performance will favor the ARM Cortex processor. The best 8-bit or 16-bit EEMBC CoreMark score is 0.78/MHz. The worst Cortex-M series score is 0.98, though some Cortex-M0+ devices exceed a score of 1.5. The Cortex-M4 can hit a score of 2.19, which is about the same as the PowerPC. Thus the Cortex-M series will steal design wins from the bottom of the market up through some very demanding control applications.
STMicroelectronics and/or NXP Semiconductors will migrate some Cortex-M series MCUs to their 40 nm processes (from 90 nm today), reducing power, increasing speed, and dropping costs. Expect other key ARM licensees such as TI and Freescale to follow suit.
In 2013, ARM parts will lose out in long-lived battery applications as their sleep currents are an order of magnitude greater than the thriftiest low-end MCUs. Expect vendors of 8- and 16-bit devices to use power as a differentiating factor from ARM devices. But a CR2032 battery will only lose about a third of its capacity over a decade from the one microampere sleep current of the Cortex-M0+, and at 40 nm sleep current will be even lower. ARM will get significant design wins in ultra-low-power applications in 2014.
ARM will deprecate the Cortex-M0. No new Cortex-M0 MCUs will appear, since the Cortex-M0+ is faster and uses less power with about the same gate count.
Currently Cortex-M series microcontrollers come in relatively large packages (28 pins or more). NXP’s 8-pin LPC810M (strangely in a large DIP) is the only exception. In contrast, some 8-bit parts are available in a 6-pin SOT-23A. ARM licensees will offer the same for ultra-small applications with few I/O needs.
The Cortex-M1 (the version designed explicitly for FPGAs) will make deep inroads against proprietary FPGA soft cores In 2012 Altera released blazing ARM Cortex-A9 IP, but the Cortex-M1 will address the huge deeply-embedded market.
Though their demise will be prolonged, 2013 will be the year that 8- and 16-bit MCUs start to fade away. Few non-ARM 32-bit MCUs will appear, leaving vendors to compete using innovative I/O and multicore designs.
Jack Ganssle has written over 600 articles and six books about embedded systems, as well as one about his sailing fiascos. He has started and sold three electronics companies and now lectures and consults about the industry. He also works as an expert witness from time to time.
See the original post at http://www.mouser.com/applications/article-2013-arm/.
Filed Under: Rapid prototyping