Internet of Stuff: Chip rivals try to stop Cortex-M7 from flexing ARM’s muscle
Processors, microcontrollers start to collide
The Internet of Things (IoT) is growing an estimated five times more quickly than the overall embedded processing market, so it's no wonder chip suppliers are flocking to fit out connected cars, home gateways, wearables and streetlights as quickly as they can.
However, the sector is so new that there is considerable uncertainty about the precise capabilities which will be required, and new bars are set in critical areas like performance:power consumption ratios all the time.
That fluidity is illustrated by British chip designer ARM’s latest addition to its Cortex-M family of cores for microcontrollers. The new M7 pushes the performance of a microcontroller almost into the realms of a microprocessor, nipping at the heels of its own Cortex-A designs and Intel’s Quark.
For chip providers, the IoT presents a huge variety of target devices, but also presents the challenge of balancing the massive volume potential of low end "things" and sensors, with the higher prices and margins of full processors.
The launch of Quark saw Intel firmly staying at the high end of the IoT, with an architecture which drove down power consumption to ARM-like levels, but retained the level of processing power needed for products like gateways. The giant will probably not venture into the low end of the embedded space, where 8-bit and 16-bit microcontrollers (MCUs) still rule, rather than 32-bit platforms like Cortex-M7, but that does not mean the low end is not growing.
In quite large percentages of those projected billions of IoT devices, the very basic circuits - ultra-low power and with simple wireless connectivity - will always suffice, because many IoT devices will not need a full TCP/IP stack or IPv6, which are the big drivers behind 32-bit, and it is still easier to drive power levels right down with an 8-bit architecture.
Companies continue to develop 8-bit MCUs around the venerable 8051 core, including Silicon Labs with its recent Si106x/108x range of wireless microcontrollers for the IoT. Thomas Barber, director of marketing for wireless products at the firm, said in an interview: "Estimates suggest there may be 30 billion edge devices.
They can't all consume 1W because we don't have enough energy, so they need to consume microWatts and they need to be inexpensive." Although Silicon Labs also offers ARM-based MCUs, Barber sees the humble 8051 core surviving, commenting: “Every time there's an 'obit' for the 8051, it keeps going and we're not betting on it going away. ARM sees the Cortex-M0 as an 8bit killer, but it will come down to customer preference."
The rising performance of MCUs
Generally, though, there is a trend for more performance at the edge, so the challenge is to bring 8-bit power efficiency levels to bigger architectures. To the delight of chipmakers, the IoT is re-placing one chip with several in many instances – for instance, an Cortex-M0+ sensor hub coupled with a Cortex-M4 or –M7 doing the apps processing. And standalone MCUs will increasingly be superseded by SoCs integrating more memory and peripherals.
And the need to IP-enable devices, and to increase their intelligence, means that much of the innovation is going into the overlap territory between souped-up 32-bit microcontrollers like ARM’s, and low power microprocessors and SoCs with full operating systems. And for the suppliers, these not only carry better prices, but can enable whole new categories of devices, such as wearables, and by so doing, drive incremental revenues.
ARM, as an IP supplier, does not need to make a firm choice. It needs only to ensure it covers every possible option customers might choose when planning their silicon. The launch of the new M7 shows it doing just that, with a 32-bit microcontroller design that could power smart home appliances, home gateways and industrial controls – the kind of higher end "things" that might also run on a micro-processor like Quark.
Thomas Ensergueix, senior product marketing manager in ARM’s CPU group, told VentureBeat: “We could only dream of this kind of performance before. It is a quantum leap forward. We are bringing intelligence to what used to be dumb objects.”
The Cortex-M7 core
This positions the M7 as ARM’s premier product in terms of growth potential in devices. Its launch is particularly significant since the UK company has recently suffered from some fears about slowing growth in smartphones and other high-power mobile devices. These have been the driver of its growth and market share (over 90 per cent), but the IoT is now set to usurp that position, even though its different price points may put pressure on ARM’s profits.
That factor makes it important that the company packs as much value into the M7 as possible in order to target higher priced chips and appliances. So the offering has some of the capabilities of a processor, but with important advantages in terms of latency and power consumption, both critical to many IoT applications.
The Cortex-M7 boosts the 32-bit MCU’s capabilities when translating data from sensors into digital information. It doubles the compute and DSP performance of the current top end family members while reducing power consumption, thanks to the 28nm manufacturing process It is also more powerful than ARM’s Cortex-R real time processor family. It will be able to support embedded devices with voice control or graphical interfaces and can also be used within smartphones or cars to drive more complex audio and touch display functions.
The M7 supports 64-bit data transfer and can execute two instructions in parallel, which will be important for markets where instantaneous response to changing conditions is required, such as lighting management. In many such cases, a collection of small microcontrollers could be replaced with a single M7, said ARM. The design runs a real time operating system, which improves latency rates tenfold compared to a full OS.
The design is targeted in particular at the intensive real time processing applications found in some industrial and automotive settings, and at some smart home gateways, and as such does overlap, in some targets, with the Cortex-R. It has already been adopted by several chip suppliers, including STMicro, Atmel and Freescale. Freescale will add M7 to its Kinetis portfolio of over 1,000 MCUs, citing applications such as motor control and power conversion. STMicro, meanwhile, unveiled the STM32 F7, based on the new core, and claiming to be the highest performance 32-bit MCU on the market.
ARM will hope such wins boost an existing trend. In the first quarter of this year, half of its new CPU IP licences were for Cortex-M, including four brand new customers, and CEO Simon Segar commented: “Cortex-M chips can be found in most of the internet of things and wearable devices that have been announced to date... We are now seeing the second generation of wearable and internet of things devices...
"This is still a very fragmented end market but with many of the chips going into these devices being based on ARM the benefits of our ecosystem make it easier for developers to create new products."
All this means that Quark may be a strong challenger to some ARM-based embedded processors, but it will also have to look over its shoulder at a new wave of higher end MCUs. It works the other way too of course. Freescale’s Geoff Lees told New Electronics recently that “although we are a lead partner for Cortex-M, we are beginning to see more IoT applications suited to low end Cortex-A processors, rather than high end Cortex-M cores. Application processors are coming to the domestic market, and that's something that was unexpected."
But ARM will not have this arena to itself in the way it has done with smartphones. As well as Intel, it may meet MIPS again in low power processor cores – the latter, now owned by ARM’s GPU core rival Imagination Technologies, is focusing hard on so-called WPUs (wearable processing units).
Perhaps it is more likely to be wrongfooted by an unfamiliar challenge, perhaps from the open source microcontroller space. Open source hardware is gathering momentum, especially in the crea-tive and fragmented, "anything goes" world of the IoT. There will always be an attraction, in a price-sensitive space like this, in finding an architecture that does not charge a tax in the form of an ARM or MIPs licence.
Qualcomm Snapdragon comes to robots
The Open Source Robotics Foundation has announced plans to extend the capabilities of its Robot Operating System (ROS) to Qualcomm’s Snapdragon 600 processors.
The Foundation will create and support a ROS release for Snapdragon running Linux or Android, and will test, refine, and integrate support for the ARM instruction set into its development efforts. The organisation says the worldwide robotics community will benefit from the opportunity to work on lower power ARM chips in order to create smaller, more efficient robots with longer battery life, while still using an open OS.
"As adoption of ROS continues to increase, our developer community wants to incorporate the latest computing platforms," commented the Open Source Robotics Foundation’s CEO Brian Gerkey. "Given the intersection between robotics and mobile and embedded systems, we believe that offering Snapdragon's SoC capabilities to our users will be a big hit."
Copyright © 2014, Wireless Watch
Wireless Watch is published by Rethink Research, a London-based IT publishing and consulting firm. This weekly newsletter delivers in-depth analysis and market research of mobile and wireless for business. Subscription details are here.