The belief that IoT would become Arm’s key growth driver never played out in the way owner SoftBank Group expected, but the British chip designer is still doing everything it can to keep IoT developers busy with a variety of new offerings that will to speed up development significantly.
On Tuesday, Arm announced its first major expansion of the Arm Total Solutions for IoT program, consisting of pre-integrated subsystem designs that take the guesswork out of chip designers, the Arm Virtual Hardware cloud service for testing Arm-based devices without a physical need Silicon and multiple software components that developers can reuse across multiple devices.
Accompanying the expansion was the unveiling of the new Cortex-M85, which Arm says is the fastest Cortex-M CPU core blueprint yet for IoT devices powered by microcontroller units.
Mohamed Awad, Vice President of IoT and embedded at Arm, narrates The registry that Arm developed the IoT program extension based on feedback from developers and device manufacturers, who are facing increasing pressure to bring new and more powerful products to market faster.
“They face this ever-increasing demand for higher performance, faster time-to-market, increased security, easier development, and then more secure processing. And that’s the only way they can continue to scale,” he said.
We’d also like to think that Arm is doing whatever it takes to prevent IoT developers from migrating to RISC-V, an open-source instruction set architecture that has been gaining notoriety following the announcement of Nvidia’s now-failed plan to acquire Arm in 2020 won. Arm is also gearing up for a possible IPO, which could reportedly happen later this year.
New speech recognition subsystems, cloud-native edge devices
A key part of the Arm Total Solutions for IoT program is Corstone, a collection of pre-integrated subsystem designs that bring Arm’s CPU core blueprints together with other key building blocks the company believes are best for different types of applications.
These other types of building blocks can include component designs for power control, system control, system peripherals, interconnects, as well as security-oriented things like secure debugging, secure enclave, and the enterprise’s TrustZone Protection Controller.
The rationale is that since Arm knows which components work well together for different applications, why not just do this integration up front so chip designers at other companies don’t have to waste valuable time figuring this out.
Awad promised developers can still make their own tweaks, and so far this style of chip design has helped companies release more than 200 designs since Corstone started three years ago.
“It’s important to say that we’re working really hard to find that crucial balance of providing all the essential ingredients for a complete solution, while still maintaining enough freedom and flexibility for developers to be creative and add differentiation,” he said.
Last October, Arm introduced the Corstone-300 design for keyword-recognition applications, and today the company expanded it with two new designs for devices with higher performance requirements: Corstone-310 for speech-recognition applications and Corstone-1000 for cloud-native edge devices.
With a focus on voice recognition, Corstone-310 is well-suited for MCU-powered devices ranging from smart speakers and thermostats to drones and factory robots.
The Corstone-310 design leverages Arm’s new MCU-class Cortex-M85 core, for which the company boasts a significant increase in both scalar performance and machine learning performance over the Cortex-M55. This difference in performance is even greater when compared to the Cortex-M7.
“I like to call it the king of the M-Class lineup. It brings Cortex-M closer to the performance we’ve come to expect from Cortex-A, but with the convenience, determinism, and low power consumption of an MCU,” he said.
Cortex-M85 leverages Arm’s Helium technology to boost machine learning and digital signal processing workloads. It also features enhanced security via Arm TrustZone technology and features a new architectural feature called Pointer Authentication and Branch Target Prediction to help developers achieve Level 2 PSA certification for their chips.
In addition to Cortex-M85, Corstone-310 also optionally supports Arm’s Ethos-U55 NPU blueprint, which contributes to it being the company’s most powerful MCU-based subsystem design.
The Corstone 1000 subsystem design, on the other hand, is intended for devices that require higher levels of performance and need to run on operating systems such as Linux. This includes point-of-sale systems, edge gateways and high-end smart cameras.
As such, Corstone-1000 relies on both a Cortex-M core and a more powerful Cortex-A core that can range from the A32 to the A53. The subsystem design also includes a secure enclave that can be PSA certified to handle sensitive information.
Corstone-1000 is part of Arm’s SystemReady certification program, which ensures it meets a set of hardware and firmware standards to allow chips with the subsystem design to “just work”. It also supports Arms Project Cassini, which aims to simplify cloud-native software experiences for devices running on Cortex-A-based processors.
Arm Virtual Hardware expands to more hardware
Another key way Arm is trying to accelerate IoT development is through Arm Virtual Hardware, a service available on Amazon Web Services that allows people to develop software for processors without holding physical silicon to have to.
The news here is that the cloud service has now been expanded to support the two new Corstone subsystem designs as well as seven Cortex-M cores ranging from M0 to M33. According to the award, this will make it much easier for developers to test new software for devices with Cortex-M designs.
“By adding support for seven of our most popular Cortex MCUs, Arm Virtual Hardware will immediately support the 80 billion devices that our ecosystem has already deployed. Software developers can immediately start developing, testing and verifying their software on all of these devices without having to deal with all of this hardware,” he said.
The service now also supports ARM-based hardware from partner companies such as NXP Semiconductors, ST Microelectronics and the makers of Raspberry Pi. This expands the scope of IoT software testing for developers to even more devices, Awad added.
“Because they can leverage the breadth of the cloud and not have to build hardware farms, they can take full advantage of modern development flows, things like continuous integration and streamlined ML DevOps and simplified security,” he said.
The other new development is that Arm Virtual Hardware can be integrated into existing development tools and environments. This includes Arm’s own Keil MCU development kit, as well as code repository GitHub and ML DevOps tools like Edge Impulse.
The latest major update to the IoT program is Arm expanding its efforts to create a “consistent set of standards” across a wide range of devices so developers can easily reuse different software components for standard things that don’t need to be new each time to be written.
Arm is doing this with the first release of its Open IoT SDK Framework, which includes the community-driven Open-CMSIS Pack, which includes a set of standardized interfaces and interoperability technologies for Cortex-M-based devices. It also includes software for IoT applications such as speech recognition and keyword recognition.
“By defining how they are accessed, we enable developers to have a level of software reuse and leverage that simply didn’t exist before,” said Awad. ®