There’s no doubt that building Internet-of-Things (IoT) systems are a complex effort bringing software, OEM hardware, and silicon in synch. Being able to develop software before hardware is available is an essential asset to getting to market on time. But with security, machine learning (ML), and connectivity all part of the puzzle, the challenge only gets more daunting.

Arm is juggling many balls in the air to address those IoT hurdles. In October, Arm unveiled its Arm Total Solutions initiative, a framework comprising a virtual hardware capability and an ecosystem for the Arm Cortex-M called Project Centauri.

Today, Arm is expanding those efforts in three ways: 

• New versions of Arm Corstone subsystems for Cortex-M and Cortex-A
• New Arm Cortex-M85 processor with high-performance and improvements to ML and security
• A broader slate of Arm Virtual Hardware offerings to enable virtual development for Arm processors, third party hardware, and embedded devices

Arm’s updated IoT roadmap. Image used courtesy of Arm

In this article, we’ll describe each of these new offerings and share insights from our pre-briefing with Arm’s Mohamed Awad, vice president of IoT and Embedded Arm.

Arm’s virtual hardware strategy is centered on using pre-integrated, pre-verified IP subsystems under the brand Corstone. In its October announcement, Arm introduced its virtual hardware platform, starting with the Arm Cortex-M based Corstone-300—a solution designed for keyword spotting. Today, the company has rolled out two more such subsystems: A Cortex-A based Corstone-1000 and the Corstone-310, which is based on Arm’s new Cortex-M85.

Corstone-1000 for Cloud-Native Edge Devices

The Corstone-1000 offers an architecture that can run operating systems like Linux. Arm says that applications such as smart cameras, point-of-sale (POS) terminals, robots, and IoT gateways require computing platforms that can run a rich operating system like Linux. Corstone-1000 is aimed at cloud-native-edge devices.

The platform leverages an Arm Cortex-A32 core, but it is designed to scale up to a Cortex-A53. The design also embeds Cortex-M3 and Cortex-M0+ cores, which creates a heterogeneous system that is both power efficient and designed for security from the ground up, according to Arm.

The Corstone-1000 platform can run security resources in a secure enclave. Image used courtesy of Arm

In order to ensure that operating systems like Linux simply work, the Corstone-1000 comes already compliant with SystemReady-IR. Part of Arm’s certification program, SystemReady-IR is a set of system certifications focused on the requirements of the embedded Linux/BSD ecosystem for Arm SoC-based systems.

The M3 and M0+ cores on the Corstone-1000 can use a separate OS and system resources to serve as a secure enclave to store secrets and process sensitive data. That makes the platform immediately ready for PSA Level 2 certification.

M85-based Corstone-310 for High-Performance

Arm’s second platform announced today is the Corstone-310 subsystem. It is pre-integrated with the new Cortex-M85 and the Arm Ethos-U55 core. Together, these result in Arm’s “highest ever performance MCU-based design.” Voice recognition is the key purpose of the Corstone-310. It is aimed at systems including smart speakers, smart thermostats, drones, and factory robots.

Arm says that the Corstone-310 was built specifically to leverage the capabilities of the new Cortex-M85 core, which is launching today. We’ll get into more detail on the M85 below. The Corstone-310 also has an optional Ethos-U55 NPU (neural processing unit) for ML.

The Corstone-310 platform is based on Arm’s new Cortex-M85. Image used courtesy of Arm

According to Arm’s Mohamed Awad, the new-Cortex M85 fills a need for MCU-based IoT system developers who find themselves forced to make tradeoffs in performance versus features. Awad explains these challenges.

"Microcontroller software developers are facing enormous challenges as the IoT evolves and they look to add more and more functionality and intelligence to their products. The developers either need to find ways to squeeze more processing per clock cycle out of their current microcontroller, or they need to drop features. Neither is ideal. Both are time consuming. Beyond that, security is always Imminent threat."

For their part, the Arm Cortex-M family processors offer a number of aspects to ease those tradeoffs, says Awad. They provide determinism, short interrupt latencies, full power management modes, and integrated hardware. For its part, Awad says the Cortex-M85 provides additional performance muscle and security functionality to further ease IoT system design roadblocks.

A Closer Look at the New Cortex-M85

Awad says that the M85 has an enhanced microarchitecture that increases scalar performance. It also makes use of Arm’s Helium technology to enable ML use cases. An example is high-fidelity audio processing without the need for a DSP.

Security is a particular challenge for IoT or embedded devices. As connected devices, there is always a risk of malicious or unintentional exposure of confidential data. To face those issues, the Cortex-M85 supports TrustZone for Arm. The core also is the first Cortex unit that integrates the Pointer Authentication and Branch Target Identification (PACBTI) extension from the new Armv8.1-M architecture.

Awad says this simplifies and accelerates the developer’s efforts to achieve PSA-certified Level 2 security. He says that MCU/MPU partners like Renesas Electronics, NXP Semiconductors, and STMicroelectronics are showing interest in the Cortex-M85 for the prospect of using it on their secure, high-performance MCUs and MPUs.

As for performance, Awad says that the M85 is the first Cortex-M core to benchmark to over 6 CoreMarks/MHz and more than 3 DMIPS (Dhrystone MIPS)/MHz. This level of performance is attained by features at the microarchitecture level. Awad says these include optimized dual-issue and selective triple-issue capability, improved branch prediction, and an enhanced memory system. The M85’s performance is 30% higher than Arm’s previously fastest core, the Cortex-M7.

Expansion of Arm Virtual Hardware Choices

In its Arm Virtual Hardware rollout last October, the company introduced its cloud-based development platform. The platform enables developers to craft and test software on fully accurate models of Arm SoCs. This enables the Arm ecosystem to easily adopt cloud-based development and CI/CD (continuous integration and continuous delivery/continuous deployment) and develop without needing large custom hardware farms.

In today’s announcement, Arm is releasing several new virtual devices to expand the Arm Virtual Hardware portfolio. Arm now has an updated Arm Virtual Hardware page. New additions will include Arm Virtual Hardware for the new Corstone designs previously discussed, plus virtual hardware support for seven Cortex-M processors ranging from the Cortex-M0 to the Cortex-M33. Meanwhile, Arm is expanding the library with third-party hardware from partners including NXP, STMicroelectronics, and Raspberry Pi.

Expansions to Arm Virtual Hardware include support for more processors, integrated dev tools, and support for third-party hardware. Image used courtesy of Arm

These expansions to the Arm Virtual Hardware ecosystem are expected to enable independent software vendors and cloud service providers to leverage the billions of Arm-based IoT and embedded devices already designed and deployed.

Project Centauri Ecosystem Additions

Last, but not least, of today’s Arm IoT announcements, Arm is adding to its Project Centauri ecosystem. Part of the October news, Centauri is an ecosystem comprised of device and platform standards and reference implementations aimed at security and cloud integration—all for the Arm Cortex-M. Today’s news includes the first release of Arm’s Open IoT SDK Framework.

The SDK framework contains the new Open-CMSIS-CDI software standard, a community-driven project hosted by Linaro.org. The standard defines a Common Device Interface (CDI) for the Cortex-M ecosystem. Awad says that eight companies are already involved in the standard, including silicon partners, cloud service providers, OEMs, and ODMs. The framework also includes software for voice recognition and keyword recognition.

The Future of Large Scale IoT

IoT has come a long way from when it was a mere buzzword over a decade ago. Today, developing IoT devices for a large-scale system is possible, but new challenges always complicate such designs. Arm’s expansion of its virtual hardware and its related ecosystem solutions have dramatically fleshed out what they claimed to aspire to six months ago. It will be interesting to observe the next steps in this path forward.