With the newly designed Cortex-M52 chip, Arm enables AI computing even to the smallest IoT devices at the edge.
Translated from Arm Pushes AI into the smallest IoT devices with Cortex-M52 chip, by Jeffrey Burt has been a journalist for over 30 years, covering tech news for more than 20 years. After more than 16 years at eweek, he worked as a freelance tech journalist covering everything from data center infrastructure and collaboration technologies to artificial intelligence, cloud computing, quantum computing, cybersecurity, and channel collaboration. AI is everywhere these days, but where it really needs to be applied is in the deepest depths of edge computing, where Internet of Things (IoT) devices are generating massive amounts of data that needs to be analyzed, where insights can be collected and taken, and where machine-Xi-optimized computing can be run. Paul Williamson, senior vice president and general manager of Arm's IoT business line, said, "Only then will we be able to truly scale IoT and drive further innovation and deployment that we believe exists. But developers face challenges. They needed to extend the hardware capabilities of their platform, but they also needed a simplified software development platform. Arm says it is offering the platform through its new Cortex-M52, the latest addition to the chip design company's low-cost and energy-efficient Cortex-M microcontroller family for the smallest IoT and embedded devices, Williamson told reporters at a press briefing. "The convergence of advances in AI and the ubiquity of connected technologies means that local intelligence is possible in small and cost-sensitive devices, and they can become smarter and more powerful," he said. With less reliance on the cloud, these devices can operate with great privacy and reliability. ”
Williamson also said, "To realize this opportunity, chip providers and developers need access to more AI capabilities to deliver the intelligence needed in the cost- and power-constrained areas typical of small embedded devices." ”
The AI-in-IoT space is a rapidly growing space as organizations turn to AI and machine Xi to make sense of the data generated by these devices (there are now around 15 billion IoT devices). Pragma Market Research expects this market segment, which generated $10.3 billion last year, to grow to $91.7 billion by 2032. Rob Enderle, principal analyst at the Enderle Group, told The New Stack:"This is at the heart of IoT success, which is to try to put as much intelligence into the endpoint as possible to reduce the load on the associated cloud services, but it has to be very efficient because there is usually very little power available in the endpoint, including batteries and sometimes solar. "ARM built the Cortex-M52 to incorporate the company's Helium technology, which is ARMv8The extension of the 1-M architecture brings performance improvements to machine Xi and digital signal processing (DSP) applications. With Helium, the Cortex-M52 will provide a boost in DSP and machine Xi performance without the need for dedicated DSPs or machine Xi accelerators, or the Arm ethos neural processing unit (NPU), which can be found in Arm's high-end Cortex-M85 and mid-range M55.
The performance gains from Helium allow developers to deploy more complex machine Xi algorithms. The Cortex-M52 was created by ARM in collaboration with ARM China's engineering team and provides a home for workloads now running on Cortex-M4 and M33.
It brings 5 to the work of machine XiA 6x performance boost, bringing 27x performance improvement and 2 times lower power consumption compared to the previous Cortex-M generation1 times.
Wafer area has also been reduced by 23%, giving wafer manufacturers more options when making trade-offs between performance and cost.
In terms of security, the Cortex-M52 uses the latest ARMv81-M Pointer Authentication and Branch Target Identification (PACBTI) extensions and ARM's Trusted Execution Environment (TrustZone) technology. It will also help chipmakers meet PSA Level 2 certified silicon wafer requirements to create PSA-certified IoT devices. For developers, it opens up a wide range of use cases, including vibration, anomaly and keyword detection, and sensor fusion.
Enderle said"It provides more options for optimizing IoT solutions. "
The new chip design provides a unified software development environment and Cortex-M toolchain. Other Cortex-M chips contain embedded**, DSP**, and neural network models. All of this is integrated into a software development process in the Cortex-M52, providing programmers with an easier development path that is compatible with common machine Xi frameworks and existing tools.
Williamson said"Developers can code for common APIs in a single language to achieve the desired performance gains in the DSP and ML elements of their applications. They don't need to know the specific hardware details of the underlying processor. "
Software developers need DSPs and machine Xi performance to take advantage of the role of AI. Previously, this meant using CPUs, DSPs, and NPUs.
He said:"This means that they will have to build the hardware, and once that is, they may have to write, debug, and link multiple processors across multiple chips or within a single design, I may need three separate toolchains, compilers, debuggers, and developers will have to have a very deep understanding of timing, memory access, and scheduling of events across multiple processors. A very complex task. "
He said the chipmaker already has Cortex-M52, which will start appearing on silicon wafers next year.
From Intel, AMD and Nvidia to a growing number of smaller companies are struggling to bring more AI capabilities to IoT and embedded devices, but according to Enderle, Arm"The potential big ** comes from the RISC-V, which has been making progress to replace these processors with ASICs. "About 10 years ago, RISC-V hit the industry landscape with the advent of an open-source chip design that RISC-V International is developing as an alternative to x86 and ARM. This is the slogan presented by RISC-V International at the recent RISC-V Summit, driving the idea that the RISC-V architecture can be used in any device, from IoT and the cloud to PCs and data center servers. It's gaining some traction, with Apple adding controllers to its silicon wafers, and other companies such as Meta, AMD, and Qualcomm are working on this architecture.
At the moment, ARM seems to be in a good position. Although the RISC-V core market share is expected to increase to 10 billion units in the next few years, it is estimated that Arm has shipped 100 billion Cortex-M devices by then.
However, RISC-V brings an atmosphere of openness that developers and organizations have become Xi to through Linux and other open source software. Companies can license it to create their own version of the chip.
It's cheaper and supporters are more open," Enderle said. "Many ARM developers have switched or will switch to this technology. Qualcomm and AMD are also working on this technology. ”
The open, free nature of the RISC-V has also been exploited by China, which has embroiled the technology in an ongoing processor controversy between China and the United States. U.S. lawmakers and Biden** are considering restricting U.S. companies that want to export RISC-V technology to China.
During a Q&A session with reporters, Williamson sidestepped questions about the Cortex-M52 as a possible response to China's growing interest in RISC-V, saying that Arm has been "highly focused on DSP and machine Xi performance for years" and that this is "the next step in bringing it to devices with lower performance, lower power consumption and more constraints." This is a key long-term development focus for us. ”
He also spoke about ARM's significant advantages over RISC-V in terms of partners and software ecosystem, pointing to the wide range of features in the ARM-based market, from the smallest embedded devices to large servers, with consistent libraries and tools.
The reason we have a strong ecosystem of partners investing in our tools is that they know that when they produce a tool, then it can use the technology across all these different areas that are deployed across all of these products and market areas. "It's a very valuable thing for us and our ecosystem. This consistency exists to drive scalability. ”
However, while Enderle agrees that ARM is leading the way in its ecosystem, RISC-V is still working hard to build its capabilities and partnerships and is gaining traction.
Recent pricing actions and lawsuits against licensed users have scared developers into the platform, while RISC-V is very similar but doesn't have those baggage," the analyst said.