Arm Ups Its IoT Intelligence Game With New Chips
When you think of IoT, the company you should really be thinking of is Arm. The semiconductor giant has staked out a claim on what it calls “the 5th wave of computing”—the crucial intersection of the emerging technologies of AI, IoT, and 5G. I’ve watched with interest as the company continues to build out its offerings in the area, and always make a point of attending its annual Arm TechCon conference (see my coverage of the 2018 and 2019 events for more background). Today Arm unveiled last week what will certainly be key parts of its portfolio moving forward—a selection of offerings and IP geared specifically towards machine learning and AI. In the realm of IoT, powerful AI is absolutely crucial in unlocking the value of the vast amounts of data generated by a world of connected devices. Furthermore, according to our studies and others, AI is becoming more and more mainstream and accepted. Let’s take a look at what Arm announced.
Arm Cortex-M55 processor
Arm’s bread and butter is providing the small, power-constrained processors necessary to power mobile, IoT and embedded devices. On-device processing is a necessity for these connected devices, reducing the time it takes to derive insights and make decisions, and alleviating some of the widespread security concerns around IoT by keeping sensitive data out of the cloud. The Cortex-M series is one of Arm’s flagship lines, designed for use within microcontrollers in a wide range of customer applications.
Today Arm announced the new Cortex-M55, which it is heralding as its “most AI-capable” Cortex-M processor to date. It’s the first offering from Arm to include the company’s new Armv8.1-M architecture, which features Arm’s Helium vector processing technology and promises to deliver more powerful, energy efficient DSP and machine learning performance. How much more powerful? Arm claims a significant 15X increase in ML performance and a five-fold increase in DSP performance from previous Cortex-M offerings. Another thing worth mentioning is the fact that Arm is making its Custom Instructions available for the processor, which the company says will allow customers to optimize the processor for specific workloads. All in all, it looks to be a serious step up from previous generations.
Some machine learning systems, however, require even more power. For those, Arm also introduced the industry’s very first microNPU, the Ethos-U55. Arm says that when paired with the new Cortex-M55, the Ethos-U55 can deliver a truly incredible 480X increase in machine learning performance over previous Cortex-M processors. Geared towards size-constrained IoT and embedded devices, the company touts the Ethos-U55’s configurability and advanced compression techniques, which it says saves power and reduces overall ML model sizes.
Software and security
While hardware designed for AI is crucial, it is all for naught without the correct software. Both the Arm Cortex-M55 and the Ethos-U55 are fully supported by the company’s Cortex-M software toolchain, which the company promises will give customers a unified development flow for their machine learning, DSP, and traditional workloads. Additionally, Arm says the processors feature optimizations and specific integrations with various popular machine learning frameworks such as TensorFlow Lite Micro.
One aspect that has hindered IoT adoption and scaling from the get-go is the ongoing security concerns around these connected devices. To help alleviate these worries, Arm says these processors and their reference designs feature Arm’s TrustZone technology to provide hardware-level, system-wide security.
This article originally appeared on forbes.com To read the full article and see the images, click here.
Nastel Technologies uses machine learning to detect anomalies, behavior and sentiment, accelerate decisions, satisfy customers, innovate continuously. To answer business-centric questions and provide actionable guidance for decision-makers, Nastel’s AutoPilot® for Analytics fuses:
- Advanced predictive anomaly detection, Bayesian Classification and other machine learning algorithms
- Raw information handling and analytics speed
- End-to-end business transaction tracking that spans technologies, tiers, and organizations
- Intuitive, easy-to-use data visualizations and dashboards