Synaptics Expands Edge AI Offerings with Advanced Adaptive MCUs for Multimodal Context-Aware Computing

Key Takeaways

Synaptics has launched the SR-Series microcontroller units for Edge AI, focusing on performance, efficiency, and ultra-low-power operations.
The SR-Series supports a range of IoT applications, featuring advanced processing capabilities and extensive peripheral options.
It combines Synaptics’ AI expertise with Arm technology to enhance device intelligence and contextual awareness in real-time applications.

Advancements in Edge AI Technology

Synaptics Incorporated has introduced its latest innovation, the SR-Series microcontroller units (MCUs), as part of its award-winning Synaptics Astra AI-Native platform. Designed for scalable context-aware Edge AI, the SR-Series comprises three tiers of operation: performance (100 GOPS), efficiency, and ultra-low-power (ULP) always-on (AON) modes, providing flexibility in power consumption and processing capabilities.

At the heart of the SR-Series is the Arm Cortex-M55 core, paired with the Arm Ethos-U55 neural processing unit (NPU). This advanced architecture is supported by the Astra Machina Micro development kit and an open-source software development kit (SDK), making it ideal for a variety of multimodal Internet of Things (IoT) workloads. The MCUs are specifically optimized for consumer, enterprise, and industrial applications, incorporating accelerators as well as adaptive vision, audio, and voice algorithms.

The compact design of the SR-Series allows for a rich set of peripherals, including multiple camera interfaces, aimed at reducing system costs and power consumption while maintaining a small footprint. This functionality makes the SR-Series suitable for a wide range of devices, such as battery-operated security cameras, various sensors, appliances, point-of-sale systems, digital signage, and scanners.

Debuting at EW2024 alongside the SL-Series microprocessor units (MPUs), the Synaptics Astra AI-Native platform aims to merge scalable, low-power computing technology at the Edge with user-friendly software and tools, along with Veros wireless connectivity. This platform draws on Synaptics’ established expertise in neural networks and AI hardware, coupled with extensive support for diverse modalities.

Vikram Gupta, Senior Vice President and General Manager of IoT processors at Synaptics, emphasized the pivotal moment in Edge AI development, stating, “Embedded developers have a unique opportunity to redefine human-machine interaction through multimodal processing and contextual awareness.” He noted that the SR-Series enhances their processing roadmap with customized intelligence for varying power levels, allowing for cognitive IoT devices capable of adapting seamlessly to their environments.

Paul Williamson, Senior Vice President at Arm, highlighted the transformative potential of ultra-low-power AI processing for sectors like retail and smart home applications, remarking that the SR-Series meets the increasing performance demands in these markets. He affirmed, “Synaptics is delivering the real-time intelligence and innovation needed to scale edge AI deployments.”

Jim McGregor, Principal Analyst at TIRIAS Research, acknowledged the necessity for multimodal processing to achieve situational awareness in future intelligent edge devices. He described the SR-Series as a scalable solution that optimizes power and performance while simplifying device integration.

The SR-Series includes three MCUs: the SR110, SR105, and SR102, each tailored for different multimodal applications. All incorporate a Cortex-M55 core with Arm Helium technology clocking up to 400 MHz. Notably, the SR110 features an additional Cortex-M4 core and an Ethos-U55 NPU, currently available for sampling, while the SR105 and SR102 focus on varying configurations of the Ethos-U55 NPU.

Additional features of the SR-Series include:
– Up to 4 MB of system memory with ULP AON memory
– Streaming vision and audio processing capabilities
– MIPI-CSI camera input and passthrough functionality
– Low-power image signal processing
– Secure OTP, TRNG, AES-256, RSA-4096, and SHA-512 capabilities

The Machina Micro kit ensures that the Astra platform provides an accessible AI development experience for both newcomers and seasoned developers alike.

The content above is a summary. For more details, see the source article.