#USA #Los Angeles #EMASS #Nanoveu #EdgeAI

EMASS Advances Edge AI with 16nm ECS-DoT

EMASS, a subsidiary of Nanoveu specializing in advanced semiconductor technologies, recently announced a significant milestone in ultra-low-power edge AI with the successful tape-out of its 16nm ECS-DoT system-on-chip (SoC). This development marks a pivotal moment as the device has entered fabrication at TSMC, propelling the company’s edge AI architecture into production.

The Evolution of ECS-DoT Architecture

The 16nm ECS-DoT SoC is an evolution of EMASS's 22nm ECS-DoT platform, showcasing impressive advancements in compute density, memory bandwidth, and system integration. The transition to a 16nm process allows for a tighter integration of subsystems, which is critical in supporting complex always-on workloads within the limited power budgets often found in edge devices.

The 22nm ECS-DoT has carved a niche in various markets, seamlessly integrating into customer products such as wearables, industrial sensors, and smart infrastructure solutions. This ongoing market success has laid a solid foundation for the new 16nm design, allowing existing designs to upgrade their performance without needing to alter the software or architecture dramatically.

This new chip design capitalizes on increased on-chip SRAM capacity, enabling the integration of more functions and the improvements of AI and DSP acceleration—without requiring any changes to the existing programming model or software tools. The result is a compact solution that combines wireless connectivity, dedicated AI accelerators, and efficient power management into a single SoC. These features significantly enhance the low-latency operation necessary for battery-powered and energy-harvesting edge devices.

Key Features and Advantages

Mark Goranson, CEO of EMASS, noted, "Reaching tape-out confirms that our ultra-low-power edge AI approach scales cleanly to more advanced nodes." The 16nm ECS-DoT is not merely a smaller version of its predecessor; it represents a new paradigm allowing for increased intelligence and capabilities in demanding applications without overextending power or cost constraints.

Key features of the 16nm ECS-DoT include:

• - Fully Integrated BLE Subsystem: Reduces the complexity and costs associated with external wireless components.
• - Expanded On-Chip Memory: Facilitates handling larger AI models and higher throughput workloads more efficiently.
• - Adaptive Fine-Grained Power Management: Optimizes energy consumption for devices operating in always-on modes while prolonging battery life.
• - Dedicated Object-Detection Accelerators: Enhances throughput for vision workloads, reducing latency significantly for edge-based vision applications.
• - Integrated Floating-Point Unit (FP16/FP32): Simplifies developer workflows, speeding up DSP and mixed-precision AI tasks without complicating the coding requirements.

Despite changing the process node to 16nm, EMASS ensures that developers can transition their applications with minimal adjustments, thus safeguarding their investments in the existing platform while unlocking new performance avenues. The move fortifies EMASS’s philosophy, termed "Atoms-to-Apps," which aligns application needs, algorithm designs, and silicon implementations into a unified development approach.

Dr. Mohamed Sabry, founder and CTO of EMASS, expressed pride in their innovative approach, stating, "Tape-out validates years of architectural decisions. We’ve demonstrated that our design can deliver higher functionality and integration without compromising energy efficiency."

Conclusion

As EMASS embarks on this next chapter with its 16nm ECS-DoT, it not only elevates the standards for edge AI applications but also commits to maintaining energy efficiency at the forefront of its innovations. The future looks bright for EMASS as it navigates this frontier of always-on intelligence, promising enhanced capabilities while staying true to the principles that have guided its growth.

For more information about EMASS and its revolutionary technology advancements, please visit nanoveu.com/emass.