Versatile Edge-AI-Processor

Ultra-Low-Power-Computing

The IP-protected analog in-memory computing (IMC) architecture reduces power consumption by up to 90% compared to digital accelerators. Ideal for battery-powered devices and energy harvesting systems.

Low latency

ADELIA's high-speed processing enables real-time applications. This is crucial for real-time analytics and alarm systems, for example.

Software toolchain

ADELIA's software ecosystem with simplified deployment of customized neural networks and pre-silicon verification.

Versatility

Despite analog in-memory computing technology, ADELIA is highly programmable via a digital cyclic fabric layer and can cover a wide range of AI applications.

• Investigation (model simulation) of the benefits of neuromorphic circuits for a customer-specific requirement
• Cost estimation for the customized solution
• Hard-IP ADELIA available in 90 nm and 22 nm processes with software toolchain for model training
• Customization of the IP core and engineering support for integration into a customer SoC
• Based on the use case analysis and existing neural networks, the networks are optimized, trained, quantized and implemented so that they run optimally on the hardware platform
• Integration of AI accelerators with existing sensor front-ends

Further information on neuromorphic computing at IIS can be found here.