OpenFive and EdgeCortix announce AI Accelerator SoC collaboration

1 min read

EdgeCortix, an AI-focused fabless semiconductor company, and OpenFive, a provider of custom silicon solutions have announced that they are working together to develop a power-efficient and high-performance custom 12nm FinFET SoC that accelerates AI processing for real-time edge applications.

The IP driving EdgeCortix’s SoC is the Dynamic Neural Accelerator (DNA) architecture, designed with a software-first approach and which provides a coupling between neural networks and an underlying low-power domain-specific hardware architecture. This is backed with software to process applications across different machine learning frameworks.

The IP has been optimised for real-time processing and targeted business sectors include: defence, autonomous vehicles, satellite and 5G communications, smart cities and manufacturing.

EdgeCortix’s SoC features a proprietary runtime reconfigurable architecture that enables the SoC to deliver over 10x efficiency gains over traditional GPUs. OpenFive is contributing its low power, deep-submicron FinFET design methodology to the collaboration, making it possible to efficiently implement the complex SoC in TSMC’s 12nm FinFET process, together with packaging and supply chain experience to ensure a smooth ramp up to volume production.

"We are very pleased with how closely the teams at OpenFive and EdgeCortix have worked together to optimise the power, performance, and area of our SoC," said Sakyasingha Dasgupta, Founder & CEO of EdgeCortix.

Dr. Shafy Elthouky, SVP and GM of OpenFive added, "We are confident that the combination of EdgeCortix’s AI domain expertise and OpenFive’s advanced silicon implementation capabilities will enable a very competitive and efficient product for the rapidly growing edge AI market."