OpenEdge Launches IP for 'On-Device AI' NPU

4x Performance Improvement Compared to Existing IP

OpenEdge Technologies, a semiconductor intellectual property (IP) platform company, announced on the 16th the launch of the neural processing unit (NPU) IP 'ENLIGHT PRO,' which delivers up to 4 times the performance of its previous products.

An OpenEdge engineer developing an NPU IP /
[Photo by OpenEdge]

This new IP product is a successor to the existing 'ENLIGHT.' It features enhanced scalability and flexibility to support new neural networks. It was developed as a neural network accelerator supporting core components of large language models (LLM) in energy-constrained environments.

ENLIGHT PRO offers 4 times the computational performance compared to its predecessor and supports fast interaction between memory and computation. Additionally, it uses the open-source RISC-V architecture to minimize semiconductor design costs. The company also provides a software development kit (SDK) with the new product to improve user convenience.

OpenEdge explained that ENLIGHT PRO is suitable for designing NPUs used in on-device AI (AI running on devices) such as fully autonomous driving, cameras, and mobile devices. The company also announced plans to expand its IP business in the autonomous driving sector going forward.

Lee Seong-hyun, CEO of OpenEdge, stated, "We plan to obtain international certification for automotive functional safety (ISO26262) in the second half of the year. Based on this, we will expand licensing agreements with global clients by utilizing verified IP for autonomous driving."