Nepes, led by CEO Byeong-gu Lee, an advanced semiconductor back-end process foundry company, recently announced the successful development of 'METIS', an intelligent semiconductor designed for edge computing.
In April 2021, Nepes Artificial Intelligence Research Institute was chosen to lead the 'Development of Intelligent Edge Computing Semiconductor for Lightening Manufacturing Inspection Equipment' project under the 'Artificial Intelligence Semiconductor Application Technology Development Project' by the Institute for Information and Communication Planning and Evaluation (IITP) under the Ministry of Science and ICT. This research was conducted in collaboration with the Korea Institute of Electronic Technology and Hanyang University.
The research team focused on creating a deep learning model for ultra-lightweight/ultra-precise manufacturing inspection equipment using an automatic lightweight software framework. The aim was to achieve ultra-lightweight/miniaturization, low-power consumption, and cost-effectiveness compared to existing server- or PC-based manufacturing inspection systems. Additionally, a deep learning acceleration IP optimized for data reuse and parallel computing processing was developed and integrated into Metis.
As a result, Nepes has developed a neural network-based deep learning defect detection algorithm that enables the transition from centralized server-based AI systems to terminal-based ones, while securing low-power SoC platform technology to support this transition. These advancements aim to enhance productivity and efficiency in manufacturing environments.
Metis incorporates the cx-BGA (Ball Grid Array) of nePACTM, Nepes' cutting-edge 2.5D & 3D package platform. nePACTM leverages multi-layer and fine RDL wiring based on fan-out technology and flip chip bonding technology, making it well-suited for highly integrated, high-performance chips like artificial intelligence semiconductors.
Yeonsook Park, head of Nepes Artificial Intelligence Research Center, stated, "Through this research, we aim to address network traffic issues in existing cloud server-based systems and develop intelligent processor technology capable of detecting edge defects. This will enable the development of various deep learning-related applications needed across the industry."
Nepes is also actively involved in national projects focused on developing core packaging technologies for system semiconductors and AI semiconductors. This includes the 'Development of chiplet heterogeneous integration ultra-high-performance artificial intelligence semiconductors' project, with a total project cost of KRW 44 billion, hosted by the Ministry of Science and ICT. Additionally, they are working to secure original technology for AI semiconductors specialized for on-device applications.