AI SoC Research Division
- HW/SW Technology for Hyperscale AI Neural Network Semiconductor(IC)The ever-growing AI neural networks for achieving near human intelligence require high performance semiconductor IC (Integrated Circuit) design technology and the system development technology which utilizes the IC including the software technology to efficiently operate these systems.
With the experiences in the developments of untra-low power visual AI processing chip and the AB9 chip which is Korea’s first developed NPU chip as well as the 5PFLOPS grade AI server (ArtBrain-K) and the accompanying AI-Ware system software, we are developing hyper-scale AI computing semiconductor IC and the AI semiconductor’s system SW and platform technology along with hyper-parallel processing core based semiconductor processor for super computer and 64 bit RISC-V core technology for next generation CPU.
<HW/SW Technology for Hyperscale AI Neural Network Semiconductor(IC)>
- AI Edge SoC Design TechnologyThe AI Edge SoC Research Section is researching intelligent edge semiconductor designing technology to achieve intelligence, low power consumption, and miniaturization of edge devices that collect and process data at the edge, intelligently selecting data. Our section is utilizing world-leading TEI (Temperature Effect Inversion) ultra-low power semiconductor design foundational technology and the newly emerging RISC-V open core to develop the world's best ultra-low voltage RISC-V edge processors, as well as ultra-low power neuromorphic artificial intelligence semiconductors that simulate the biological mechanisms of neurons and synapses that transmit and process information in spike units. Our laboratory is also developing intra-body communication semiconductor technology that uses the human body channel to transmit and receive edge data.
<AI Edge SoC Design Technology>
- AI Sensing SoC TechnologyTo develop an intelligent radar sensor semiconductor chip (RF transceiver, radar signal processing, data processing CMOS ASIC) for long-distance human detection useful in large-scale disaster sites (fire, collapse, burial, etc.), We are developing core technology for “Distributed Radar” that can dramatically increase the performance of existing radars so that biosignals of survivors can be detected farther and deeper by overcoming areas and obstacles. Meanwhile, we are developing radar parts and system technologies that can perform non-contact breathing, heartbeat detection, and posture estimation of living organisms at close range and complete spectroscopic imaging of objects, as well as core semiconductors for Sub-THz CMOS transmission and reception. Existing high-performance radar technology for military, port, and aircontrol purposes, which has been mainly used for identifying aircraft and ships, requires the use of high-power, expensive compound semiconductors and large-scale antennas. As the commercial market demand for sensors is rapidly growing, we are developing CMOS semiconductors, artificial intelligence signal processing, and data processing technologies for this purpose. Small-sized radar technology is used for people detection, people identification, route tracking, autonomous driving, and access monitoring. Unlike CCTV, there is no controversy over invasion of privacy, and distance inaccuracy and delay time can be reduced when reading images. And small-sized radar sensors are very promising as a next-generation intelligent sensing technology because they can detect drones up to several kilometers with low power and can be used as collision avoidance devices by being mounted on UAM, and we are developing a competitive intelligent radar CMOS ASIC for this.
<AI Sensing SoC Technology>