Machine Learning

KeyASIC offers innovative chipset and board solutions covering both IoT & AI. First is the SPG101 SoC, which is best used as main application MCU connecting IoT sensors and interfacing with various wireless connectivity as well as AI accelerators. SPG101 run on an ARM-based CPU and is capable of running rich OS such as full Linux kernel. It supports external data storage up to 32GB SD2.0 flash as well as providing comprehensive peripheral interfaces to support IoT sensor and wireless application.


ORAMA and EXYPNOS are 2 purposely-built artificial intelligence ASIC to run inference at the edge. ORAMA is a highly edge-optimized dual-core AI accelerator with capability of 16 GFLOPS while EXYPNOS is a powerful combination of IoT-and-AI-into-one SoC that run on ARM-based CPU, which also contains 16 cores with the capability of 512 GFLOPS. Both chips offer comprehensive peripheral interfaces in the likes of USB2.0, SPI, UART and GPIO. To support high-end application, EXYPNOS is designed with high-speed PCIe2.0 interface.


We also created development kits and boards for each of the IoT and AI SoCs. These boards enable easier prototyping, validation, and deployment of neural network inference applications at the edge. Besides that, we combined SPG101 and ORAMA into a custom board C2D2 to support customer that would like to go for rapid-prototype-to-production. All development kits come with complete SDK that allow developers to design, fine-tune and deploy neural network on a host computer. Get started today with our development kits to reduce design risk.

Main Features : 


●     Machine analysis at the edge computing 

●     Chipset & development boards come with complete SDK & guide 

●     Optimized power consumption over performance

●     Rich set of tools, frameworks, and libraries (not limited to below)

            *     Google TensorFlow

            *     Microsoft CNTK

            *     Caffe

●     Target application:

            *     Image & video processing

            *     Deep Neural Networks &Vector matrix computations

            *     SDR navigation & communication

            *     Radiolocation & 3D computer vision

            *     Hybrid high-performance computing systems

How it works