LG Introduces LG8111 AI SoC with Eris Reference Board

10-01-2021 |   |  By Sam Brown

LG Electronics has recently introduced a new AI development board that includes a host of tools and capabilities ideal for AI applications. What does the LG8111 SoC integrate, what AI features it provides for designers, and how does it demonstrate the trend in edge computing?

LG Electronics announces the LG8111 Eris Reference Board

Recently, LG Electronics announced the Eris Reference Board release that allows users to prototype with the LG8111 AI SoC. The development board integrates all the components needed to build a single-board-computer including an ARM Cortex A53, multiple cache levels, camera engine with a dual ISP pipeline, and HDR preprocessing and video encoder with a voice engine supporting full HD video H.264.

Furthermore, the board supports Linux operating systems including Ubuntu, programming languages such as C and C++, hosts a range of I/O including GPIO, I2C, MIPI, PWM, SD, UART, and USB, Wi-Fi, and has an industrial operating temperature range of -40°C to +85°C. The small compact size of the Eris Reference Board not only makes it ideal for prototyping with the LG8111 AI accelerator SoC, but it also makes it ideal for integrating into solutions that are to be used in the field. 

The Eris Reference Board is supported by the Amazon Web Services platform (AWS), specifically the AWS IoT Greengrass service. The IoT Greengrass Service is an IoT open source edge runtime that helps engineers develop IoT solutions from development to deployment. The platform also allows engineers to monitor and manage IoT solutions no matter where they are.


What is the LG8111 SoC?

While the Eris Reference Board allows engineers to use the LG8111, the question remains; what is the LG8111?

The LG8111 is LG Electronics solution to the development of edge-computing AI to tackle power consumption, speed, and security. To start, the LG8111 supports AI tasks that involve neural nets such as those involved in video, audio, language processing, and control processing. According to LG, the neural net engine found in the LG8111 mimics the human brain and can efficiently process deep learning algorithms.

The ability to run AI in hardware is not the only advantage of the LG8111; its low power and low latency capabilities mean that AI can be moved closer to the edge. Instead of relying on a cloud service to process data, the LG8111 allows many AI tasks to run locally, thereby decreasing latency. The use of local processing also reduces power with the use of dedicated AI hardware and the lack of dependency on Wi-Fi (Wi-Fi is notoriously power-hungry). 

Furthermore, the ability to run AI algorithms locally helps to increase security in a system. One feature that devices of the future will support is preprocessing gathered data so that anyone who gains access to the preprocessed data will not extract private information. For example, a camera that performs facial recognition could stream its video content to a remote data centre for AI processing. However, this runs the risk of leaking private data whereby an attacker can access either the camera or the server. Thus, if the data is preprocessed by the device first, the resulting image data (which contains the unique information needed to perform facial recognition), is obscured enough that the original image cannot be reconstructed.

The LG8111 also takes security further and integrates cryptographic functions that help to accelerate security operations. AI platforms that the LG8111 include TensorFlow, TensorFlow Lite, and Caffe.

Read More


By Sam Brown

Related articles