Edge AI Platforms for Industrial and Robotic Systems
Edge AI systems bring computing power to machines, making it possible to run AI models directly in robots, factories, and medical devices.
ADLINK Technology has introduced a new set of edge AI platforms built on NVIDIA IGX Thor and Jetson Thor, aimed at running large AI models in industrial and robotic systems. The systems are designed to deliver higher compute, better connectivity, and built-in safety features compared to earlier IGX Orin-based platforms, making them suitable for applications such as humanoid robots, medical imaging, and autonomous machines.
– Advertisement –
The lineup includes two main categories. The DLAP-IGX Series targets high-end deployments that need real-time processing and strict safety control. The DLAP-700 Series focuses on compact systems for space-constrained and embedded use cases, while still supporting multi-model AI workloads.
The DLAP-IGX Series combines an integrated NVIDIA Blackwell GPU with an optional discrete GPU to support demanding AI tasks. It delivers up to 4,293 TFLOPS (FP4-sparse) and supports high-speed data transfer through an NVIDIA ConnectX-7 SmartNIC with dual 200GbE ports. The platform includes a functional safety setup with a dedicated safety island on the SoC, a safety MCU on the board, and a board management controller for monitoring and control. It also supports PCIe Gen5 expansion and USB 3.2 for connecting peripherals.
Within the compact range, the DLAP-701 is designed for applications such as medical image analysis and high memory bandwidth workloads. It runs on NVIDIA Jetson T5000 or T4000, delivering up to 2,070 FP4 TFLOPS. The system uses a 14-core ARM Neoverse-V3AE CPU with up to 128GB LPDDR5X memory. It includes dual GbE LAN and a QSFP port supporting 4×25GbE, and operates between -10°C and 35°C.
The DLAP-711 is built for humanoid robots, vision sensing systems, and autonomous mobile robots. It uses the same processing setup as the DLAP-701 but adds more networking options, including four GbE ports, two 100M LAN ports, and a QSFP interface. It supports operation in a wider temperature range from -20°C to 65°C and comes in a smaller form factor.
All platforms support NVIDIA’s AI software stack, including Isaac and Holoscan, enabling development and deployment of AI models on a unified environment. The systems are positioned for use cases where AI needs to run directly on machines with consistent performance, safety control, and low latency.
Click here for the original announcement.