Impulse Embedded, a prominent UK supplier of industrial computing solutions, has introduced new turnkey edge AI systems based on NVIDIA’s Jetson AGX Thor. This next-generation platform is designed for robotics and autonomous machines, delivering up to 2,070 FP4 TFLOPS of AI performance. Equipped with a 14-core Arm Neoverse-V3AE CPU, up to 128GB LPDDR5X, and 100GbE-class connectivity, the Jetson AGX Thor offers data-centre-class capability at the edge. This allows for real-time generative AI, advanced sensor fusion, and high-performance robotics.
Compared to the Jetson AGX Orin, which offers up to 275 INT8 TOPS, the Thor platform provides over 7.5 times more AI compute power and is 3.5 times more energy-efficient. This enhancement enables new application classes, such as autonomous vehicles and large-scale vision analytics with low latency. Thor is integral to NVIDIA’s roadmap for platforms like Isaac, Metropolis, Holoscan, and GR00T, ensuring software compatibility and accelerating innovation.
Chris Durose, head of technology and R&D at Impulse Embedded, stated, “With Jetson AGX Thor, our customers can deploy multimodal, foundation-model-driven robotics and real-time generative AI directly on device without the latency or data-sovereignty challenges of the cloud. We’ve paired Thor’s performance with field-proven, industrial platforms so engineering teams can prototype and deploy faster.”
The Advantech MIC-743-AT is a standout product, featuring the Jetson AGX Thor T5000. This edge AI inference computer offers up to 2,070 FP4 TFLOPS and a 14-core Arm Neoverse-V3AE CPU. It combines Thor’s GPU acceleration with high-throughput connectivity to facilitate seamless data transfer between sensors, cameras, and systems, allowing complex models to operate with real-time communication.
The MIC-743-AT’s fanless design ensures silent, maintenance-free operation, and its wide operating temperature range makes it suitable for industrial and mission-critical environments. Its modular I/O architecture allows for expansion, accommodating additional networking and sensor interfaces as projects develop. This compact, industrial-grade platform is ideal for demanding edge deployments in manufacturing, smart cities, and healthcare AI.
Key features include up to 128GB LPDDR5X memory for intensive models and high-bandwidth sensor data, QSFP28 for multi-25GbE integration — perfect for multi-camera arrays, LiDAR, and high-speed networks. The system supports applications such as AI-driven inspection, predictive maintenance, multi-sensor fusion, healthcare, robotics, advanced image recognition, and on-premises generative AI.
Impulse provides comprehensive support beyond hardware delivery, assisting customers in integrating Jetson platforms into production-grade solutions. From custom configuration to thermal and lifecycle management, Impulse ensures a smooth transition from prototype to deployment.
