en.Wedoany.com Reported - Innodisk showcased the comprehensive results of its collaboration with Qualcomm Technologies, Inc. at Computex 2026, expanding its edge platform product line based on the "AI on Dragonwing" computing series launched earlier this year. This product line covers four generations of Qualcomm Dragonwing processors, delivering scalable performance from 20 to 700 TOPS across a wide range of edge AI workloads.
The shift toward edge artificial intelligence in factory automation, smart cities, and defect detection has increased the demand for efficient inference. For battery-powered AMRs (Autonomous Mobile Robots) and AGVs (Automated Guided Vehicles), minimizing power consumption is critical to maintaining operational runtime and reducing thermal costs, making this high-performance, low-power architecture more suitable for next-generation deployments. With Dragonwing, Qualcomm Technologies aims to deliver industry-leading energy efficiency and multimodal computing for the industrial sector. This scalable processor portfolio spans the Qualcomm Dragonwing IQ8, IQ9, IQ10, and IQ-X series, with product lifecycle support extending to 2036 or later. Innodisk leverages this advantage to launch the "AI on Dragonwing" hardware product line, including modules, starter kits, and deployable systems, complemented by its IQ Studio software toolkit to streamline the process from evaluation to deployment through BSPs (Board Support Packages), benchmarking, and model optimization.
Shyam Krishnamurthy, Senior Vice President of Product Management at Qualcomm Technologies, stated that Dragonwing processors are designed to deliver industry-leading performance per watt and promise long-term supply, laying the foundation for large-scale physical AI deployments. He added that Innodisk's expertise in industrial edge integration makes it a key partner in bringing the AI on Dragonwing ecosystem to market.
The Computex product line covers four Dragonwing platforms tailored for different application scenarios. The Dragonwing IQ8 series powers the EXMP-Q801 COM-HPC Mini module, delivering up to 40 TOPS (dense) of computing power with wide temperature support, suitable for vision inference and sensor fusion. The Dragonwing IQ9 series scales up to 100 TOPS (dense) via a dual-NPU design, powering the EXMP-Q911 module, EXEC-Q911 starter kit, and APEX-A100 fanless edge AI system, addressing applications such as industrial inspection, Driver Monitoring Systems (DMS), and multi-stream analysis. The Dragonwing IQ10 series powers the EXMR-QA01, extending the product line to up to 350 TOPS (dense) with an 18-core Oryon CPU, supporting up to twenty 16-megapixel MIPI cameras, and integrating a Safety Island (SAIL) module to assist heavy-duty AMR navigation, humanoid robots, and VLM (Vision Language Model) and VLA (Vision Language Action) edge workloads. The Dragonwing IQ-X series stands out with native Windows compatibility; the EXDU-QX11 edge AI system, powered by the Dragonwing IQ-X processor, delivers up to 45 TOPS and supports dual operating systems—Windows 11 IoT Enterprise LTSC and Ubuntu 26 LTS—simplifying deployment in PLCs (Programmable Logic Controllers), HMIs (Human-Machine Interfaces), and edge controllers.
Two live demonstrations at the Innodisk booth validated deployment readiness. The first demonstration featured the EXMP-Q911 module powered by the Dragonwing IQ9 processor, simultaneously executing CNN defect detection, DMS, and VLM-driven PPE (Personal Protective Equipment) detection pipelines under a strict total system power consumption of 30W, achieving an ultra-low average inference latency of 4.4 milliseconds—a 2.6x speed advantage over traditional architectures. The second demonstration showcased vision-driven autonomous robot navigation on the EXFC-Q911 fanless system, which managed a high-throughput 8-channel GMSL2 camera pipeline and 3D depth sensors, delivering 360-degree spatial awareness and millisecond-response real-time obstacle avoidance at just 32W power consumption.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
