The AIBOX-9075 is powered by Qualcomm Dragonwing IQ9 series industrial-grade SoC: the IQ-9075. It integrates an 8-core all-big-core Kryo Gen6 CPU with 230 KDMIPS of general-purpose computing power, and a dual Hexagon NPU delivering up to 200 TOPS of peak INT8 performance, perfectly suited for high-load edge AI and concurrent multi-tasking. Built with a secure subsystem and four real-time cores, and leveraging the Qualcomm Linux software stack, it supports Ubuntu/Yocto operating systems and mainstream AI inference frameworks. Featuring an industrial wide-temperature design, it comes standard with GMSL2, CAN-FD, RS485, and optically isolated I/O, making it ideal for edge computing, industrial control, intelligent surveillance, and robotics scenarios.
Octa-Core Heterogeneous Processor: IQ-9075
The entire unit is equipped with the Qualcomm IQ-9075 processor, manufactured on a 5nm LPE process. It integrates an octa-core all-big-core Kryo Gen6 CPU with a maximum clock speed of 2.36GHz and a built-in dual Hexagon NPU. Supporting hardware collaborative acceleration across CPU/GPU/NPU, it enables the private deployment of on-device large models and generative AI. With large model inference performance reaching 12 TPS @ 13B parameters, it efficiently handles various complex artificial intelligence computations and large model inferences.
High-Performance 8K Codec VPU
The unit features a high-performance VPU (Video Processing Unit) that supports concurrent multi-channel encoding and decoding at 8K@60fps, 4×4K@60fps, and 32×1080P@30fps. It can simultaneously encode and decode 2×4K@60fps + 2×4K@60fps streams.
Built-in Adreno 663 GPU
The built-in Adreno 663 GPU supports 4K@60fps output and is compatible with mainstream graphics APIs including OpenGL ES, Vulkan, OpenCL, and OpenVX.
36GB LPDDR5 High-Bandwidth Memory
The board is equipped with 36GB of LPDDR5 (ECC) high-speed memory and 128GB of UFS 2.2 flash storage. It also supports PCIe 4.0×4 M.2 NVMe SSD expansion, fully ensuring high-speed read/write and stable operation for on-device large model loading, HD video processing, and massive data handling.
All-Scenario Multi-Standard Network Connectivity
Equipped with dual 2.5G Gigabit Ethernet RJ45 interfaces, it supports TSN (Time-Sensitive Networking) industrial real-time networks. It can expand Wi-Fi 6 and Bluetooth 5.2 via M.2 slots, and 4G/5G wireless networks via Mini PCIe interfaces. With low-latency and highly reliable transmission characteristics, it fully meets the application needs of industrial automation, real-time motion control, and high-speed data backhaul scenarios.
Rich Industrial Expansion Interfaces
The entire unit is configured with 2 CAN-FD ports, 2 RS485 ports, and 12 optically isolated DI/DOs, comprehensively adapting to high-reliability link connections, real-time data interaction, and multi-device collaborative control scenarios on the industrial floor.
It also features 8 GMSL2 high-speed vision interfaces that support the synchronous connection of HD cameras with transmission distances of up to 20 meters. With strong anti-interference capabilities, it stably adapts to complex industrial electromagnetic environments, ensuring millisecond-level synchronization of multi-channel images and seamless integration with multi-camera scenarios such as machine vision, intelligent security, autonomous driving, and industrial inspection.
All-Aluminum Alloy Casing, Efficient Heat Dissipation
The unit features an all-aluminum alloy unibody construction with a built-in fan for active, highly efficient heat dissipation. It supports ultra-wide temperature stable operation from -40°C to 85°C and enables 7×24 hours of uninterrupted continuous operation. Relying on stringent industrial-grade reliability design, it perfectly adapts to extreme high/low-temperature scenarios such as vehicle-mounted applications, outdoor exposure, and industrial plant environments.
Comprehensive Full-Stack Software Ecosystem Support
Leveraging the Qualcomm Linux software stack, it natively adapts to Ubuntu/Yocto systems and is compatible with mainstream AI inference frameworks such as TensorFlow, PyTorch, and ONNX. It supports the private on-device deployment of large language models like Deepseek-R1, Gemma, Llama, Qwen, and Phi, as well as image generation models like Stable Diffusion. It can connect to model marketplaces and massive model repositories, allowing for rapid model conversion, tuning, and validation, streamlining the full-process deployment of on-device AI.
Extensive Application Areas
Widely applicable across industries including robotics, private deployment of large models, edge computing, intelligent surveillance, industrial automation, and smart cities.
