AIPC-631 — Compact Jetson AGX Thor AI for local deployment

    AIPC-631Compact Jetson AGX Thor AI for local deployment

    AIPC-631 is a compact NVIDIA Jetson AGX Thor AI system designed for high-performance local AI deployment where data and models must remain on site. Its dual 10GbE networking and small footprint suit edge inference, robotics, industrial automation, and other bandwidth-intensive AI applications.

    Highlights

    • Equipped with Jetson AGX THOR module
    • Unified memory architecture, supporting up to 128GB memory with a bandwidth of up to 273GB/s
    • Supports 14-core Arm® Neoverse®-V3AE 64-bit
    • 2 x 10GbE RJ45 High-speed network
    • 4 x USB3.2 Gen1 Type-A, 1 x USB3.2 Gen-1 Type-C
    • Supports 2TB large-capacity SSD storage
    • Local deployment, all data and models are processed locally without uploading to the cloud

    Certifications

    Description

    AIPC-631 is a compact NVIDIA Jetson AGX Thor AI system designed for high-performance local AI deployment where data and models must remain on site. Its dual 10GbE networking and small footprint suit edge inference, robotics, industrial automation, and other bandwidth-intensive AI applications.

    T4000 and T5000 configurations combine Arm Neoverse processors, NVIDIA Blackwell GPUs, up to 128GB unified memory, and up to 2070 TOPS. The system also provides NVMe storage, dual HDMI 2.1 outputs, five USB ports, Wi-Fi and Bluetooth expansion, Ubuntu support, and the NVIDIA AI software ecosystem.

    Specifications

    Computing
    Processor
    T4000: 12-core Arm® Neoverse®-V3AE 64-bit, T5000: 14-core Arm® Neoverse®-V3AE 64-bit
    Memory
    T4000: 64GB; T5000: 128GB
    Storage
    1 x M.2 M Key 2280 NVMe
    Graphics
    T4000: 1536-core NVIDIA Blackwell architecture GPU with 64 fifth-gen Tensor Cores Multi-Instance GPU (MIG) with 6 TPCs, T5000: 2560-core NVIDIA Blackwell architecture GPU with 96 fifth-gen Tensor Cores Multi-Instance GPU (MIG) with 10 TPCs
    AI Performance
    T4000: 1200 Sparse FP4 TFLOPs, 600 Sparse/300 Dense FP8 TFLOPs, T5000: 2070 Sparse FP4 TFLOPs, 1035 Sparse/500 Dense FP8 TFLOPs
    Watchdog
    255 levels of programmable seconds per minute
    I/O & Connectivity
    Display
    2 x HDMI 2.1 output, Up to 4096 x 2160 @60Hz
    USB
    4 x USB3.2 Gen1 Type-A, 1 x USB3.2 Gen1 Type-C
    LAN
    2 x RJ45, 10GbE
    Expansion
    1 x M.2 E Key 2230 (for WiFi5/6 &Bluetooth 5.3)
    Button
    1 x Power button, 1 x Recovery
    Mechanical & Power
    Power
    19V DC IN; 200W
    Dimensions
    130x125x75mm (5.1x4.9x3.0 in)
    Weight
    1.3kg (2.9lb)
    Environmental
    Operating Temperature
    -20~40℃ (-4~104 ℉)
    Relative Humidity
    10%~90% (non-condensing)
    Vibration
    With SSD solid-state drive: Operating state, 2Grs@5~500Hz,IEC 60068-2-64
    Shock
    With SSD solid-state drive: working state, 15G@11ms half sine wave
    Certifications & Software
    EMC
    CE FCC
    OS Support
    Ubuntu24.04 (Jetpack7.1)
    Toolchain
    CUDA, cuDNN, CuBLAS, NCCL, TensorRT
    Container Tool
    NVIDIA Container Toolkit
    AI Development Kit
    NVIDIA NIM, NVIDIA AI Blueprints (customizable workflows), NVIDIA AI Workbench (development environment manager)
    Frameworks
    Supports PyTorch, TensorFlow, MATLAB, etc.