AIR-427A Series: Edge AI System Powered by NVIDIA IGX Thor
AIR-427A SeriesNVIDIA IGX Thor is designed to meet the stringent demands of physical AI applications, where deterministic performance, functional safety, and system reliability are critical requirements for deployment. Built for industrial and medical-grade environments, the platform combines high-performance AI compute, real-time processing, and safety capabilities in a unified architecture, forming a solid foundation for safety-critical edge intelligence.
- Powered by NVIDIA IGX T7000, delivering up to 2070 TFLOPS (FP4-Sparse) inference performance
- Powered by the NVIDIA Blackwell architecture iGPU and an optional NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition to deliver up to 5,581 TFLOPS (FP4-Sparse)
- Functional safety & system reliability
- Multiple I/O Expansion: 2x PCIe Gen 5, 1 x M.2 M Key (NVME) 1x M.2 B Key (5G/LTE), and 1x M.2 E Key (WiFi)
- Supports 2x 200 GbE with QSFP28
- Supports CAN, COM, GPIO with Header
- Enterprise validation
Frequently Asked Questions
The AIR-427A provides the following I/O:
- 1x USB 3.2 Gen2 Type C
- 4x USB 3.2 Gen2 Type A
- 1x VESA DisplayPort 1.4a
- 2x RJ-45 (1GbE each) and 2x QSFP28 (200GbE each)
- Audio line-out and mic-in
It also has expansion headers for CAN (1x CAN header, 20-pin), COM (4x COM header, 20-pin), and GPIO (1x GPIO header, 20-pin).
The unit accepts a supply voltage of 100~240V AC. It supports ATX/AT mode, with ATX as the default power type. The processor module has a TDP of 130W.
Operating temperature is 0 ~ 45 °C without a PCIe card installed, or 0 ~ 40 °C with a PCIe card installed. Operating humidity is rated at 95% @ 40 °C, non-condensing. Vibration rating is 3 Grms @ 5 ~ 500 Hz, random, 1 hr/axis. Note: operating temperature varies by the specific PCIe card used, so no universal guarantee applies, and devices with graphics cards installed are not suitable for high-vibration environments.
Certifications include CE/FCC Class B, CB, UL, CCC, and BSMI. The unit does not carry a RED certificate.
Storage: 1x M.2 M Key slot supporting NVMe (PCIe Gen5).
Memory: 128GB of 256-bit LPDDR5X system memory (273GB/s), plus 96GB of dedicated dGPU memory (1,792GB/s).
Expansion: 2x PCIe Gen5 slots (x8, x16), 1x M.2 B Key slot (5G/LTE), and 1x M.2 E Key slot (WiFi/BT). TPM 2.0 is available as an optional feature.
The AIR-427A ships with Ubuntu 24.04 LTS running IGX OS 2.0. Software support includes compatibility with the Advantech Edge AI SDK.
The AIR-427A measures 260.3 x 380.75 x 190 mm (W x D x H). It supports wall mounting. Weight is listed as TBD in the datasheet.
The system includes an M.2 B Key slot for 5G/LTE modules and an M.2 E Key slot for WiFi/BT modules, allowing wireless connectivity to be added via these expansion slots.
The AIR-427A is powered by the NVIDIA IGX T7000 module, delivering up to 2070 TFLOPS (FP4-Sparse) inference performance from its integrated Blackwell architecture GPU (2560 cores, 5th-generation Tensor Cores, 10 TPCs). With an optional NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition GPU added, performance can reach up to 5,581 TFLOPS (FP4-Sparse). Note that listed graphics card models have only been partially tested, so physical fitment should be verified.
The AIR-427A provides 2x RJ-45 ports at 1GbE each, and 2x QSFP28 ports supporting 200GbE each, using an NVIDIA ConnectX-7 PHY with BMC.
NVIDIA IGX Thor is designed to meet the stringent demands of physical AI applications, where deterministic performance, functional safety, and system reliability are critical requirements for deployment. Built for industrial and medical-grade environments, the platform combines high-performance AI compute, real-time processing, and safety capabilities in a unified architecture, forming a solid foundation for safety-critical edge intelligence.
| Product Available On | Apr 14, 2028 |
|---|---|
| Manufacturer | Advantech |
| Short Description | NVIDIA IGX Thor is designed to meet the stringent demands of physical AI applications, where deterministic performance, functional safety, and system reliability are critical requirements for deployment. Built for industrial and medical-grade environments, the platform combines high-performance AI compute, real-time processing, and safety capabilities in a unified architecture, forming a solid foundation for safety-critical edge intelligence. |