Radxa Fogwise AIRbox Q900 Industrial Edge AI Box with 200 TOPS Power

Fogwise AIRbox Q900 Industrial Edge AI Box

With a Qualcomm Dragonwing IQ-9075 SoC that delivers up to 200 TOPS of AI performance in sparse mode, alongside offering 36GB LPDDR5 RAM, 128GB UFS 3.1 storage, and an M.2 M Key slot for PCIe Gen4 SSD expansion, Radxa has unveiled the Fogwise AIRbox Q900, a diminutive but durable industrial-grade Edge AI box that happens to be an

upgrade to the Fogwise Airbox. Targeting real-time inference in robotics, smart cities, manufacturing, and research, the small unit provides high-performance computing with industrial dependability.

With an Adreno 663 GPU that can generate up to 1.2 TFLOPS, the AIRbox Q900 boasts an octa-core Kryo Gen 6 CPU based on the Arm Cortex-A78C. Hardware-accelerated video codecs that facilitate simultaneous multi-stream decoding and encoding have been incorporated by an Adreno VPU 765. Radxa asserts the aforementioned combination grants the Q900 to perform sophisticated multimodal, voice, and vision workloads locally, including 7B-parameter language model support at up to 22 tokens per second inference speed.

Fogwise AIRbox Q900 Industrial Edge AI Box Internal Structure 

Radxa Fogwise AIRbox Q900 Specifications:

• SoC: Qualcomm DragonWing IQ-9075
  • CPU:
    • Octa-Core Kryo Gen 6 (Cortex-A78C-based) application cores @ up to 2.36 GHz
    • Quad-Core Cortex-R52 real-time cores @ up to 1.85GHz
  • GPU: Adreno 663 GPU delivering up to 1.2 TFLOPS FP32 with secure GPU compute; supports Vulkan 1.2, OpenGL ES 3.2, OpenCL 2.0 FP, Adreno NN Direct
  • NPU:
    • 200 TOPS @ INT8 (Sparse AI Computing Power)
    • Hexagon Tensor Processor (HTP) with quad HVX and dual HMX
  • VPU: Adreno VPU 765
    • Video Decode:
      • AV1 / HEVC / H.265 / H.264 / VP9 / MPEG-2
      • 1x 8Kp60 / 2x 8Kp30 / 4x 4Kp60 / 8x 4Kp30 / 16x 1080p60 / 32x 1080p30
    • Video Encode:
      • H.264 / H.265 / HEIF / HEIC
      • 2x 4Kp60 / 4x 4Kp30 / 8x 1080p60 / 16x 1080p30
    • Concurrent 2x 4Kp60 encode + 2x 4Kp60 decode
  • Display (Video Output): 1x Standard HDMI 2.0 port up to 4Kp60
• System Memory: 36GB 96-bit LPDDR5 @ 6400MT/s (with ECC support)
• Storage:
  • 1x 32MB SPI Flash (Onboard)
  • 1x UFS & eMMC connector
  • 1x Onboard 128GB UFS 3.1 (Gear4 x2)
  • 1x M.2 M Key Slot (PCIe Gen4 x4)
• Networking:
  • Wired: 2 x 2.5GbE RJ45 LAN ports with TSN
  • Wireless:
    • Wi-Fi 6 / BT 5.4 via mini PCIe socket
    • 4G / 5G cellular network connectivity via mini PCIe socket + Nano SIM Card slot
    • Supports up to 3 external antennas
• USB:
  • 1x USB 3.1 Gen2 Type-A HOST port
  • 1x USB 3.1 Gen2 Type-A OTG port
  • 1x USB Type-C Serial Console port
• Misc:
  • Power and EDL buttons
  • PWM fan connector (PWM fan built-in)
  • RTC battery connector
• Power Supply: 12V / 5.4A DC via 5.5 x 2.5 mm DC jack
• Dimensions: 104 x 84 x 45 mm (Durable aluminium alloy enclosure)
• Temperature Range: 0°C ~ 60°C

Radxa Fogwise AIRbox Q900 Specifications

TensorFlow, PyTorch, ONNX, and other AI frameworks are supported by Radxa. It has the ability to run Stable Diffusion, LLaMa, and GPT locally. CasaOS, a container-based management platform with a graphical user interface and app store, is pre-installed on the Fogwise AIRbox Q900, which runs Ubuntu and Yocto Linux.

The Fogwise AIRbox Q900 is available for purchase through Arace Tech at $599 for the 36GB LPDDR5 RAM + 128GB UFS 3.1 configuration. More details can be found on the product page and official press release. 

All images courtesy of Radxa.