Radxa rCore-Q6490 Is a Compact, Energy-Efficient SoM with 12 TOPS AI Performance

rCore-Q6490 Energy-Efficient Edge AI SoM

Radxa has introduced the rCore-Q6490, an Energy-Efficient Edge AI SoM built around the Qualcomm Dragonwing QCS6490 platform. The compact system-on-module combines an 8-core Kryo 670 CPU, Adreno 643 GPU, and Hexagon 770 AI Engine in a 42 × 36 mm form factor. Its 12 Dense TOPS AI performance makes it suitable for intelligent embedded systems that need strong on-device computing without relying entirely on cloud processing.

The module targets edge AI, robotics, machine vision, smart displays, industrial control, and embedded applications. It supports up to 16GB of LPDDR5 memory and provides PCIe Gen3, USB 3.1 Gen 1, UFS, eMMC, and multiple camera interfaces. Radxa also offers commercial and industrial temperature options, while the company lists product longevity through July 2036.

rCore-Q6490 Energy-Efficient Edge AI SoM Top and Bottom

Radxa rCore-Q6490 SoM Specifications:

• SoC: Qualcomm Dragonwing QCS6490
  • CPU: 8-core 64-bit Kryo 670 CPU with 1x Kryo Gold Prime @ 2.7GHz, 3x Kryo Gold @ 2.4GHz, and 4x Kryo Silver @ 1.9GHz
  • Process: 6nm
  • GPU: Qualcomm Adreno 643 with OpenGL ES 3.2, OpenCL 2.0, Vulkan 1.x, and DirectX 12
  • AI Engine: Qualcomm Hexagon 770 with up to 12 Dense TOPS, dual HVX, Hexagon CP 2.0, and Tensor Accelerator
• Memory and Storage:
  • Memory: Up to 16GB LPDDR5, 32-bit dual-channel
  • eMMC: On-module eMMC 5.1
  • microSD: SDIO 3.0
  • UFS: UFS 3.1
  • NVMe: Supported through PCIe Gen3 x2
• Video and Display:
  • Video Decode: Up to 4K60, 2x 4K30, or 4x 1080p60 for H.264, H.265, and VP9
  • Video Encode: Up to 4K30 or 4x 1080p30 for H.264 and H.265
  • HDR: HDR10 and HDR10+ playback
  • Display: 1x 4-lane eDP/DisplayPort and 1x 4-lane MIPI DSI
  • USB-C: 5Gbps interface with DisplayPort Alt Mode
• Camera and AI Vision:
  • ISP: Qualcomm Spectra 570L ISP
  • Camera Interfaces: 5x 4-lane MIPI CSI
  • Camera Resolution: Up to 64MP at 30fps with zero-shutter-lag capture
  • AI Applications: Designed for robotics, machine vision, smart terminals, and edge AI systems
• Audio: SoundWire interface through the Qualcomm Low-Power Audio Subsystem (LPASS)
• Expansion and Connectivity:
  • PCIe: 1x PCIe Gen3 x2 for NVMe SSD expansion and 1x PCIe Gen3 x1 for wireless expansion
  • USB: 1x USB 3.1 Gen 1 (5Gbps) and 1x USB 2.0
  • Serial Interfaces: I2C, I3C, SPI, and UART
• Mechanical:
  • Package: 540-ball BGA
  • Dimensions: 42 × 36mm
  • Operating Temperature: 0°C to 60°C, with commercial and industrial variants available

Interfaces

The rCore-Q6490 supports Radxa OS, Ubuntu, Deepin, and Yocto, while the QCS6490 platform also supports Linux distributions such as Fedora, Armbian, and Arch, along with Windows 11 IoT Enterprise. Developers can use Qualcomm AI Hub to access pre-optimized AI models and simplify deployment on the module.

Block Diagram

Radxa currently lists the rCore-Q6490 as “Coming Soon,” so the company has not announced an official retail price yet. The module will sell through Radxa’s approved partner network rather than directly through a standard public checkout. Users can check the official product page for availability and regional purchasing options.

Images used courtesy of RADXA