Specifications
SoC
Qualcomm Dragonwing QCS8550, 4 nm process
CPU
8-core 64-bit Kryo — 1x GoldPlus @ up to 3.2 GHz, 4x Gold @ up to 2.8 GHz, 3x Silver @ up to 2.0 GHz
AI Accelerator
Optional Hexagon processor — up to 48 dense INT8 TOPS (dual eNPU V3 engines, four HVX units, HMX); a non-NPU configuration is also available
GPU
Qualcomm Adreno 740 with hardware-accelerated ray tracing, OpenGL ES 3.2, Vulkan 1.2, OpenCL 3.0 Full Profile and Adreno NN Direct
Memory
Up to 16 GB 64-bit LPDDR5X
Storage
128 GB UFS 3.1 plus 32 MB SPI NOR flash
Removable Storage
microSD via SD 3.0 interface; NVMe connectable through PCIe
Video Decode
H.265, H.264, VP9 and AV1 up to 4K240 or 8K60
Video Encode
H.264 and H.265 up to 4K120 or 8K30
Display
Two 4-lane MIPI-DSI interfaces plus DisplayPort 1.4 over USB-C; up to two 4K60 displays with Multi-Stream Transport
Camera
Qualcomm Spectra ISP with up to six 4-lane MIPI-CSI interfaces; 108 MP @ 30 fps capture with zero shutter lag, 200 MP non-ZSL snapshots
Audio
SoundWire and digital microphone interfaces driven by a Qualcomm Hexagon V73M dual-cluster, four-thread DSP
High-Speed I/O
1x PCIe Gen 4 x2, 1x PCIe Gen 3 x2, USB 3.1 Gen 2 up to 10 Gbps, DisplayPort 1.4 Alt Mode, eUSB 2.0
Low-Speed I/O
UART, I2C, I3C and SPI via QUP; I2S, PWM and GPIO per block diagram
Form Factor
36 × 32 mm, 516-ball BGA package
Software
Radxa OS, Ubuntu, Debian, Armbian, fnOS and Windows planned; Qualcomm AI Hub for on-device model deployment
Development
Radxa provides a dedicated development carrier board for the module
Target Applications
Autonomous mobile robots, intelligent cameras, edge AI gateways and multimodal AI devices
Overview
The Radxa rCore-Q8550 is a coin-sized system-on-module that packs Qualcomm's Dragonwing QCS8550 platform into a 36 × 32 mm, 516-ball BGA package. It pairs an eight-core Kryo CPU running up to 3.2 GHz with an Adreno 740 GPU and, in its AI configuration, a Hexagon processor rated at up to 48 dense INT8 TOPS through dual eNPU V3 engines, four HVX units and HMX matrix acceleration.
Designed for robotics and machine vision, the module carries up to 16 GB of 64-bit LPDDR5X, 128 GB UFS 3.1 storage, PCIe Gen 4 x2 and Gen 3 x2 links, and a Spectra ISP able to drive up to six 4-lane MIPI-CSI cameras at 108 MP with zero shutter lag. Radxa ships a companion development carrier board and lists Radxa OS, Ubuntu, Debian, Armbian, fnOS and Windows support, with Qualcomm AI Hub available for on-device model deployment.
Key Benefits
48 dense INT8 TOPS in a 36 × 32 mm footprint for robot perception and multi-camera vision. Flagship connectivity: PCIe Gen 4 x2, USB 3.1 Gen 2 and UFS 3.1 storage. Multi-camera by design: up to six 4-lane MIPI-CSI inputs with 108 MP ZSL capture. Solder-down deployment: a 516-ball BGA module that scales from prototype carrier boards to volume production. Broad OS support with a Qualcomm AI Hub path for model deployment.
Applications
Autonomous mobile robots and AGVs, intelligent multi-camera security systems, machine-vision inspection, edge AI gateways, multimodal AI kiosks and devices, and drone or UAV perception stacks.
Request a Quote — RADXA RCORE-Q8550 — 48-TOPS QUALCOMM DRAGONWING EDGE AI MODULE
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →