Radxa is back with another powerful compute module, and this time it’s called the Radxa CM4. Think of it as a next-gen platform designed for Edge AI, multimedia, and high-performance embedded systems — all packed into a compact 55×40 mm module.
Although it’s not yet available for purchase, Radxa has already dropped full documentation and hardware specs, giving developers plenty of time to explore what’s coming next.
At the heart of the CM4 is the Rockchip RK3576(J) SoC, combining:
- Octa-core CPU: 4× Cortex-A72 + 4× Cortex-A53, up to 2.2 GHz
- GPU: ARM Mali-G52 MC3 with OpenGL ES 3.2, Vulkan 1.2, and OpenCL 2.1 support
- NPU: Up to 6 TOPS (INT8), compatible with TensorFlow, PyTorch, ONNX, and Caffe
This means the CM4 isn’t just about raw performance — it’s built for AI inference, computer vision, and other ML workloads right at the edge.
Here’s where things get exciting: the multimedia engine supports modern codecs like AV1, VP9, and H.265. You get:
- 8K @ 30 fps decoding
- 4K @ 120 fps decoding
- 4K @ 60 fps encoding
Add to that dual-display support via HDMI, DisplayPort, eDP, or MIPI DSI, and you’ve got yourself a pretty capable media box or edge AI visualization device.
- Up to 16 GB LPDDR4X RAM
- Up to 256 GB eMMC storage
- Expansion via SPI flash, UFS 2.0, or SDMMC
And for the tinkerers out there:
- 2× PCIe Gen2 lanes
- 2× SATA 3.1 for storage
- USB 3.2 (Gen1 & Gen2)
- M.2 M-Key slot for NVMe SSDs
 |
| M.2 M Key Connector Provides PCIe 2.1 1-lane for Effortless Expansion,Supports |
Basically, plenty of options to turn this into a mini-server, AI accelerator, or media powerhouse.
The Radxa CM4 doesn’t hold back on networking:
- 1× Gigabit Ethernet PHY (PoE supported)
- WiFi 6
- Bluetooth 5.4 with external antenna support
The CM4 integrates neatly into Radxa’s broader ecosystem, complete with a dedicated IO board exposing:
- Display & camera connectors
- Gigabit Ethernet
- Multiple USB ports (2.0, 3.0, and Type-C OTG)
- Storage expansion (UFS, MicroSD, NVMe)
- GPIO, RTC, fan connector, and even a 3.5 mm audio jack
Best part? It’s also mechanically and electrically compatible with Raspberry Pi CM4 carrier boards, including the official Pi IO Board and third-party options from WaveShare. That means if you’ve already been working with Raspberry Pi CM4 designs, you won’t be starting from scratch here.
Out of the box, Radxa promises support for:
- Debian
- Linux
- Buildroot
- Android
With long-term availability confirmed until at least September 2035, developers and businesses can rely on stability for long-term projects.
- CPU: 4× Cortex-A72 + 4× Cortex-A53, up to 2.2 GHz
- GPU: Mali-G52 MC3
- NPU: Up to 6 TOPS (INT8)
- RAM: Up to 16 GB LPDDR4X
- Storage: Up to 256 GB eMMC, UFS 2.0, SDMMC
- Networking: Gigabit Ethernet, WiFi 6, Bluetooth 5.4
- Video: 8K @ 30 fps decode, 4K @ 120 fps decode, 4K @ 60 fps encode
- Expansion: PCIe Gen2, SATA 3.1, USB 3.2, M.2 NVMe
- Camera: 1× 4-lane MIPI CSI or 2× 2-lane MIPI CSI
- Size: 55 × 40 mm
The Radxa CM4 isn’t on sale yet, but full product documentation is already live on the official Radxa product page. Pricing hasn’t been revealed, but with its feature set, it’s clearly aimed at developers who need powerful compute modules for AI, multimedia, and embedded solutions.
And here’s a big plus: Radxa has committed to long-term availability until 2035, making it ideal for industrial projects.
The Radxa CM4 feels like a serious competitor to Raspberry Pi CM4 and other compute modules in the same class. With 8K video support, AI acceleration, WiFi 6, and PCIe expansion, it has all the makings of a high-performance embedded solution.
If you’ve been waiting for something more powerful than the Pi CM4 for edge AI projects, multimedia applications, or industrial use, this one is definitely worth keeping an eye on.
0 Comments