Edgeble AI’s Neurable Compute Module 2, or Neu2 for shorts, is a system-on-module for computer vision applications based on the Rockchip RV1126 quad-core Cortex-A7 camera processor that follows the 96Boards SoM form factor.
I first found the Neu2 and Neu6 (Rockchip RK3588) in the release log for the Linux 6.3 kernel, but at the time I found there was not enough information about those. The specifications for the Neu6 are still wrong (e.g. “64-bit processor with 4x Cortex-A7 core”) at the time of writing, so I’ll check the Neu2 system-on-module and its industrial version – the Neu2K based on RK1126K – for which we have more details.
Edgeble Neu2 SoM specifications:
- SoC – Rockchip RV1126/RV1126K with
- CPU – Quad-core Arm Cortex-A7 @ 1.5GHz, RISC-V MCU @ 200MHz; (14nm SMIC process)
- GPU – 2D graphics engine
- NPU – 2 TOPS with INT8/INT16
- 4K H.264/H.265 video encoder up to 3840 x 2160 @ 30 fps + 720p @ 30 fps encoding
- 4K H.264/H.265 30fps video decoder up to 3840 x 2160 @ 30 fps encoding + 3840 x 2160 @ 30 fps decoding
- Process – SMIC 14nm low power process
- System Memory
- Neu2 – 2GB LPDDR4, Inline ECC
- Neu2K – 4GB LPDDR4, Inline ECC
- Storage – 16GB eMMC flash
- Camera I/F – 2x MIPI CSI-2 connector on-module
- Networking – Dual-band industrial WiFi and Bluetooth module (Fn-link 8223A-SR)
- Misc – Precision RTC
- 4x board-to-board connectors (X1-X4) as per 96Boards SoM specifications with
- Networking – Gigabit Ethernet (RGMII) with TSO network acceleration
- USB – 1x USB OTG, 2x USB Host
- Audio – 2x I2S (2ch), 1x I2S (1ch)
- Serial – Can Bus
- Low speed I/Os – 4x I2C, 4x UART, 4x SPI, 10x GPIO
- Dimensions – 85 x 50 x 6 mm
- Temperature Range
- Neu2 – Consumer grade: 0°C to +80°C
- Neu2K – Industrial grade: -40°C to +85°C
- Certifications – RoHS, Pre-certified WiFi/BT
Edgeble provides support for mainline Linux, Yocto Kirkston with pre-validated OTA system, and TensorFlow for the module. U-boot and the Linux kernel source code can also be found on GitHub. I could not find anything about TensorFlow per se, but there’s an AI Toolkit that does not seem open-source since you can request a demo, but still runs with 95% mainline Linux kernels.
The company also provides a carrier board for the module called “Edgeble Neu2 IO board”.
Neu2 IO carrier board specifications:
- Storage – MicroSD card slot
- Display – 1x MIPI DSI 40-Pin FPC connector
- Audio – 1x Speaker, 1x MIC, 1x MIC Array connectors
- Gigabit Ethernet RJ45 port
- mini PCIe slot for 4G module with SIM card slot
- USB – 2x USB Type-A ports, 1x micro USB OTG port
- Serial – CAN bus, RS485, and RS232 terminal blocks
- Expansion – 40-pin GPIO header
- Misc – RTC Battery slot, Reset key, Power key, Recovery key,
- Power Supply
- 5V via USB Type-C port
- On-board PoE (optional)
- Dimensions – 130 x 100 mm
The company can also provide a 10-inch 1080p LCD panel, a GC2093 + GC2053 Dual Image sensor, and/or an IMX415 4K Image sensor as shown in the photo above with what appears to be older revisions of the module and carrier board since the camera connectors are on the carrier board and not on the module.
Edgeble does not sell the module and development kits online, and instead, interested companies are asked to contact them. The product page has more details notably about the post/pre-processing capability of the ISP and VPU in the Rockchip RV1126 processor, and also a somewhat outdated (Linux 6.1) table showing the progress with Linux, U-boot, and EAI Toolkit (maybe this project?) upstreaming.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.