Specifications
SBC Model
OK1126B-S / OK1126BJ-S (development board)
SoM Model
FET1126Bx-S (40 x 40 mm)
CPU
Rockchip RV1126B: 4x Cortex-A53 up to 1.6 GHz
CPU Variant (BJ)
RV1126BJ: 4x Cortex-A53 up to 1.3 GHz
NPU
3 TOPS @ INT8, INT8/INT16 mixed precision
GPU
None — 2D RGA only
Memory
1 GB / 2 GB / 4 GB LPDDR4
Storage
8 / 16 / 32 / 64 GB eMMC
Operating System
Linux 6.1.141 with full BSP, RKNN toolchain
ISP (VICAP)
RAW8/10/12/14/16; up to 12 MP @ 20 fps input
AI ISP
Up to 8 MP @ 30 fps, maximum resolution 4096 x 4096
Camera Input
Up to 2x 4-lane MIPI CSI @ 2.5 Gbps per lane; MIPI CSI/LVDS/SubLVDS; DVP 8-16 bit
Video Encode
HEVC / H.264 / JPEG up to 12 MP @ 30 fps, parallel encode
Video Decode
H.265 and H.264 up to 3840 x 2160 @ 30 fps
Display
1x MIPI-DSI 4-lane @ 1.5 Gbps per lane, RGB 24-bit, max 1920 x 1080 @ 60 fps
Industrial Buses
Up to 2x CAN-FD (CAN 2.0 A/B), 8x UART, 5x I2C, 2x SPI, 27x PWM
Ethernet
1x RGMII Gigabit MAC plus built-in 100 Mbps PHY
ADC
Up to 24 single-ended channels, 13-bit, 2 MSPS
GPIO
Up to 118 GPIO; 40-pin Raspberry Pi-compatible header
Extremes
FET1126BJ-S SoM rated -40 °C to +85 °C; RV1126B SoM -20 °C to +85 °C
Power
DC 5 V
Board Size
120 x 75 mm (SBC), 40 x 40 mm (SoM)
Overview
The Forlinx OK1126Bx-S pairs a 120 x 75 mm carrier board with a 40 x 40 mm FET1126Bx-S system-on-module built on Rockchip's RV1126B and RV1126BJ. The chip carries four Cortex-A53 cores up to 1.6 GHz and a 3 TOPS INT8 NPU, deliberately omitting a 3D GPU to spend the silicon budget on the vision path instead: a VICAP image signal processor taking RAW8 to RAW16 at up to 12 MP, an AI-ISP stage that uses the NPU for spatial noise reduction at up to 8 MP, and hardware H.265/H.264/JPEG encode up to 12 MP at 30 fps.
That makes it a vision-first board rather than a general-purpose SBC. Two 4-lane MIPI CSI receivers run at 2.5 Gbps per lane and support MIPI CSI, LVDS or SubLVDS PHY configurations, with an additional 8-16 bit DVP path — so stereo camera pairs, or a camera plus a second sensor type, attach natively. Models run through Rockchip's RKNN toolchain with TensorFlow, PyTorch, Caffe and MXNet conversion, and typical targets are face detection, safety-helmet recognition, fire and smoke alarms and intrusion detection running entirely locally.
For industrial integration the SoM exposes 237 pins over castellated holes and LGA, bringing out CAN-FD, up to eight UARTs, I2C, SPI, PWM, a 13-bit 24-channel ADC, RGMII Gigabit Ethernet and a 100 Mbps PHY. GPIO is laid out to the Raspberry Pi 40-pin standard so existing HATs and wiring can be reused. The BJ variant of the module is rated -40 to +85 °C for 24/7 outdoor and industrial duty.
Key Benefits
3 TOPS NPU with a dedicated AI-ISP delivers usable image quality in low light without extra illumination hardware. Dual 4-lane MIPI CSI plus CAN-FD and 8 UARTs suit multi-camera machine vision and vehicle or machinery interfaces on one board. -40 to +85 °C on the BJ module and a Raspberry Pi-compatible pinout cut both qualification risk and integration time.
Applications
Embedded vision and smart security cameras, industrial AI inspection, AIoT gateways, machine-vision terminals with CAN-connected equipment, edge inference in outdoor cabinets, and OEM products built on the castellated SoM.
Request a Quote — FORLINX OK1126BX-S — ROCKCHIP RV1126B EDGE AI VISION SBC
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →