Rhino Boost Q86

Product Overview

Rhino Boost Q86 is a high-performance edge AI compute card from Chengdu APLUX Intelligence Technology Ltd, based on the Qualcomm Snapdragon QCS8625 chipset. It integrates a 60 TOPS (INT8) neural processing unit (NPU) and can stably run large language models (LLMs) and multimodal vision-language models (VLMs) with 0-14B parameters at 13W power consumption, delivering strong local AI compute for edge devices.

The card uses a Mini PCIe form factor and can be quickly integrated into mainstream platforms such as Raspberry Pi and RK development boards through adapter boards. It can also be embedded as a standalone module into various industry terminals.

Product Positioning

Rhino Boost Q86 supports two major application modes:

• Standalone Controller Mode: The compute card runs independently and provides AI services externally via Ethernet (OpenAI-compatible APIs)
• Ethernet Coprocessor Mode: Works with a host board (Raspberry Pi/RK platform). The host handles peripheral management while the compute card focuses on AI inference

Application Scenarios

• Smart Security: Smart NVR, face/abnormal behavior recognition, real-time multi-stream video analysis
• Industrial Field: SOP inspection, component defect detection, robotic arm collaborative control
• Consumer Electronics: Smart NAS, AI live-stream camera, home service robots
• Edge Computing: Edge gateways, local data processing, low-latency interactive terminals

Specifications

Item	Parameters
Platform	Qualcomm Snapdragon QCS8625 (4nm FinFET process)
AI Performance	15 TOPS (FP16) / 60 TOPS (INT8 Dense) / 120 TOPS (INT8 Sparse)
CPU	8-core heterogeneous architecture: 1x3.4GHz + 5x3.2GHz + 2x2.3GHz
GPU	Adreno 750, 5.2 TFLOPS-FP16, supports OpenGL ES 3.2 / Vulkan 1.3 / OpenCL 3.0
VPU	Decode: 4K@240fps / 8K@60fps Encode: 4K@120fps / 8K@30fps
Memory	12/16GB LPDDR5X
Storage	128GB UFS 3.1
Network Interface	2.5G Ethernet (RJ45)
Interfaces	Mini PCIe (PCIe Gen3 x2 RC), USB 3.1, USB 2.0, UART, I2C
Power Supply	Typical 12V
OS	Ubuntu 22.04
Dimensions	47.0 x 64.0 x 3.25 mm (Mini PCIe package)
Operating Temperature	-20C ~ +45C

Product Appearance

Front

Back

Note

The physical appearance is subject to the delivered product. Images are for reference only.

Mechanical Dimensions

Item	Parameters
Dimensions	47.0 x 64.0 x 3.25 mm (excluding thermal accessories)
Weight	Approx. 15g
Operating Temperature	-20C ~ +45C
Storage Temperature	-40C ~ +75C
Operating Humidity	10% ~ 90% RH (non-condensing)
Cooling Method	Passive cooling (heatsink or active fan recommended)

Interface Definition

Rhino Boost Q86 exposes all interfaces through the Mini PCIe edge connector, including power, data, control, and debug interfaces.

• Interface summary:

Interface Type	Specification and Function
Power Input	5V~18V wide-voltage input (VCC_IN, Pin 2/4/6/8)
Data Interface	1x2.5Gbps RJ45 Ethernet, 1xPCIe RC Gen3 (2 lane)
USB Interface	1xUSB 3.1 (SuperSpeed), 1xUSB 2.0 (HighSpeed)
Control Interface	1xUART (1.8V, debug serial), 1xI2C (1.8V)
Power On/Off	PWRKEY (active low), POWER_EN (power enable)
Emergency Download	FORCE_USB_BOOT (pull high to 1.8V to enter EDL mode)

• Key pin definitions:

Pin Name	Pin No.	I/O	Description
VCC_IN	2, 4, 6, 8	PI	Four input power pins, supports 5~18V wide-voltage supply. Add large capacitors and Zener diodes externally for surge protection.
1.8V-OUT	32	PO	1.8V power output, 100mA. Output is available only after power-on.
GND	1, 3, 5, 7, 35, 24, 36, 44	-	Ground
PCIE0_REFCLK_M	9	AO	PCIe reference clock -
PCIE0_REFCLK_P	11	AO	PCIe reference clock +
PCIE0_PWR_EN	10	DO	PCIe peripheral power enable, 1.8V level
PCIE0_RESET_N	12	DO	PCIe reset signal, 1.8V level
PCIE0_WAKE_N	38	DI	PCIe wake interrupt request signal, 1.8V level
PCIE0_CLK_REQ_N	22	DI	PCIe clock request signal, 1.8V level
PCIE0_RX0_M	27	AI	PCIe Lane0 RX -
PCIE0_RX0_P	29	AI	PCIe Lane0 RX +
PCIE0_TX0_M	31	AO	PCIe Lane0 TX -
PCIE0_TX0_P	33	AO	PCIe Lane0 TX +
PCIE0_RX1_M	17	AI	PCIe Lane1 RX -
PCIE0_RX1_P	19	AI	PCIe Lane1 RX +
PCIE0_TX1_M	21	AO	PCIe Lane1 TX -
PCIE0_TX1_P	23	AO	PCIe Lane1 TX +
DD-	37	AI/AO	RJ45 differential signal, supports 2.5G and backward compatible with 100M/1000M
DD+	39	AI/AO	RJ45 differential signal
DC-	41	AI/AO	RJ45 differential signal
DC+	43	AI/AO	RJ45 differential signal
DB-	45	AI/AO	RJ45 differential signal
DB+	47	AI/AO	RJ45 differential signal
DA-	49	AI/AO	RJ45 differential signal
DA+	51	AI/AO	RJ45 differential signal
DEBUG_UART_TX	13	DO	DEBUG_TX data output, 1.8V level. Default module debug port.
DEBUG_UART_RX	15	DI	DEBUG_RX data input, 1.8V level. Default module debug port.
I2C_SDA	14	OD	I2C signal, 1.8V level. Internally pulled up to 1.8V through 2.2K.
I2C_SCL	16	OD	I2C signal, 1.8V level. Internally pulled up to 1.8V through 2.2K.
USB0_HS_DP	26	AI/AO	USB 2.0 differential data positive. When high, module USB2.0 switches to pin 26/28.
USB0_HS_DM	28	AI/AO	USB 2.0 differential data negative. When high, module USB2.0 switches to pin 26/28.
VBUS	30	DI	USB power input for USB insertion detection. When high, module USB2.0 switches to pin 26/28.
USB1_HS_DM	40	AI/AO	USB 2.0 differential data negative. When low, module USB2.0 switches to pin 40/42.
USB1_HS_DP	42	AI/AO	USB 2.0 differential data positive. When low, module USB2.0 switches to pin 40/42.
USB1_SS_TX0_M	46	AO	USB 3.1 differential TX -
USB1_SS_TX0_P	48	AO	USB 3.1 differential TX +
USB1_SS_RX0_M	50	AI	USB 3.1 differential RX -
USB1_SS_RX0_P	52	AI	USB 3.1 differential RX +

Note

For carrier-board package dimensions and pin layout details, refer to the A8625MY1 Hardware Design Guide (contact FAE).

Accessories

Rhino Boost Q86 provides multiple adapter boards and debugging accessories to support access from different host platforms:

Accessory	Model	Compatible Host	Description
Debug baseboard	A8625MY1-SUB-REVA	Universal	Provides complete debug interfaces, including power, serial, RJ45, and USB
Raspberry Pi adapter board	CEK2601-Pi5A8625MY1	Raspberry Pi 5	Connected via PCIe FPC cable, jointly developed by Firefly Workshop and APLUX

Note

For detailed accessory information, see the Raspberry Pi 5 AI Acceleration Solution section.

System Platform

Supported operating systems:

• Ubuntu 22.04 + AidLux Desktop (for graphical development and validation)

Quick Access

---

Quick Start

ADB login and basic connectivity verification

Core Board & BSP

Source retrieval, environment setup, build, and firmware flashing

Peripheral & Interface Dev

Board introduction, installation, BOM reference, and drivers

Pi 5 AI Acceleration

Raspberry Pi 5 + Rhino Boost Q86 + CEK2601-Pi5A8625MY1 adapter board

AI App Development & Deployment

AidLite inference, model optimization, generative AI, and voice

Resources

System images, tools, and reference design materials

FAQ & Troubleshooting

Hardware, system, and AI development FAQs and troubleshooting