Radxa VMARC-Q9075
Introduction
The Radxa VMARC-Q9075 is a high-performance AI computing module based on the Qualcomm Dragonwing IQ-9075 processor. It uses a 314-pin MXM 3.0 connector and measures 82 mm × 80 mm, making it suitable for space-constrained embedded applications that require high computing performance, AI inference, and multimedia processing capabilities.
The module integrates an 8-core Kryo Gen 6 CPU, a 4-core Cortex-R52 MCU, a Qualcomm Adreno 663 GPU, and a Hexagon Tensor Processor (HTP), delivering 100 Dense / 200 Sparse TOPS@INT8 of AI performance. It is equipped with 36GB LPDDR5 memory and 128GB UFS 3.1 storage and supports multiple high-resolution cameras, display outputs, 2.5GbE, USB 3.2 Gen 2, PCIe Gen4, and expansion signals including CAN, UART, SPI, I2C, and GPIO.
The Radxa VMARC-Q9075 Development Kit consists of the compute module and a matching IO board. It provides dual 2.5GbE ports, M.2 NVMe SSD, Wi-Fi / Bluetooth, and 4G / 5G expansion capabilities, as well as multiple USB, CAN, RS-232, RS-485, display, and camera interfaces. It can be used to quickly evaluate performance, validate interfaces, develop AI applications, and build product prototypes.
Product Views
- Radxa VMARC-Q9075
- Radxa VMARC-Q9075 IO Board
- Radxa VMARC-Q9075 Development Kit



Specifications
- Radxa VMARC-Q9075
- Radxa VMARC-Q9075 IO Board
- Radxa VMARC-Q9075 Development Kit
| Category | Specification |
|---|---|
| Processor | Qualcomm Dragonwing IQ-9075 |
| CPU | 8-core Kryo Gen 6, up to 2.36GHz |
| MCU | 4-core Cortex-R52, up to 1.85GHz |
| GPU | Qualcomm Adreno 663 - Supports secure general-purpose GPU computing with 1.2 TFLOPS FP32 performance - Supports Vulkan 1.2, OpenGL ES 3.2, and OpenCL 2.0 FP - Supports Adreno NN Direct |
| AI Performance | 100 Dense / 200 Sparse TOPS@INT8 - Hexagon Tensor Processor (HTP) with four HVX units and two HMX units - Supports mainstream deep-learning frameworks including TensorFlow, PyTorch, ONNX, Paddle, Caffe, and DarkNet |
| Video Codec | Adreno VPU 765 Video decoding: - Supports AV1, HEVC, H.264, H.265, VP9, and MPEG2 - Performance: 1× 8K@60fps / 2× 8K@30fps / 4× 4K@60fps / 8× 4K@30fps / 16× 1080p@60fps / 32× 1080p@30fps Video encoding: - Supports H.264, H.265, HEIF, and HEIC - Performance: 2× 4K@60fps / 4× 4K@30fps / 8× 1080p@60fps / 16× 1080p@30fps |
| Memory | LPDDR5 - Capacity: 36GB - Bus width: 96-bit - Maximum data rate: 6400 MT/s |
| Storage | 1× onboard 32MB SPI Flash 1× onboard 128GB UFS 3.1 (Gear4 x2) |
| Display | 1× MIPI DSI display signal 1× eDP display signal 1× HDMI TX signal (DP-to-HDMI signal) 1× DP TX signal |
| Camera | 1× 4-lane MIPI CSI camera signal 1× 2-lane MIPI CSI camera signal |
| Network | 2× 2.5GbE PHY |
| USB | 2× USB 3.2 Gen 2 signals (10Gbps) 4× USB 2.0 signals |
| PCIe | 2× PCIe Gen4 x2 |
| GPIO | 2× CAN 2× I2S 4× UART 2× SPI 5× I2C 14× GPIO |
| Power | 12V DC input |
| Connector | 314-pin MXM 3.0 connector |
| Dimensions | 82 mm × 80 mm |
| Category | Specification |
|---|---|
| Compatible Compute Module | VMARC-Q9075 |
| Storage | 1× M.2 M Key 2280 slot (NVMe SSD) |
| Network | 2× 2.5GbE RJ45 ports 1× M.2 E Key 2230 slot (Wi-Fi / Bluetooth module) 1× M.2 B Key 3042 slot (4G / 5G cellular module) 1× Nano SIM card slot |
| Display | 1× 4-lane MIPI DSI display connector (39-pin) 1× eDP connector 1× standard HDMI connector (Type-A) 1× DP connector (DisplayPort) |
| Camera | 1× 4-lane MIPI CSI camera connector (31-pin) 1× 2-lane MIPI CSI camera connector (15-pin) |
| USB | 4× USB 3.1 Type-A ports 1× USB 3.2 Gen 2 port 3× USB 2.0 ports 1× USB Type-C debug port |
| Other Interfaces | 1× power button 1× EDL button 1× status LED 1× fan connector 1× RTC battery connector 4× CAN bus interfaces 2× RS-232 interfaces 1× RS-485 interface 1× 40-pin GPIO header (UART / SPI / PWM / GPIO, etc.) |
| Power | DC5525 power jack or XT60 power connector (19–24V input) |
| Compute Module Connector | 314-pin MXM 3.0 connector |
| Dimensions | 120 mm × 120 mm |
| Category | Specification |
|---|---|
| Compute Module | Radxa VMARC-Q9075 |
| IO Board | Radxa VMARC-Q9075 IO Board |
| CPU | 8-core Kryo Gen 6, up to 2.36GHz |
| MCU | 4-core Cortex-R52, up to 1.85GHz |
| GPU | Qualcomm Adreno 663 - Supports secure general-purpose GPU computing with 1.2 TFLOPS FP32 performance - Supports Vulkan 1.2, OpenGL ES 3.2, and OpenCL 2.0 FP - Supports Adreno NN Direct |
| AI Performance | 100 Dense / 200 Sparse TOPS@INT8 - Hexagon Tensor Processor (HTP) with four HVX units and two HMX units - Supports mainstream deep-learning frameworks including TensorFlow, PyTorch, ONNX, Paddle, Caffe, and DarkNet |
| Video Codec | Adreno VPU 765 Video decoding: - Supports AV1, HEVC, H.264, H.265, VP9, and MPEG2 - Performance: 1× 8K@60fps / 2× 8K@30fps / 4× 4K@60fps / 8× 4K@30fps / 16× 1080p@60fps / 32× 1080p@30fps Video encoding: - Supports H.264, H.265, HEIF, and HEIC - Performance: 2× 4K@60fps / 4× 4K@30fps / 8× 1080p@60fps / 16× 1080p@30fps |
| Memory | LPDDR5 - Capacity: 36GB - Bus width: 96-bit - Maximum data rate: 6400 MT/s |
| Storage | 1× onboard 32MB SPI Flash 1× onboard 128GB UFS 3.1 (Gear4 x2) 1× M.2 M Key 2280 slot (NVMe SSD) |
| Display | 1× 4-lane MIPI DSI display connector (39-pin) 1× eDP connector 1× standard HDMI connector (Type-A) 1× DP connector |
| Camera | 1× 4-lane MIPI CSI camera connector (31-pin) 1× 2-lane MIPI CSI camera connector (15-pin) |
| Network | 2× 2.5GbE RJ45 ports 1× M.2 E Key 2230 slot (Wi-Fi / Bluetooth module) 1× M.2 B Key 3042 slot (4G / 5G cellular module) 1× Nano SIM card slot |
| USB | 4× USB 3.2 Gen 2 Type-A ports (10Gbps) 1× USB 3.2 Gen 2 port (10Gbps) 3× USB 2.0 ports 1× USB Type-C debug port |
| Other Interfaces | 1× power button 1× EDL button 1× status LED 1× fan connector 1× RTC battery connector 4× CAN bus interfaces 2× RS-232 interfaces 1× RS-485 interface 1× 40-pin GPIO header (UART / SPI / PWM / GPIO, etc.) |
| Power | DC5525 power jack or XT60 power connector (19–24V input) |
| Connector | 314-pin MXM 3.0 connector |
| Dimensions | 120 mm × 120 mm |
Block Diagram

Interface Description
- Radxa VMARC-Q9075
- Radxa VMARC-Q9075 IO Board
- Radxa VMARC-Q9075 Development Kit

| No. | Description | No. | Description | No. | Description |
|---|---|---|---|---|---|
| 1 | LPDDR5 | 2 | Qualcomm Dragonwing IQ-9075 | 3 | 314-pin MXM 3.0 connector |
| 4 | SPI NOR Flash | 5 | Qualcomm QCA8081 | 6 | Onboard UFS |

| No. | Description | No. | Description | No. | Description |
|---|---|---|---|---|---|
| 1 | Fan connector | 2 | 40-pin GPIO header | 3 | 4× CAN bus interfaces |
| 4 | 2× RS-232 interfaces | 5 | RS-485 interface | 6 | XT60 power connector |
| 7 | 314-pin MXM 3.0 connector | 8 | DP connector | 9 | 4× USB 3.2 Gen 2 Type-A ports |
| 10 | Nano SIM card slot | 11 | eDP connector | 12 | 1× 4-lane MIPI CSI |
| 13 | 1× 2-lane MIPI CSI | 14 | 1× M.2 M Key 2280 slot | 15 | USB Type-C debug port |
| 16 | Power button | 17 | USB 3.2 Gen 2 port | 18 | 3× USB 2.0 ports |
| 19 | RTC battery connector | 20 | EDL button | 21 | DC5525 power jack |
| 22 | HDMI connector | 23 | 2× 2.5GbE RJ45 ports | 24 | 1× 4-lane MIPI DSI |
| 25 | 1× M.2 B Key 3042 slot | 26 | 1× M.2 E Key 2230 slot |

| No. | Description | No. | Description | No. | Description |
|---|---|---|---|---|---|
| 1 | Fan connector | 2 | 40-pin GPIO header | 3 | 4× CAN bus interfaces |
| 4 | 2× RS-232 interfaces | 5 | RS-485 interface | 6 | XT60 power connector |
| 7 | 314-pin MXM 3.0 connector | 8 | DP connector | 9 | 4× USB 3.2 Gen 2 Type-A ports |
| 10 | Nano SIM card slot | 11 | eDP connector | 12 | 1× 4-lane MIPI CSI |
| 13 | 1× 2-lane MIPI CSI | 14 | 1× M.2 M Key 2280 slot | 15 | USB Type-C debug port |
| 16 | Power button | 17 | USB 3.2 Gen 2 port | 18 | 3× USB 2.0 ports |
| 19 | RTC battery connector | 20 | EDL button | 21 | DC5525 power jack |
| 22 | HDMI connector | 23 | 2× 2.5GbE RJ45 ports | 24 | 1× 4-lane MIPI DSI |
| 25 | 1× M.2 B Key 3042 slot | 26 | 1× M.2 E Key 2230 slot |
Applications
Edge AI Computing and Intelligent Gateways
With 100 Dense / 200 Sparse TOPS@INT8 of AI performance, dual 2.5GbE ports, and Wi-Fi, Bluetooth, and 4G / 5G expansion capabilities, the system can aggregate data, perform AI inference, and make real-time decisions at the edge. It is suitable for industrial gateways, edge servers, and multi-protocol data acquisition devices.
Machine Vision and Video Analytics
Support for 4-lane and 2-lane MIPI CSI camera inputs and multiple streams of 8K, 4K, and 1080p video encoding and decoding makes the system suitable for visual inspection, object recognition, behavior analysis, multi-channel video aggregation, and intelligent security applications.
Industrial Automation and Equipment Control
The 4-core Cortex-R52 MCU and interfaces including CAN, RS-232, RS-485, UART, SPI, I2C, and GPIO can connect sensors, actuators, and industrial control equipment. The system is suitable for automated production lines, machine condition monitoring, quality inspection, and equipment control systems.
Robotics and Autonomous Devices
The multi-core CPU, GPU, and high-performance AI accelerator can work together on environmental perception, object detection, and path-planning tasks. The extensive camera, network, and control interfaces simplify the expansion of sensors and motion-control modules, making the system suitable for mobile robots, service robots, inspection devices, and other autonomous systems.
Smart Displays and Interactive Terminals
MIPI DSI, eDP, HDMI, and DP display capabilities enable high-resolution, multi-display, and embedded display solutions for digital signage, smart retail, self-service terminals, information displays, and interactive devices.