Hardware Overview of Rubik Pi 3

Follow Us On

‎

Rubik Pi 3

Table of Contents [Sticky]

RUBIK Pi 3 is a compact QCS6490-based SBC designed for edge AI, smart cameras, and robotics. The most important hardware facts for evaluation are simple: the board exposes camera connectors, HDMI, Ethernet, multiple USB ports, a 40-pin header, and M.2 expansion; internally it uses a QCS6490-based compute module; and the official vendor resource set now includes datasheets, 3D resources, and hardware schematic.

This page is part of Rubik Pi 3. It answers common searches such as “What is the QCS6490?”, “What are the main features of the QCS6490 chipset?”, “RUBIK Pi 3 specs”, and “RUBIK Pi 3 schematic”.

Board-level overview

RUBIK Pi 3 is best understood as a product-development board rather than just a hobby SBC. It combines:

a Qualcomm Dragonwing QCS6490-based compute platform,
a carrier-board-style set of interfaces for cameras, networking, storage, and displays,
and maker-friendly expansion points for fast prototyping.

The current vendor datasheet identifies the following major board elements:

USB-C with power delivery and data
HDMI output
Gigabit Ethernet
2 x USB 3.0 Type-A plus 1 x USB 2.0 Type-A
2 camera connectors
40-pin connector
M.2 Key M
debug Micro USB
RTC connector
fan connector
3.5 mm audio jack

What is the QCS6490?

The Qualcomm Dragonwing QCS6490 is an embedded application processor for industrial and commercial IoT systems. It combines general-purpose CPU cores, graphics, AI acceleration, image signal processing, video acceleration, and expansion I/O in a single SoC. For a board like RUBIK Pi 3, that means one chip family can cover camera ingestion, AI inference, display, networking, and local application logic without needing separate accelerator boards.

QCS6490 is a high-end embedded SoC for smart cameras, robotics, AI boxes, industrial handhelds, and edge multimedia devices where low-power on-device AI and rich I/O matter as much as raw CPU speed.

Main features of the QCS6490 chipset

Compute and AI

The QCS6490 family combines an octa-core Kryo 670 CPU cluster with Qualcomm AI Engine acceleration. In Qualcomm's current product brief, the processor is positioned with up to 12 dense TOPS, which is enough to support real-time edge inference tasks such as classification, detection, segmentation, tracking, and multimodal preprocessing when paired with an efficient pipeline design.

Camera and vision

The processor family is designed for camera-heavy devices. Qualcomm's current materials describe support for up to five concurrent cameras and a triple-ISP architecture. This is one of the reasons QCS6490 appears in robotics, smart camera, surveillance, and AI-box discussions.

Multimedia and display

The same silicon also includes a GPU, a dedicated video processing path, and display support suitable for local visualization, preview, overlay composition, and encode / decode workflows. For RUBIK Pi 3 users, that translates into practical support for GStreamer-based video applications and AI-assisted multimedia products.

I/O and expansion

QCS6490 is relevant for product development because it is not just an AI chip. It includes PCIe, USB, serial buses, GPIO, and storage options needed to connect sensors, SSDs, peripherals, and carrier-board logic.

QCS6490 technical specifications

QCS6490 technical summary relevant to RUBIK Pi 3 users
Category	Summary
CPU	8-core Kryo 670 CPU, up to 2.7 GHz
AI	Qualcomm AI Engine, up to 12 dense TOPS
GPU	Adreno 643
Camera	Triple ISP, support for multiple concurrent cameras
Video	4K-class encode / decode capabilities on the platform
Memory support	LPDDR4x / LPDDR5 support at SoC level
Expansion	PCIe Gen 3, USB, UART, SPI, I2C, I3C, GPIO
Supported OS	Linux, Ubuntu, Android

RUBIK Pi 3 board specifications

These are the specs of the RUBIK Pi 3:

RUBIK Pi 3 developer-board configuration
Item	Board-level value
Memory	8 GB LPDDR4x
Storage	128 GB UFS 2.2
Camera inputs	2 x 4-lane MIPI CSI D-PHY connectors
Networking	Gigabit Ethernet, Wi-Fi 5, Bluetooth 5.2
USB	1 x USB-C, 2 x USB 3.0 Type-A, 1 x USB 2.0 Type-A
Expansion	40-pin header, M.2 Key M connector
Audio	3.5 mm jack
Board size	100 mm x 75 mm

Camera hardware

Camera modules available

Current vendor documentation references support for Raspberry Pi ecosystem camera modules including IMX477, IMX219, and IMX708, with version-specific notes about supported variants.

Validated camera options
Module	Sensor	Driver
Raspberry Pi HQ Camera	IMX477	Included in official image
Raspberry Pi Camera Module 2	IMX219	Included in official image
Raspberry Pi Camera Module 3	IMX708	Included in official image
Arducam PiVariety	Mira220	Contact RidgeRun

For custom imaging requirements, RidgeRun can help with board-specific sensor enablement. See also Mira220 Camera Driver for Rubik Pi 3.

40-pin header

The 40-pin header makes RUBIK Pi 3 useful not only as an AI board but also as an integration board for:

GPIO-triggered capture systems
UART-attached microcontrollers
I2C environmental sensors
SPI peripherals
robotics prototypes

RUBIK Pi 3 is compatible with the WiringPi library. Users can get the information of all the pins with gpio readall

 +-----+-----+---------+------+---+RUBIK Pi 3+---+------+---------+-----+-----+
 | QC  | wPi |   Name  | Mode | V | Physical | V | Mode | Name    | wPi | QC  |
 +-----+-----+---------+------+---+----++----+---+------+---------+-----+-----+
 |     |     |    3.3v |      |   |  1 || 2  |   |      | 5v      |     |     |
 |   4 |   8 |   SDA.1 | alt1 | 1 |  3 || 4  |   |      | 5v      |     |     |
 |   5 |   9 |   SCL.1 | alt1 | 1 |  5 || 6  |   |      | 0v      |     |     |
 | 105 |   7 | GPIO105 | alt1 | 0 |  7 || 8  | 1 | alt1 | TxD     | 15  | 10  |
 |     |     |      0v |      |   |  9 || 10 | 0 | alt1 | RxD     | 16  | 11  |
 |   8 |   0 | GPIO. 8 | alt1 | 0 | 11 || 12 | 1 | In   | GPIO. 1 | 1   | 101 |
 |  24 |   2 | GPIO.24 |   In | 0 | 13 || 14 |   |      | 0v      |     |     |
 |  25 |   3 | GPIO.25 |   In | 0 | 15 || 16 | 0 | In   | GPIO. 4 | 4   | 26  |
 |     |     |    3.3v |      |   | 17 || 18 | 0 | In   | GPIO. 5 | 5   | 27  |
 |  49 |  12 |    MOSI | alt1 | 0 | 19 || 20 |   |      | 0v      |     |     |
 |  48 |  13 |    MISO | alt1 | 1 | 21 || 22 | 0 | In   | GPIO. 6 | 6   | 44  |
 |  50 |  14 |    SCLK | alt1 | 0 | 23 || 24 | 1 | alt1 | CE0     | 10  | 51  |
 |     |     |      0v |      |   | 25 || 26 | 0 | In   | CE1     | 11  | 55  |
 |  12 |  30 |   SDA.0 |   In | 1 | 27 || 28 | 1 | In   | SCL.0   | 31  | 13  |
 |  32 |  21 | GPIO.21 |   In | 0 | 29 || 30 |   |      | 0v      |     |     |
 |  33 |  22 | GPIO.22 |   In | 0 | 31 || 32 |-1 |  -   | GPIO.26 | 26  | -1  |
 |  34 |  23 | GPIO.23 |   In | 0 | 33 || 34 |   |      | 0v      |     |     |
 | 103 |  24 | GPIO.24 |   In | 0 | 35 || 36 | 1 | alt1 | GPIO.27 | 27  | 9   |
 |  35 |  25 | GPIO.25 |   In | 0 | 37 || 38 | 0 | In   | GPIO.28 | 28  | 102 |
 |     |     |      0v |      |   | 39 || 40 | 0 | In   | GPIO.29 | 29  | 104 |
 +-----+-----+---------+------+---+----++----+---+------+---------+-----+-----+
 | QC  | wPi |   Name  | Mode | V | Physical | V | Mode | Name    | wPi | QC  |
 +-----+-----+---------+------+---+RUBIK Pi 3+---+------+---------+-----+-----+

To set a gpio as input or output, use the corresponding gpio number (Physical column) as follows:

gpio mode <number> <in or out>

To set the level of a GPIO use:

gpio write <number> <1 or 0>

USB, storage, and expansion

RUBIK Pi 3 exposes multiple USB roles:

2 x USB 3.0 Type-A in host mode
1 x USB 2.0 Type-A in host or device mode
1 x USB 3.1 Gen 1 Type-C in host or device mode

The board also includes an M.2 Key M connector for storage / expansion-oriented workflows, meaning it supports faster storage and reduces the need for USB-attached external SSDs in edge-logging and AI-box use cases.

Power, thermal, and mechanical considerations

Power

Use USB-C PD 3.0 power input with a 12 V / 3 A adapter. Never use the USB port of a computer to power the RUBIK Pi.

Thermal

Like most edge AI boards, RUBIK Pi 3 should be treated as a thermally managed platform, not a bare board that can be ignored once it boots. If your workload involves sustained inference, multiple cameras, or encode / decode concurrency, plan for airflow, an active cooler, or an enclosure with controlled thermal characteristics.

Mechanical resources

The current vendor Hardware Resources page lists not only datasheets but also 3D resources and case models.

RidgeRun perspective on hardware evaluation

A board can look perfect on paper and still fail in productization because of sensor integration, zero-copy video movement, driver gaps, or thermal / power interaction under sustained workloads. On QCS6490-class platforms, RidgeRun typically adds value in three hardware-adjacent areas:

camera and sensor enablement,
multimedia pipeline integration using accelerators,
and board support customization for product-specific peripherals.

If your project has already moved beyond “Can it boot Ubuntu?” and into “Can it ship a reliable smart camera or robot?”, continue to Rubik Pi 3/Consulting and Customization.

Frequently asked questions

Is RUBIK Pi 3 a QCS6490 SBC?: Yes. RUBIK Pi 3 is a single-board computer built on a QCS6490-based compute platform and intended for edge AI and embedded development.
What are the main features of the QCS6490 chipset?: The QCS6490 combines an 8-core CPU, Adreno GPU, Qualcomm AI Engine, advanced camera ISP capabilities, video processing, and embedded I/O such as PCIe, USB, serial buses, and GPIO.
Where can I find RUBIK Pi 3 specs?: You can start with this page for a practical summary, then use the official vendor datasheet and Hardware Resources page for the current downloadable board documents.
Where can I find the RUBIK Pi 3 schematic?: The current official Hardware Resources page lists a hardware schematic entry. Use that page as the authoritative vendor source.
Can I use Raspberry Pi cameras with RUBIK Pi 3?: Current official documentation references validated camera modules such as IMX219, IMX477, and IMX708, but exact supported variants and image-version behavior should be verified before deployment.

Related pages

❯