Rock Pi X by Radxa
An affordable x86 brain for your robot
The ROCK Pi X is a small, low-cost Intel computer that runs Windows and Linux. Install ARCx on it, mount it in your robot, and use the robot from a web browser on any phone, tablet or computer.
How the ROCK Pi X works with Synthiam. It is the robot's on-board computer. It runs ARCx on 64-bit Windows 10/11 or Ubuntu 24.04 LTS, or ARC on Windows, just like a desktop PC. Servos and sensors connect through an EZB controller over USB or Wi-Fi; cameras, microphones and speakers plug straight into the board. We recommend ARCx, which runs headless and is used from any browser.
What is the ROCK Pi X?
The ROCK Pi X is a single-board computer (SBC) from Radxa: a full computer, with processor, memory, Wi-Fi and USB, on one compact board. It is Radxa's first x86 SBC, similar in concept to the UP Board and LattePanda, and typically more affordable.
"x86" matters because it is the same processor family as most Windows PCs, so Windows software and drivers work on it. It has an HDMI port for a monitor and a 3.5 mm audio jack with microphone support, so it is easy to set up on a workbench with a screen and keyboard and then mount inside a robot.
Key features, and what they mean for a robot
- Intel Cherry Trail Z8350 quad-core processor: the board's brain; four cores handle several robot tasks at once.
- 64-bit LPDDR3 memory, 1, 2 or 4 GB: working memory for the operating system and ARCx. More memory means smoother multitasking.
- HDMI up to 4K at 30 Hz: robots rarely need 4K, but a monitor is handy during setup.
- 3.5 mm audio jack with microphone: speaker output, voice recognition and microphone input.
- 802.11ac Wi-Fi and Bluetooth 4.2: fast wireless networking and Bluetooth accessories such as game controllers.
- Gigabit Ethernet: a reliable wired link, often best during development.
- USB ports: cameras, storage and robot controllers.
- 40-pin color expansion header: for add-ons; not needed for Synthiam, which uses an EZB controller for robot I/O.
- Real-time clock: keeps time for logs and schedules, especially with a backup battery fitted.
- USB PD and QC power: powered from compatible USB-C chargers and power banks.
ROCK Pi X at a glance
| Type | x86 single-board computer |
|---|---|
| Maker | Radxa |
| Processor | Intel Cherry Trail Z8350, quad-core |
| Memory | 1, 2 or 4 GB LPDDR3 |
| Models | Model A and Model B |
| Connectivity | 802.11ac Wi-Fi, Bluetooth 4.2, Gigabit Ethernet, USB, HDMI, 3.5 mm audio with mic |
| Power | USB-C, USB PD and QC |
| Synthiam uses it for | Running ARCx or ARC on board the robot |
| Website | rockpi.org, getting started |
Radxa lists the ROCK Pi X as discontinued (production stopped in 2022). Existing boards work as described here.
Model A, Model B and memory options
The ROCK Pi X comes as Model A and Model B, each with 1, 2 or 4 GB of memory. For the exact hardware differences between the models, check the manufacturer's specifications for your board version. The chart from our friends at Explaining Computers is a helpful overview.
Our recommendation: choose 4 GB if you can. An operating system, ARCx, a camera and several robot skills all share that memory.
What you can do with it in Synthiam
Control from any browser
ARCx runs on the board; you use it from a phone, tablet or laptop. Several people at once, and the robot keeps running when the browser closes.
Servos and sensors
An EZB controller over USB or Wi-Fi adds servos, PWM, digital, analog, I2C and UART ports for movement panels, animations and sensors.
Camera and vision
USB cameras with Camera Source (usb) for face, color, glyph and object tracking.
Voice and sound
The audio jack with microphone support gives the robot ears and a voice: speech recognition, text to speech and sound playback.
AI agents
ARCx is an MCP server, so AI agents can see, drive and program the robot, alongside conversational AI robot skills.
Scripts, Custom UI, remote
Blockly, JavaScript and Python; touch-friendly Custom UI screens; and Exosphere for remote control and telepresence.
Set up ARCx on the ROCK Pi X
- Install a 64-bit operating system. Windows 10 or 11 (64-bit) with the drivers below, or Ubuntu 24.04 LTS (64-bit) Server or Desktop. ARC needs Windows.
- Install ARCx. Follow Install ARCx on Windows 64 or Install ARCx on Linux x64.
-
Open ARCx from another device. Start ARCx Server (Bind Any) on Windows, or on Ubuntu:
Then browse to./ARCx -b http://0.0.0.0:5000http://<rock-pi-address>:5000. -
Connect an EZB controller. Add the Ezb Connection robot skill and enter the controller's port
(
COM3,/dev/ttyUSB0) or IP address, then connect. On Ubuntu, add your user to thedialoutgroup first. - First test. Move a servo with the Servo robot skill or drive with a movement panel, then add a camera and speech.
-
Run headless. Open the project at start-up (the Windows shortcut
-foption, or asystemdservice on Ubuntu, as the install guides describe) and unplug the monitor.
Windows drivers for a smooth setup
Some ROCK Pi X boards need specific Windows drivers before Wi-Fi, audio, graphics acceleration and other hardware work correctly. Synthiam has assembled a driver package to get your board running reliably.
Download ROCK Pi X Windows 10 drivers (x64)
Follow the steps in order. After extracting the ZIP file, open the included README and follow its step-by-step instructions carefully. Installing drivers in the recommended order prevents missing devices and unstable behavior.
Both run on it; ARCx is recommended
ARCx Recommended
- 64-bit Windows 10/11 or Ubuntu 24.04 LTS.
- Headless; used from any web browser by several people at once.
- MCP for AI agents; robot skills from the Synthiam store.
ARC
- Windows desktop application, installed like on any PC.
- Viewed on an HDMI monitor or through remote desktop software.
- A good choice if you already have ARC projects.
Small computers run robots very well but sometimes need tuning, especially with Windows. Our Tips to Make a Robot guide covers freeing storage space, improving performance (start-up apps, power settings, updates), powering the board reliably and running headless with remote desktop tools.
When something does not work
-b http://0.0.0.0:5000 (Ubuntu), browse to the board's IP address rather than 127.0.0.1, and allow port 5000 through the firewall.
Frequently asked questions
Give your robot an x86 brain
Install ARCx on the ROCK Pi X and control your robot from any browser.
Oh - and if you have a Rock Pi X in the invmoov, you can connect it to exosphere and turn it into a telepresence robot. So you can be out for dinner or at your friends and control the robot in your house. You can move stuff around, pick things up, what ever... That's pretty awesome
I should set that up as a live hack one day . Use the inmoov as a telepresence robot and control it remotely from dinner or something haha
Wow, things keep getting better and better. Jeremie is like your mad scientist at Synthiam! It's ALIVE!!!!
hey everyone - I created a quick guide in the support section on improving performance and freeing storage space with SBC (single board computers). This applies to the rock pi x, atomic pi, lattepanda, etc...
You can view it here: https://synthiam.com/Support/Install/free-space-with-windows-10
Thanks for the tips DJ, Optimizing Disk, RAM and CPU will be crucial especially when we want to run a lot of skills and multiple clients.
One area we are lacking is GPU support for ML at edge and the dependance on X86 Architecture is a challenge. Migrating to .NET core with multi architecture support for Lin/Win/Mac opens up a lot more cross platform capabilities. With ARC supporting ARM and CUDA on NVIDIA (Nano, Xavier etc) would be ideal for robotics platforms. Obviously that requires a fair amount of work on Synthiam part. Another option is to look at partnerships with companies like AMD. ROCm is maturing, they are an X86 platform and are launching edge devices enabled by chips like the V2000 so this could be an option to explore in future.
@Nink, All I read and understood was Bla, bla, bla, bla. LOL! I'm so left behind. Thank God for guys like you and DJ that can conger up the magic we all dreamed about when we were kids.
Unfortunately I can’t conjure anything up Dave. All I know is my robot needs a brain. My logic is if I post a lot of three letter acronyms I copied from a random Reddit post, DJ will come back and say, here you go, and it works.
Sadly core doesn’t have a GUI yet. It’s on the dev horizon. Migrating to core wouldn’t be difficult and a significant part of our code would migrate well. Until core matures and becomes something more than a web server.
I also don’t really believe in the ml running local. It doesn’t make sense to consume that much power in the robot locally - specifically since the global internet communication infrastructure is barely at capacity. It was significantly over engineered expecting less data compression. But with media compression the way it is, bandwidth and low latency is readily available for cloud computing.
offloading ml to the cloud is the right thing to do. I don’t even want to consider what batteries would be needed to power a gpu for "useful" ml.
Like, come on... tensor flow is a joke to run locally even on the most powerful hardware. There’s a lot of technologies we kinda skip at Synthiam / mostly because we wait for it to mature and stabilize.
just look at google dialog flow. Ugh that’s the worst. It’s changed so much that it’s impossible to integrate. I should have waited longer before making a skill. Now it’s just a mess.
The open source and education/exploration space is really fragmented and unreliable. By the time you figure out how to implement something, the technology had changed.
so back to core... until it matures AND has a GUI, we can’t use it. However I should add that our entire cloud infrastructure and website is core
entirely.
I love the EZB and this is why I got into EZ-Robot. Suddenly I didn't need a 10 pound laptop duct taped to a robot platform. Everything was offloaded to a desktop and now we had extremely intelligent nimble robots. THANK YOU!!!
Moores law kicked in and now you can put an SBC like Rock Pi X directly on the robot so I don't need an external computer anymore. 7nm has brought in 10w SBC's with CPU/GPU/TPU cores onboard. This opens up a whole new world for Autonomous robots. Network goes down, robots keep running. Sure you can still offload complex ML to the cloud you don't need to process local when network is up, but there is enough smarts on board to allow the robot to continue to function off line. You really don't want a robot that a 12 year old kid can shut down with a $10 jammer he bought on Ebay.
Hopefully as MAUI matures ARC moves over, it sounds like you have given it a lot of thought and it's on your roadmap. I understand that avalonia probably isn't a strategic move when an official UI is in development.