Parallel PC is a DIY all-in-one computer featuring a custom body made from wooden panels and 3D-printed parts, built around a 15-inch LED display. What makes it special is that it houses not one but two different single-board computers. One is the ARM-based Raspberry Pi Compute Module 5 with its official evaluation board, and the other is an x86-based LattePanda MU.
The idea behind this project was to take one of my previous builds—the WoodWorks Fusion PC—and transform it into a dual-compute system. With the press of a single button, the active computer connected to the main display can be switched instantly. If you want to work on Raspberry Pi projects, you switch to the Pi. If you need an x86 environment to run Windows or, in our case, Bazzite, you switch over to the LattePanda. One enclosure, one monitor, two completely different computing platforms.
Built using a combination of wooden panels and custom 3D-printed parts, this project is the starting point of the Parallel PC series. The goal is to keep evolving this enclosure—retrofitting new hardware, adding more single-board computers, and experimenting with different ideas over time.
LATTEPANDA MU FULL EVALUATION BOARD SETUP
For the main computer in The Parallel Desk, we are using the LattePanda MU, powered by the Intel Core i3-N305 processor.
The LattePanda MU is a compact, SO-DIMM–style compute module that integrates a full x86 PC onto a small form factor. At its core is the Intel Core i3-N305, an 8-core processor based on Intel’s Gracemont architecture, designed for high efficiency while still delivering solid desktop-class performance. This makes it well suited for light gaming, emulation, media playback, and general desktop workloads.
The module comes with 16 GB of LPDDR5 memory onboard, providing high bandwidth and low power consumption, along with onboard NVMe storage, eliminating the need for external SATA or USB boot devices. Graphics are handled by Intel UHD Graphics, which supports modern display standards and hardware-accelerated media decoding. Despite its small size, the MU is essentially a complete PC, requiring only power, I/O, and display connections to function.
One of the most interesting aspects of the LattePanda MU is its modular design. Instead of building around a traditional motherboard, the compute module simply plugs into a carrier or evaluation board—much like laptop RAM—making system integration clean and flexible.
To make the MU usable as a full desktop system, we paired it with the LattePanda MU Full Evaluation Board, which acts as a carrier and breakout platform for all of the module’s interfaces.
The evaluation board exposes standard PC I/O, including multiple USB ports, full-size HDMI and DisplayPort outputs, Gigabit Ethernet, and audio I/O, allowing the MU to be used like a conventional desktop computer. It also provides PCIe expansion, enabling support for high-speed peripherals such as NVMe SSDs, network cards, or other PCIe devices.
The LattePanda MU was chosen because it brings true x86 performance into a very small footprint. Paired with the full evaluation board, it behaves like a traditional desktop PC while remaining compact enough to coexist with a Raspberry Pi Compute Module inside the same enclosure.
Check out one of my previous LattePanda MU-based projects.
https://www.hackster.io/Arnov_Sharma_makes/the-pvm-panda-69cdfa
RASPBERRY PI CM5 SETUP
The second single-board computer used in this project is the Raspberry Pi Compute Module 5. It is powered by the Broadcom BCM2712 quad-core 64-bit Arm Cortex-A76 processor running at 2.4 GHz, offering a significant performance improvement over previous Compute Module generations. The module is available with LPDDR4-4267 SDRAM in 2 GB, 4 GB, 8 GB, and 16 GB variants, all supporting error correction. Storage options include a Lite version with no onboard storage, as well as eMMC capacities of 16 GB, 32 GB, and 64 GB.
For connectivity, the Compute Module 5 supports dual-band Wi-Fi (2.4 GHz and 5 GHz) with IEEE 802.11 b/g/n/ac, Bluetooth 5.0 with BLE, and a Gigabit Ethernet PHY with IEEE 1588 support. Expansion and interfacing options include PCIe Gen 2 with a single x1 root complex at 5 Gbps, USB 3.0 and USB 2.0, and up to 30 configurable GPIO pins operating at 1.8 V or 3.3 V. Additional interfaces such as UART, I2C, SPI, SDIO, DPI (RGB display), I2S, PWM, and GPCLK make the module highly flexible for embedded applications.
In terms of multimedia capabilities, the Compute Module 5 provides two HDMI 2.0 outputs capable of 4K 60 fps, along with two four-lane MIPI connectors for camera and display connectivity. Graphics support includes a 4K 60 fps HEVC decoder, OpenGL ES 3.1, and Vulkan 1.2. Power is supplied through a single 5 V input with USB Power Delivery support, allowing a current draw of up to 5 A. With an operating temperature range of −20°C to +85°C, the module is well-suited for long-term and industrial deployments.
Unlike standard Raspberry Pi boards, the Compute Module 5 does not include onboard USB, HDMI, or Ethernet connectors. Instead, these interfaces are exposed through high-density edge connectors, allowing the module to be used with an evaluation board or a custom carrier board.
The official Compute Module 5 IO Board acts as a breakout platform, exposing all major interfaces of the module. It features a standard 40-pin GPIO header, two full-size HDMI 2.0 ports, two USB 3.0 ports, Gigabit Ethernet with PoE support, and an M.2 PCIe slot for expansion. It also includes two four-lane MIPI DSI/CSI-2 connectors for direct display and camera connections. The board is powered via USB-C, ensuring stable power delivery for high-performance use cases.
One of the main reasons for choosing the Compute Module 5 for this project is its NVMe support and dual full-size HDMI outputs, which make it ideal for Raspberry Pi–based development inside a custom PC enclosure. Since most of my projects are built around Raspberry Pi platforms, having a Compute Module inside this system was an essential design choice for future projects.
Check out a previous CM5-based project of mine.
Woodwork Fusion PC
The original WoodWorks Fusion PC was a custom all-in-one computer built from scratch using a combination of wood and 3D-printed parts, inspired by CyberDesk-style aesthetics.
The goal was not performance but form—creating a visually striking enclosure around deliberately outdated hardware, basically a potato PC.
At its core, the system used a 4th-generation Intel i3 desktop CPU mounted on a Mini-ATX motherboard, paired with 12 GB of DDR3 RAM and a GT 710 GPU. While the hardware was clearly obsolete, the enclosure was designed to be modular, allowing components to be easily upgraded in the future without changing the overall structure.
The body was primarily constructed from plywood, chosen for its strength and ease of fabrication. 3D-printed L-brackets were used to join panels together, forming a rigid cuboid enclosure capable of housing the motherboard, power supply, storage drives, and display.
DISPLAY
In the previous version of this project, we reused an old LCD monitor from around 2012. It was a Samsung 4:3 display, a format that was very common during the early days of LCD screens.
At the time, most content and applications were designed around this aspect ratio. As display technology evolved, wider screens became the standard, driven by changes in content consumption, improved productivity workflows, and the demand for more immersive experiences in multimedia, professional work, and gaming.
For this revised build, we switched to a much slimmer 15-inch LED display, salvaged from a low-cost monitor purchased for under $25.
We stripped down a low-cost monitor and reused its TFT panel for this project. The monitor’s HDMI driver board was retained, but its original power supply was intentionally omitted. Instead, the display is powered by the same SMPS used for the other internal components. This approach helped reduce the overall size of the wooden frame, making the final build more compact.
DFROBOT SERVICE
Special thanks to DFRobot for providing the hardware used in this project. Both the LattePanda MU N305 and the Full Evaluation Board were supplied as review units for testing purposes.
DFRobot had no control over the build process, testing methods, or results shared in this project, and all opinions and conclusions are entirely my own. If you’re interested in electronics-related products such as modules, sensors, and single-board computers, you can check out their offerings on the DFRobot website.
HDMI SPLITTER
Our goal was to use a single LED display panel as the main display while running two SMPS units. Since only one SBC can drive the display at a time, an HDMI splitter became the obvious solution. We used a three-input, one-output HDMI splitter from Sounce. Opening it up was straightforward—the enclosure doesn’t use screws and is instead held together with snap locks. Using a prying tool, we were able to open the casing and access the circuit inside.
The board features proper HDMI connectors: one output port and three input ports. Each HDMI input has an associated LED indicator that lights up to show which input is currently selected. The main control element on the board is a push button, which allows us to switch the active input connected to the output.
DUAL SBC SAME DISPLAY SETUP
To test the HDMI splitter, we set up a minimal configuration using two SBCs. The HDMI output of the LattePanda was connected to the first HDMI input of the splitter, while the Raspberry Pi Compute Module was connected to the second HDMI input.
A monitor was connected to the HDMI output of the splitter. By default, the display shows the output from the first HDMI input. When the push button is pressed, the display output switches from the first SBC to the second.
UPDATED 3D DESIGN
The 3D model for this project was created by reusing and modifying the model from our previous WoodWorks Fusion PC build. Reusing parts from the older project was one of the main goals, as we wanted to evolve the original design rather than start from scratch.
In the updated model, the old display was removed and replaced with a new LED panel. The new screen is slightly smaller and significantly thinner than the previous LCD monitor, which required changes to the overall size of the wooden frame and panels. Along with this, the display holder and the front screen mount were also redesigned.
The original motherboard, power supply, and internal hardware were completely removed, leaving the model empty before adding the new components. We then integrated the 3D model of the LattePanda MU expansion board into the frame. The SMPS power supply was also modeled and positioned on the side wooden panel, slightly above the grill area.
The grill section itself was redesigned and now serves as a mounting point for both the HDMI splitter and the display driver board. On the top side, a mesh-style component was modeled, which also functions as the holder for the Raspberry Pi Compute Module.
Finally, a panel-mounted push button was added to the front of the model, which will be used to control the HDMI splitter output.
Below are the 3D model parts we used in this project and their details.
- L Bracket with four Holes—a total of six of them are required, and they will be printed using brown PLA with 25% infill. These L brackets were the only things we reused from the previous woodwork fusion PC project.
- L Bracket Long Version—This will be used to hold the display in place, and we need two of them with the same print settings and 25% infill.
- Lattepanda MU Holder—The MU Holder was printed in White Hyper PLA, with an infill of 50%.
- Screen Holder Front—four pieces were printed with Black Hyper PLA with 50% infill each.
- Raspberry Pi CM5 board Frame/Grill—This grill-holder part was printed in Hyper Black PLA with an infill of 30%.
- SIDE GRILL—This part was printed with White Hyper PLA.
- Display Holder—We printed this part (two holders) with grey PLA with an infill of 25%.
- Nametag—We printed this part with white Hyper PLA with an infill of 25%.
This is link for 3D MODEL https://a360.co/3MQalQh
RESULT- RUNNING STEAM
Here’s the final result of this simple yet time-intensive build: Parallel PC—Wood Edition. It’s a balanced blend of wood and 3D-printed components, brought together to create a custom all-in-one PC that features not one, but two different computing platforms. One system is an ARM-based SBC, while the other is an x86 machine powered by a mobile processor.
Having two computers in a single enclosure allows us to switch between platforms depending on the task. We can work on Raspberry Pi–based projects and embedded development, then seamlessly switch over to x86 for workloads that require it. In our case, the LattePanda MU runs Bazzite, giving us access to Steam and enabling proper PC gaming—all within the same machine.
The very first thing we did on the Parallel PC was open Steam on the LattePanda MU.
And, as tradition demands, the sacred question was asked: can it run Doom?
The answer was yes. Doom (1993) ran buttery smooth—which honestly wasn’t shocking. At this point Doom has been run on everything from calculators to receipt printers, so the N305 barely breaks a sweat.
Naturally, we got cocky and asked the next question: can it run Doom Eternal? Surprisingly yes again. The game hovered around a playable 30 FPS and felt totally usable. And remember—this is without a dedicated GPU, which makes it even more impressive.
Riding that high, we went full send and tried Cyberpunk 2077. That’s where reality stepped in. At around 5–10 FPS, the experience was more slideshow than game—but hey, it did launch and run, and that alone feels like a small victory.
Next up was NieR: Automata, which ran incredibly smoothly and ended up being the best-performing title we tested. We followed that with Left 4 Dead 2 and The Elder Scrolls V: Skyrim, both of which ran without any issues. These tests confirmed that the setup shines when it comes to older and moderately demanding games.
In a future revision, we plan to add a dedicated GPU and run these tests again to see just how far we can push this system.
Another big win with running Bazzite on the LattePanda MU is that it’s not just about gaming. You can drop straight into desktop mode and get a full Linux environment with a proper terminal—perfect for tinkering, testing, and doing regular Linux stuff in between game sessions.
RASPBERRY PI OS (TRIXIE)
Next, using the front-mounted button, we switch the display output from the LattePanda MU to the Raspberry Pi Compute Module 5, which is running the new Trixie OS—a port of Debian Trixie.
Once switched, the system behaves like a full Raspberry Pi 5–class setup. It supports everything people typically use a Pi for, including general Linux desktop work, coding and scripting, hardware interfacing, and embedded project testing. It can even run Steam—we’ve already shared a YouTube Short where we tested Half-Life and Minecraft Java on a Pi 5.
The system can be used for writing and compiling code, running Python scripts, experimenting with GPIO, and working with networking tools and services. Tasks such as controlling LEDs and sensors, running a local web server, or building and testing IoT applications are all possible using the integrated CM5 inside the Parallel PC.
One planned upgrade for Version 2 is to add a breakout board that brings the GPIO pins from inside the enclosure to the front panel, allowing quick connections to the CM5 using jumper wires.
CONCLUSION & WHAT'S NEXT?
Overall, this setup works remarkably well in its current form. The dual-system approach allows seamless switching between an ARM-based Raspberry Pi environment and an x86 desktop system, making the machine flexible and genuinely useful in day-to-day workflows. The shared display, compact layout, and modular construction validate the core idea behind Parallel PC—Wood Edition.
That said, there are a few areas planned for improvement in future revisions. Audio output is currently missing, and the LattePanda MU would benefit from a dedicated GPU for improved graphics performance. On the Raspberry Pi side, a front-mounted expansion board and GPIO breakout would make prototyping more accessible. Additional covers on the left side of the enclosure are also planned to clean up the overall look.
There’s also plenty of room to experiment further. Adding internal partitions within the frame would allow more compute modules to be integrated, potentially turning this into a multi-computer platform. Future ideas include adding sensors, experimenting with a RISC-V–based system, expanding storage, and making aesthetic upgrades such as painting or finishing the wooden panels.
Overall, the project has been a success and serves as a strong foundation for future iterations. A new revision is already in the works, which will incorporate many of these improvements and push the concept even further.
Thanks for reaching this far, and I will be back with a new project pretty soon.
Peace.
