Greetings everyone, and welcome back.
This Article is a bit different from my usual projects. In this one, I’ll be showing how I set up Bazzite on a LattePanda MU to run Steam.
I’m currently working on a custom PC project built around single-board computers like the LattePanda and Raspberry Pi. Instead of including the Bazzite setup as a small part of that main build, I decided to create a separate article so I could explain the process properly and in more detail. Bazzite is a Linux-based operating system focused on gaming, built on Fedora and optimized for running Steam and other PC games, making it a great fit for compact x86 systems like the LattePanda MU.
This article covers the installation of Bazzite OS along with a few basic tests, including running some older games to check performance.
Let’s get started.
MATERIALS REQUIREDThe following were the materials used in this project:
- Lattepanda MU N305 with cooler
- Lattepanda MU Full Evaluation Board
- Bazzite OS
- USB Drive
- Monitor
- Game controller (I'm using an Xbox controller)
- STEAM GAMES
For this setup, we are using the LattePanda MUpowered by the Intel Core i3-N305 processor. The LattePanda MU is a compact x86 compute module designed to deliver desktop-class performance in a very small form factor.
At the core of the module is the Intel Core i3-N305, an 8-core processor based on Intel’s Gracemont architecture. It is optimized for efficiency while still providing enough performance for everyday desktop tasks, media playback, and light gaming. The module comes with onboard LPDDR5 memory, offering high bandwidth and low power consumption, which makes it well suited for compact and embedded PC builds.
The LattePanda MU integrates memory and storage directly on the module, allowing it to function as a complete PC when paired with a carrier or evaluation board. Graphics are handled by Intel UHD Graphics, making it suitable for media playback, older games, and moderately demanding workloads.
Its modular design is a key advantage. Instead of a traditional motherboard, the compute module simply plugs into a carrier board, making system integration clean and flexible—ideal for compact custom PCs and experimental setups like running Bazzite and Steam in a small enclosure.
Another big plus is how community-friendly the platform is. DFRobot, who make the LattePanda MU, provide complete PCB files and schematics for both the Lite carrier and the full evaluation board. This means anyone can study the design, modify it, or even create their own custom carrier board. In fact, people in the community have already designed their own MU breakout boards, which is pretty cool and shows how open and flexible the ecosystem around the MU really is.
The LattePanda MU is available in two main variants, both sharing the same compact compute-module form factor but differing significantly in performance and use case.
At the lower end is the Intel Processor N100 variant. This model features a 4-core, 4-thread CPU with a maximum turbo frequency of up to 3.4 GHz. It is paired with 8 GB of LPDDR5 memory running at 4800 MT/s and Intel UHD Graphics with 24 execution units, clocked up to 750 MHz. With a configurable TDP range of 6 to 35 W, this version is ideal for low-power desktops, embedded systems, kiosks, and lightweight Linux or Windows workloads where efficiency matters more than raw performance.
The higher-end option is the Intel Core i3-N305 variant, which is the version used in this project. It comes with an 8-core, 8-thread CPU capable of boosting up to 3.8 GHz, effectively doubling the core count compared to the N100. This model includes 16 GB of LPDDR5 memory at 4800 MT/s and a more capable Intel UHD Graphics configuration with 32 execution units, reaching up to 1.25 GHz. Its configurable TDP range is 9 to 35 W, allowing it to scale from efficient operation to higher sustained performance when needed.
Both variants share the same 6 MB cache, support in-band ECC memory, and use Intel UHD Graphics, but the i3-N305 offers noticeably better multitasking, graphics performance, and overall responsiveness. In practical terms, the N100 is well suited for basic desktop use and embedded applications, while the i3-N305 is better suited for heavier workloads, light gaming, and projects like running Bazzite and Steam in a compact PC.
In short:
- N100 MU → efficient, lower power, entry-level x86 compute
- i3-N305 MU → higher performance, better graphics, more headroom for demanding tasks
Both benefit from the same modular design and open carrier ecosystem, making it easy to choose the variant that best fits a specific project.
You can check out more about MU from its well-documented wiki page.
https://docs.lattepanda.com/content/mu_edition/introduction/
https://github.com/LattePandaTeam/LattePanda-Mu/tree/main/Mechanicals
EVALUATION BOARDThe LattePanda MU Full Evaluation Board is a fully featured carrier board designed to expose nearly all interfaces of the LattePanda MU compute module, making it ideal for evaluation, development, and full system prototyping without the need for a custom PCB.
It provides standard desktop-style I/O, including dual full-size HDMI ports for display output, multiple USB ports for peripherals, Gigabit Ethernet for networking, and a 3.5 mm audio jack for sound. For expansion and storage, the board exposes the MU’s PCIe interface through an M.2 slot, allowing the use of NVMe SSDs and other high-speed peripherals, making it suitable for running full desktop operating systems like Bazzite without relying on external USB storage.
Power delivery is designed to support the MU’s configurable TDP range, ensuring stable operation across both low-power and higher-performance workloads, while additional headers provide access to low-level signals for debugging and hardware testing. By exposing all major interfaces in an accessible form, the Full Evaluation Board significantly shortens development time and enables rapid experimentation, making it a strong fit for projects like Parallel PC, where flexibility and expandability are key.
Check out its wiki page for more details.
https://docs.lattepanda.com/content/mu_edition/full_eval/#get-started
For storage, we are using a Gen4 NVMe SSD in the 2230 form factor (M-key). While the LattePanda MU Full Evaluation Board does include M.2 slots, these are E-key and B-key, which are not compatible with M-key NVMe SSDs. Using a SATA-based SSD would have significantly reduced OS performance, so instead we used the PCIe x16 to M.2 adapter included with the evaluation board kit. This allows the NVMe SSD to run over PCIe and be used as the primary system storage.
The LattePanda MU itself does not include onboard Wi-Fi or Bluetooth, and the evaluation board also does not provide a built-in wireless module. To add wireless connectivity, we installed an Intel 7265NGW M.2 E-key module, which provides both Wi-Fi and Bluetooth support, allowing the system to connect to wireless networks and peripherals without relying on external adapters.
DFROBOT SERVICESpecial thanks to DFRobotfor 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.
https://www.dfrobot.com/product-2902.html
https://www.dfrobot.com/product-2821.html
LATTEPANDA MU SETUPWe started the MU assembly by first installing the included cooler on top of the processor. The cooler was secured to the MU using three screws, making sure to remove the protective cover from the pre-applied thermal grease before placing it over the processor die.
Next, the MU was inserted into the SO-DIMM slot on the evaluation board and secured in place using two M2 screws. Finally, the cooler’s power cable was connected to the fan header provided on the evaluation board.
EDITING TDP IN BIOSBefore starting the installation process, we made one important change in the BIOS. The default TDP of the device was set to 15 W, and we increased it to 20 W.
This change was made to give the LattePanda MU more sustained performance headroom under load. At the default 15 W limit, the CPU tends to throttle earlier during CPU- or GPU-heavy tasks, which can affect performance in demanding applications like gaming, compilation, or system updates. By increasing the TDP to 20 W, the processor is able to maintain higher clock speeds for longer periods, resulting in smoother performance and more consistent frame rates—especially important when running Bazzite and Steam.
Since the system uses an adequate power supply and cooling setup, this higher TDP setting remains within safe operating limits while delivering noticeably better real-world performance.
BAZZITE BOOTABLE USBTo download Bazzite, we first visited the official Bazzite website and navigated to the download page. From the hardware selection menu, we chose Desktop as the target platform. Next, we selected Intel UHD/HD/Iris as the GPU option, since the LattePanda MU uses Intel integrated graphics. For the desktop environment, we chose KDE.
https://bazzite.gg/#image-picker
After filling in these options, we were presented with two images: Bazzite Deck and Bazzite Deck Legacy. We went ahead with the standard Bazzite Deck image. It’s also worth noting that if you’re already running a Fedora Atomic desktop, Bazzite can be installed by rebasing instead of doing a fresh installation.
To make the bootable USB, we used Rufus. We first plugged in a USB drive and opened Rufus, which automatically detected it. Then we selected the Bazzite ISO file and left all the settings exactly as they were—MBR for the partition scheme and BIOS as the target system.
After hitting Start, Rufus did its thing, and a couple of minutes later, the bootable drive was ready to go.
SETTING UP BAZZITE OSIn the BIOS, we go to the boot menu and select the bootable USB to start the installation process. After a few minutes, we’re taken to the Bazzite installer, where we select the installation language.
From there, we head to System → Installation Destination and choose our NVMe SSD as the primary storage. Once that’s done, we start the installation. The whole process takes around 15 to 20 minutes, and after it finishes, we’re booted straight into Bazzite.
STEAMOnce Bazzite boots up, we’re greeted with the Steam setup screen. After signing in, we’re dropped straight into Steam. Using the keyboard, we navigated to the settings menu, enabled Bluetooth, and connected a Bluetooth speaker along with a game controller.
With that done, the whole setup effectively turns into a Steam machine—an underpowered one, sure, but a Steam machine nonetheless.
DOES IT RUN DOOM?No hardware test is complete without asking: can it run Doom?
The answer was a very confident yes. Doom (1993) ran perfectly smooth, which honestly wasn’t much of a surprise. At this point, Doom has been ported to everything from calculators to receipt printers, so watching it run on an Intel N305 felt almost ceremonial.
The system barely broke a sweat, quietly reminding us that this tiny SBC has far more power than the hardware Doom was ever meant to run on.
SILKSONGWe also tested a recent indie title, Hollow Knight: Silksong, and it ran really well on this setup. That’s not too surprising, though—Silksong is a beautifully animated 2D game that relies more on art style and smooth animations than raw GPU power. Because of that, it plays perfectly to the strengths of the N305 and Intel integrated graphics, delivering smooth frame rates and responsive gameplay without stressing the system. It’s a great example of how this setup shines with well-optimized indie games.
CYBERPUNK 2077On the complete opposite end of the spectrum is Cyberpunk 2077—arguably one of the most brutally unoptimized games you could throw at this setup. It absolutely hammers the GPU, shows zero mercy to integrated graphics, and turns the experience into more of a benchmark than a game.
It’s a perfect reminder of where the limits are… and where integrated graphics stop rendering frames and start rendering excuses.
BAYONETTA 1We next tested Bayonetta 1, which ran exceptionally smoothly. Performance was consistently above 60 FPS, with no noticeable frame drops, stuttering, or input lag. The experience felt comparable to running the game on a standard desktop PC, confirming that the system can handle less demanding titles very comfortably.
DOOM ETERNAL BONUSBefore this setup, we also tested a different configuration using another carrier board. In that version, we used the Lite carrier board paired with a SATA SSD, along with an M.2-to-PCIe x16 adapter to add a GT 1030 as an external GPU. With this setup, we tested Doom Eternal, and the performance was noticeably better. The game ran at a stable 40–50 FPS, which is quite impressive considering the age of the GPU and the overall hardware configuration.
CONCLUSIONThe overall conclusion of this project is simple: whatever we throw at it works—and works well enough to be genuinely playable. What really stood out to me was how comfortably this setup handled games like Bayonetta, Doom Eternal, and even NieR: Automata. That’s honestly an impressive feat when you consider that all of this is running on an SBC with integrated graphics.
Don’t let the large carrier board fool you—the actual LattePanda MU itself is roughly the size of a credit card. And despite that tiny footprint, it’s capable of running workloads and games that a Raspberry Pi can only dream of. I did try running Steam on a Raspberry Pi, and while it technically works, the experience was very mixed. In contrast, this setup feels like a near-perfect match for Bazzite.
There’s also plenty of room to grow. Adding a dedicated GPU would take this system to a completely different level in terms of performance.
Overall, this SBC can now officially “run Doom, ” and it fits perfectly into our upcoming Parallel PCproject, where it will play a central role.
Special thanks to DFRobotfor 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. If you’re into electronics, be sure to check out DFRobot for a wide range of products, including modules, sensors, and single-board computers.
Comments