Compiling Software on the Minisforum M2, M2 Pro and BOSGAME VTA-439

For this article, I’m comparing three mini PCs by compiling software from source: the Minisforum M2, Minisforum M2 Pro-358H and BOSGAME VTA-439. My earlier CPU benchmarks provide a useful starting point, but I also want to see how these machines handle a practical workload that draws on processor cores, caches, memory and storage.

I’m using fooyin, a highly customisable music player, as the test application. Its substantial codebase makes it a useful compilation test, with enough work to keep each machine busy for several minutes.

Hardware

The Intel processors have more physical cores, while the AMD processor provides more hardware threads through simultaneous multithreading. The Intel cores are a mixture of different core types, so the counts alone don’t tell the whole story. The M2 Pro also has faster memory than either competitor.

Building fooyin

I’m compiling fooyin version 0.14.1 with the same source code and build configuration on all three machines, each running CachyOS. With the development tools and dependencies installed, I use CMake to configure the project and Ninja to build it.

First, download the source code and its submodules, then enter the project directory.

git clone --recurse-submodules --branch v0.14.1 \
  https://github.com/fooyin/fooyin.git
cd fooyin

I configure a Release build with compiler caching, precompiled headers and testing disabled.

cmake -S . -B build -G Ninja \
  -DCMAKE_BUILD_TYPE=Release \
  -DBUILD_CCACHE=OFF \
  -DBUILD_PCH=OFF \
  -DBUILD_TESTING=OFF

I then compile the application using all available processor threads.

/usr/bin/time -f 'Elapsed time: %e seconds' \
  cmake --build build --clean-first --parallel "$(nproc)"

The command removes previous build outputs before compiling, so existing object files aren’t reused. time measures elapsed wall-clock time, including the cleaning step.

Benchmark methodology

I used the performance CPU governor on all three machines and ran each benchmark three times, with a clean build for every run. The results are the arithmetic mean of those three timings. Downloading the source and the initial CMake configuration are excluded.

A clean build removes compiled outputs, but does not clear the operating system’s filesystem cache. These measurements therefore represent repeated clean builds, rather than builds starting with an empty filesystem cache. Lower times indicate better performance.

Compilation performance

The M2 Pro finishes comfortably ahead of both competitors. The M2 and BOSGAME are closely matched; their small difference deserves less emphasis without examining the variation between individual runs.

Analysis

The M2 Pro completes the build about 9% sooner than the BOSGAME. That reverses the order seen in my earlier CPU benchmarks, where the Ryzen AI 9 HX 470 outperformed the Core Ultra X7 358H. The standard M2 also does well, finishing only slightly behind the BOSGAME despite having fewer hardware threads.

Compilation mixes parallel work with stages that must wait for other tasks to finish. More threads can help, but their usefulness depends on the build. Logical threads also share a physical core’s resources, so having 24 threads rather than 16 does not guarantee a shorter compilation time.

The M2 Pro’s faster memory may contribute to its lead, alongside differences in cache, clock speeds and power limits. Its memory has a roughly 33% higher nominal transfer rate, but that does not establish an equivalent bandwidth advantage or explain how much memory performance contributes here. This benchmark measures the complete systems rather than isolating those factors.

Summary

The M2 Pro is the quickest of these three machines at compiling fooyin with this build configuration. The result makes a useful counterpoint to my earlier CPU benchmarks and gives me another reason to test the jobs I actually use these machines for. One application cannot settle which system is best for every developer, but here the M2 Pro has a clear lead.

Complete list of articles in this series