On 15 May 2025, Phoronix published an article comparing the performance of several Linux distributions on a Framework Laptop 13. Out of curiosity we are going to check a pair of things with a similar evaluation. In this case the hardware being put to test is a ThinkPad E14 Gen 7, with an AMD Ryzen 5 220 CPU and 32 GB of RAM.

Linux distributions

Here there are six distinct Linux distributions, all installed with their supposedly default settings.

For instance, Arch Linux was installed on an ext4 partition, an example in the installation guide was pointing toward that choice. Instead Gentoo's recommended filesystem has always been xfs:

XFS is the recommended all-purpose, all-platform filesystem.

The partition was mounted with the noatime flag. The compilation flags were the same seen in the handbook:

COMMON_FLAGS="-march=native -O2 -pipe"

CachyOS and Fedora were both installed on btrfs partitions, with CachyOS utilizing the noatime flag. CachyOS, Debian and Gentoo were also tested with an important change to their default installations: CachyOS on an ext4 partition, Debian and Gentoo with a more recent kernel.

Phoronix Test Suite

The majority of tests were executed through the Phoronix Test Suite 10.8.4. The benchmarks were chosen from the most popular ones, as provided by the list-recommended-tests command. A subset of these tests and their options is shown below, for the complete results check:

OpenBenchmarking.org - uploaded results

LevelDB

Test type: System

Gentoo is notably slow here, this could be an edge case where xfs is penalized. Other tests show a less surprising performance pattern.

SQLite

Test type: Disk

Btrfs underperforms in this test. This was expected, see this other Phoronix article:

Bcachefs, Btrfs, EXT4, F2FS & XFS File-System Performance On Linux 6.15

SuperTuxKart

Test type: Graphics

Phoronix Test Suite recorded a screen resolution of 3072x1920 for Debian and Ubuntu, while the actual hardware resolution was 1920x1200. This could explain the sluggish performance in this benchmark.

The situation is the same with Vulkan. While the FPS doubled, shadows and other effects were missing in this version of the test.

Unvanquished

Test type: Graphics

Unvanquished runs flawlessly on Debian and Ubuntu and, interestingly, Arch Linux is faster than CachyOS.

MBW

Test type: Memory

Timed Linux Kernel Compilation

Test type: Processor

CachyOS and Debian were the fastest while Gentoo was severely beaten at its own game.

BRL-CAD

Test type: System

Geometric mean

Here is the geometric mean of all the results, including Waifu2x-NCNN Vulkan, 7-Zip Compression and Blender.

However, excluding LevelDB (Fill Sync), SQLite, and SuperTuxKart, we get a more balanced and perhaps fairer representation.

CachyOS is the fastest, followed by Debian.

PassMark

Another bunch of tests were executed with PassMark PerformanceTest Linux (11.0.1004).

In this case Debian and Gentoo Linux are the fastest, while Arch Linux is the slowest. CachyOS (on btrfs) encountered an issue during the memory benchmark:

Memory Mark:                       2618
  Database Operations              5781 Thousand Operations/s
  Memory Read Cached               37405 MB/s
  Memory Read Uncached             36773 MB/s
  Memory Write                     28453 MB/s
  Available RAM                    3560 Megabytes
  Memory Latency                   58 Nanoseconds
  Memory Threaded                  57572 MB/s

Memory Mark was reporting only 3560 MB of available RAM. This should be a one-time anomaly with PassMark.

Geekbench

Last, there is Geekbench 6.5.0.

Geekbench shows results in line with the Phoronix Test Suite.

Conclusion

Overall, what Phoronix shows in its article is confirmed here: CachyOS appears to be the most performant Linux distribution, followed by Debian. While Arch and Gentoo offer great opportunities for customization, they are behind, unless you invest more time in optimization. There is no substantial difference between a 6.12 or a 6.18 kernel in these tests. Many of the benchmarks out there have quirks that needs to be taken into account when interpreting the results.