Following the recent benchmarking of Ubuntu Server 25.04 in its near final state compared to prior Ubuntu Linux releases, I turned my attention to Fedora Server 42. On the same AMD EPYC 9005 “Turin” server I carried out some comparison benchmarks of Fedora Server 42 compared to the prior Fedora Server 41 and other Linux distribution releases for seeing how Fedora 42 is competing with other Linux distributions on this 5th Gen AMD EPYC dual socket server.
Fedora Server 42 has shaped up quite nicely, just like the great experience with Fedora Workstation 42. With Fedora Server 42 shipping the Linux 6.14 kernel and having both GCC 15.0.1 and LLVM Clang 20 compiler toolchains available, it’s a very fresh and up-to-date experience. GCC 15.1 stable isn’t even out yet but in usual spring tradition Fedora 42 is already using a near-final GCC 15 snapshot by default, as great news for HPC compute users and others wanting the newest compiler optimizations and hardware target support.
Fedora Server 42 also has PHP 8.4 and Python 3.13 and other fresh software packages for those planning to deploy Fedora Server 42 onto production web servers.
With Fedora’s more liberal update policy and Linux 6.13 already on Fedora 41 (soon to be Linux 6.14), the performance delta from Fedora 41 to Fedora 42 is lower than with Linux distributions like Ubuntu non-LTS releases that don’t update their kernel version mid-cycle. For the purposes of today’s initial Fedora Server 42 benchmarking the following operating systems were freshly tested on the same AMD EPYC server:
– AlmaLinux 9.5
– Fedora Server 41
– Fedora Server 41 + Updates
– Fedora Server 42
– Ubuntu 24.04 LTS
– Ubuntu 24.04 LTS + HWE
– Ubuntu 24.10
– Ubuntu 25.04 Beta
Due to Fedora’s more liberal update policy, Fedora 41 was tested both on a clean install and then with updates for going from Linux 6.11 to 6.13 and other software upgrades. AlmaLinux 9.5 was tossed in for reference of where Red Hat Enterprise Linux 9 performance roughly is today.
The same server was used for all testing with dual AMD EPYC 9755 128-core processors, 1.5TB of DDR5-6000 RAM, Samsung 3.84TB NVMe SSD, and using an AMD Volcano reference server platform. The same hardware was used for testing throughout at the respective default environments: the hardware differences in the system table just come down to how the software exposes the CPU frequency and other information depending upon the kernel / CPUFreq driver and the like.
