The Performance Of Five Linux Distributions From Early 2016 To The End Of 2018

With the end of another year upon us, there has been the start of many year-end benchmark comparisons looking at how various aspects of Linux performance has evolved over 2018. In this comparison though is going back further than that and seeing how five Linux distributions have experienced performance changes over the past nearly three years -- using the CentOS, Clear Linux, Fedora, and openSUSE Linux distribution releases from early 2016 to their latest releases as of right now with their stable updates.

Complementing the other 2018 Linux benchmark comparisons published so far and more being wrapped up in the coming days, this comparison shows the Linux distribution performance on the same hardware from their state in early 2016 to that at the end of 2018. With having the same exact system used in one of the early 2016 Linux OS comparisons, the system was reloaded with the very latest versions of all the distributions tested to see how the performance has changed.

That system is still running strong is the Intel Core i7 5775C with Iris Pro 6200 graphics onboard, MSI Z97-G45 GAMING motherboard, 16GB of DDR4 RAM, and 120GB CT120BX100SSD1 Crucial BX100 SATA 3.0 SSD. That system is still running strong and racked up in the same configuration these years later. The Intel Core i7 5775C remains a strong and capable CPU particularly with its Iris Pro graphics and having 4 cores / 8 threads, 3.3GHz base frequency, and 3.7GHz turbo frequency. That CPU was running at stock speeds throughout all of the testing back in 2016 and now. The only change to point out is utilizing the latest motherboard BIOS now, which is necessitated for security reasons but otherwise is configured in the same manner as it was back in 2016.

Here is the rundown of the operating systems tested then and now on this Intel Core i7 system:

CentOS - CentOS 7 remains the latest version still though in this time there has been back-ports to its Linux 3.10 kernel and now being up to version 7.6. There is still the same GCC 4.8.5 compiler, a move from GNOME Shell 3.14 to 3.28, and XFS remains the default file-system in use. Also in 2016, CentOS defaulted to the CFQ I/O scheduler on this system but now defaults to the deadline scheduler. At least in 2019 we will see EL8 / CentOS 8 for a modern enterprise Linux stack.

Clear Linux - Intel's own Clear Linux platform has seen many upgrades in this time in moving from Linux 4.4 and GCC 5.3 to Linux 4.19 and GCC 8.2. There has also been countless other upgrades as well as different decisions made like migrating from the CPUFreq scaling driver to P-State, which now puts it in line with the other Linux distributions, though Clear still defaults to the "performance" governor. Clear Linux also is the only distribution tested now defaulting to MQ-DEADLINE where as in 2016 on this hardware was using CFQ.

Fedora - Going from Fedora 23 to Fedora 29 with upgrade meant going from Linux 4.3 and GCC 5.3.1 to Linux 4.19 and GCC 8.2.1 along with a plethora of other package updates. Fedora continues with the CFQ I/O scheduler on this hardware.

Ubuntu - Moving from Ubuntu 16.04 with Linux 4.4 to Ubuntu 18.10 with Linux 4.18 is another significant upgrade. Back then Ubuntu was also defaulting to the deadline I/O scheduler but now using CFQ.

openSUSE - The openSUSE testing went from openSUSE 42.1 Leap on Linux 4.1 to openSUSE Tumbleweed with Linux 4.19 and like the other distributions besides CentOS 7 saw many package updates. OpenSUSE also transitioned in this time from deadline to CFQ for its I/O scheduler on the SATA 3.0 SSD.

All of the 2018 Linux distributions are also mitigated against Spectre and Meltdown vulnerabilities, which wasn't a thing back in 2016. The other operating system settings/packages all were at their defaults. More details in the table below. Keep in mind the hardware was kept the same and with its default settings; any reported changes there just come down to how the hardware/information was exposed to the OS in a particular distribution.

The Phoronix Test Suite with its per-test-profile versioning makes doing this 2016 against 2018 comparison on the same hardware quite feasible thanks to being able to reproduce with the same exact application/package versions of the software under test and with the same test installation/run scripts in a standardized and reproducible manner.