How Linux Performance Changed In 2017 With Clear Linux & Ubuntu
Written by Michael Larabel in Operating Systems on 28 December 2017. Page 1 of 5. 25 Comments

The latest in our streak of year-end benchmarking is seeing how Linux performance has evolved over the course of 2017. For that we tested Intel's performance-optimized Clear Linux distribution as well as Ubuntu using releases from the start of the year to their current state for seeing how the performance compares using the same system.

For the Ubuntu comparison with it releasing only every six months and not offering any alternative model, the testing came down to Ubuntu 16.10 versus a recent daily snapshot of Ubuntu 18.04. Granted, this puts the comparison beyond just one year (2017) especially with Ubuntu 16.10 having been under feature freeze well before its October 2016 release, but still an interesting comparison with what's possible for Ubuntu. This means going from Linux 4.8 to Linux 4.14.9 (Ubuntu 18.04 currently still uses 17.10's Linux 4.13 kernel, but will ship with 4.15 next April, for testing I loaded it with the latest stable Linux 4.14 from the Ubuntu Mainline Kernel PPA to provider a fresher look) and going from GCC 6.2.0 to GCC 7.2.0 as the default system compiler. EXT4 remains the default Ubuntu file-system and uses no I/O scheduler with the NVMe SSD used during testing.

Clear Linux was the other distribution used for our 2017 Linux performance comparison. With Clear's rolling-release nature but still permitting access to older builds of the OS packages, it makes for convenient testing of start-of-2017 to end-of-2017 testing. When rolling back Clear Linux, build 12410 was used that was their last OS update of 2016. Clear Linux 19950 was then tested as the newest update as of testing. Going from Clear 12410 to 19950 meant going from Linux 4.8.12 to Linux 4.14.7 and on the system compiler side going from GCC 6.3.0 to GCC 7.2.1. Also changed by Clear Linux over the course of 2017 was going from using the CPUFreq scaling driver unconditionally to now using P-State: both times using the performance governor, now that P-State has matured a lot on Linux for better CPU frequency scaling on Linux with modern Intel CPUs. Interestingly, Clear Linux also now uses the Kyber I/O scheduler even with the NVMe SSD, rather than no I/O scheduler as other Linux distributions generally are using for NVMe storage and was the case with the older EOY2016 Clear release.

Each operating system release was tested with a variety of benchmarks carried out in a fully-automated and standardized manner using the Phoronix Test Suite.

The same system was used throughout testing with the Intel Core i7 7740X CPU (running at stock speeds, reported differences on the system table just come down to how turbo frequency reporting is handled via P-State/CPUFreq), Gigabyte X299 AORUS Gaming CF motherboard, and 8GB of DDR4-3000MHz memory with a 120GB Corsair Force MP500 NVMe SSD.

Related Articles
Trending Linux News