Fedora vs. Ubuntu vs. openSUSE vs. Clear Linux For Intel Steam Gaming Performance
With this week Clear Linux now being able to run Steam, I was excited to see how this performance-minded Linux distribution out of Intel's Open-Source Technology Center would compare to other more popular Linux distributions when it comes to Intel Linux gaming performance. Here are some benchmarks of this traditionally workstation/server-oriented Intel Linux distribution running some Steam Linux games.
For those that haven't read our past articles with Clear Linux benchmarks, it tends to perform very well for a variety of reasons. Delivering the best Linux performance possible on Intel hardware is one of their driving forces and include features like aggressive CFLAGS/CXXFLAGS by default, using CPUFreq over P-State for CPU frequency scaling, building software packages with function multi-versioning to support both older and newer processors, selectively choosing Clang vs. GCC compilers for each package based upon the performance, carrying some out-of-tree patches, and various other optimizations/tweaks as outlined at ClearLinux.org.
It took a while for Steam to work on Clear Linux due to the variety of 32-bit libraries depended upon by Steam. Unfortunately, only Intel graphics benchmarks for this comparison as that's all that is supported by Clear Linux at this time: the open-source Radeon and Nouveau drivers aren't even built on Clear Linux. If you haven't looked at our past Intel OpenGL benchmarks using Clear Linux, here's a sampling of the non-Steam Linux gaming performance:
The magnitude of the performance differences vary, but overall on a variety of Intel graphics hardware from multiple generations we routinely see Clear Linux at the top due to their quick updating of new driver releases paired with their fine-tuned optimizations.
Clear Linux is a rolling-release distribution and in its shape tested (v12760) it shipped with the Linux 4.9 kernel, Xfce 4.12 desktop, xf86-video-modesetting DDX, Mesa 17.0-devel, Vulkan, and X.Org Server 1.19.1. Fedora 25's Xfce spin was tested to avoid the overhead of (X)Wayland when testing Fedora 25 with GNOME as well as to more closely mirror the desktop configuration of Clear Linux. This Fedora 25 stock configuration with all stable updates had Linux 4.9., Xfce 4.12, xf86-video-modeseting, X.Org Server 1.19.0, and Mesa 13.0.3.
OpenSUSE Tumbleweed was tested too, this rolling-release distribution was on Linux 4.9, KDE Plasma 5, Mesa 13.0.3, X.Org Server 1.19.0, and was also the only distribution tested that used xf86-video-intel rather than xf86-video-modesetting as the default.
Ubuntu 16.10 was the last distribution tested with the stock Linux 4.8 kernel, Mesa 12.0.3, and X.Org Server 1.18.4. Additionally, a secondary run of Ubuntu 16.10 was done when using the Ubuntu Mainline Kernel PPA and Padoka PPA. This updated stack brought Ubuntu 16.10 to the Linux 4.10 kernel and Mesa 17.0-devel, providing the newest packages of any of the distributions tested albeit in an unofficial manner. Unless otherwise stated, all distributions were tested at their defaults in order to allow these tests to be reproducible and done in an out-of-the-box manner as seen by the majority of desktop Linux users.
I had attempted to test SteamOS Beta too, but it's current configuration didn't support the latest-generation Intel graphics and thus was only running on LLVMpipe with its default packages. Also, I had installed Antergos 17.1-Rolling on the system, but due to its library handling, my test scripts wouldn't work there with Steam.
The same system was used throughout testing and consisted of an Intel Core i5 7600K, ASUS PRIME Z270-P, 2 x 8GB DDR4-2400MHz memory, Samsung 950 PRO 256GB NVMe SSD, and Kabylake GT2 HD Graphics 630. All of these OpenGL and Vulkan Steam Linux benchmarks were done in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software.