When publishing the OpenGL performance results yesterday showing Windows 8 generally leading with a performance advantage over Ubuntu Linux, there was the usual large portion of the Linux community in disbelief. For proving a point, here are now results showing the Windows 8 Intel OpenGL performance compared to Ubuntu Linux when testing the KDE and Xfce desktops.
To no real surprise, many complaints poured in after yesterday's test results showing Windows 8 leading in some benchmarks over Linux. It's just the usual Linux users suffering from the common case of five stages of benchmark loss. Many of them were quick to say the results are invalid since testing was done with Unity, rather than some lighter weight desktop. Well, the fact of the matter is, Unity desktop is the default desktop within Ubuntu Linux, which in turn is the most common desktop Linux distribution in the world. This is what's relevant to the most number of Linux users.
As I've written dozens (or even hundreds) of times now after writing thousands of articles on Phoronix for nearly the past decade, benchmarking in the stock/default configuration is preferred since that is what's real-world applicable to the largest audience. With open-source software, all levels of the operating system can be tweaked to an unlimited extent. Benchmarks of an operating system configuration in a highly specific and highly tuned state aren't really relevant to most Linux desktop users.
The other special case is that at Phoronix I always continue to be happy to provide a separate test run in a "highly configured" case should the upstream project provide clear and concise documentation about their recommended tunables and other options for achieving maximum performance. However, over the years of calling out these projects to do so, there's been little documentation and not many projects accepting the challenge to improve their public performance/optimization documentation or Wiki pages. Without such documentation, many end-users are unaware of the optional configuration without endless hours of tuning and testing. Instead of making real progress, people prefer to just burn the messenger. The focus of most Phoronix benchmarks is to deliver results that are meaningful to the largest possible population of Linux users and can be independently reproduced.
Anyhow, aside from using Unity in yesterday's testing because it's the default desktop to Ubuntu Linux, it was also used since earlier Phoronix testing of Ubuntu 13.04 "Raring Ringtail" has shown that many of the performance challenges previously attributed to the Unity desktop and Compiz window manager have been addressed. In past Ubuntu Linux releases, the Unity performance has generally been poor, but as shown in recent months, the Unity/Compiz performance has improved a lot and is comparable to the other desktop environments.
To address the criticism of those feeling Unity OpenGL performance results are invalid, new tests were conducted from the ASUS S56CA-WH31 Ultrabook with KDE and Xfce. Specifically:
- KDE from Ubuntu 13.04 while changing the default KWin option to suspend desktop effects on full-screen windows.
- Xfce 4.10 from Ubuntu 13.04 in a stock configuration.
All testing was from the same Ubuntu installation as used earlier with the updated Linux 3.9 kernel and Mesa 9.2-devel while from the same Intel Core i3 3217U ultrabook.