There have been a flurry of comments this week following my post why software defaults are important and why in the Linux benchmarks at Phoronix.com the tests are most often carried out in their default/stock configurations: it's what most everyone uses. There have been comments by Ted Ts'o on file-system default mount options and whether they are sane or not in the non-enterprise distributions and others have questioned if defaults like Compiz on in Ubuntu by default makes sense. Does using Compiz still hurt your graphics performance?
Some comments to my "software defaults are important, so test the software at its defaults" posting have been along the lines of: "As for default config as a baseline, I'm all for it. I have nothing against testing the default. But I personally think that Ubuntu has made some bad decisions -- like using Compiz, which cannot suspend compositing on the fly AFAIK -- which harms performance on free drivers disproportionately." Is that though still the case or is just FUD continuing to be spread?
In May of last year, I published an article entitled The Cost Of Running Compiz that looked at the performance impact of running this compositing window manager on Intel, open-source Radeon, Catalyst, and NVIDIA setups. Those results from nearly one year ago showed that the open-source Intel and ATI Radeon drivers took an approximate 15% performance hit when Compiz was running rather than GNOME's Metacity when using Kernel mode-setting and DRI2. With the proprietary Catalyst driver there was little impact along with NVIDIA's binary driver, but there still was a performance penalty in some tests.
Since it's been nearly one year since those tests were carried out, and the Linux graphics drivers have advanced a great deal, particularly the open-source Gallium3D drivers, I have performed a new set of benchmarks to see if running Compiz with full-screen games at the server's native resolution still hurts the overall performance. The Compiz performance is also becoming more important since this compositing window manager is to play a critical role in Canonical's Unity Desktop for Ubuntu 11.04.
For this testing, Ubuntu 10.10 was used as the Linux distribution, but Compiz 0.9.4 was tested, which is the version currently used by Ubuntu 11.04. The latest open-source and closed-source graphics drivers were also used. The performance with various drivers and graphics cards were compared between running Compiz and Metacity with the GNOME 2.32 desktop.
The test system was an AMD Phenom 9500 quad-core CPU with an ECS A70-GXM-A motherboard, 4GB of DDR3 system memory, and 250GB Seagate SATA HDD. The graphics cards tested included an ATI Radeon X1950PRO, ATI Radeon HD 4670, ATI Radeon HD 5450, NVIDIA GeForce 8600GT, and NVIDIA GeForce GT 220. No Intel integrated graphics were tested in this article as those test systems were pre-occupied with other tests. The Radeon Gallium3D and Catalyst drivers were tested on the ATI side and the Nouveau Gallium3D and proprietary NVIDIA drivers were tested for the GeForce hardware.
Key packages installed on Ubuntu 10.10 x86_64 include the Linux 2.6.38 kernel (from 8 March 2011), the Linux 2.6.35 kernel when using the Catalyst driver (it's not yet 2.6.38 compatible), X.Org Server 1.9.0, Compiz 0.9.4, GNOME 2.32.0, GCC 4.4.5, NVIDIA 270.30 binary driver, Mesa 7.11-devel Gallium 0.4, Catalyst 11.2 / fglrx 8.82.8, xf86-video-nouveau Git, and xf86-video-ati 6.14.99 Git. The Git code-bases were obtained on 8 March 2011.
Via the Phoronix Test Suite, OpenArena, World of Padman, Nexuiz, and Urban Terror were tested under Compiz and Metacity with each of the graphics cards and drivers.