The performance of the upcoming Mesa 10.4 might be better out-of-the-box for R600g and RadeonSI Gallium3D driver users if a new patch is accepted to re-enable HyperZ by default.
HyperZ is an important performance-boosting feature that's been available in the open-source AMD Linux drivers for years but tends to often get flipped on/off every once in a while when bugs are reported about HyperZ causing corruption or stability problems by users. Currently in Mesa Git the HyperZ support is disabled by default in the R600g and RadeonSI drivers.
In celebration of Ubuntu 14.10's Utopic Unicorn release today, here's some fresh benchmarks of one of the most requested topics: 2D/3D benchmarks of different desktop environments. In this article is a look at six of the popular desktop offerings found in Ubuntu 14.10.
The desktops tested in their near-final state on Ubuntu 14.10 x86_64 included Unity 7.3.1, KDE 4.14.1, Xfce 4.10, LXDE 0.6.2, Openbox 3.5.2, and GNOME Shell 3.12.2. Tests are also being done of Kubuntu's PPA for Plasma 5 packages, but those results will be saved for its own article. Testing the MATE packages in Ubuntu 14.10 was also attempted but when logging into the MATE session it was endlessly spawning a bunch of new windows and just wasn't working right at least in the configuration attempted.
With Ubuntu 14.10 "Utopic Unicorn" due for release today, here's some benchmarks showing how the standard Unity 7 desktop on Ubuntu 14.10 is comparing to the still-experimental Unity System Compositor and using XMir for running traditional Linux OpenGL games.
From a standard Intel Core i7 Haswell system with HD Graphics I ran benchmarks with the development snapshot of Ubuntu Utopic as of yesterday to see how well the stock Unity 7.3.1 environment is comparing to when it's run with unity-system-compositor installed and using Mir support with XMir for running a variety of standard OpenGL benchmarks as well as some 2D X11 benchmarks.
Emil Velikov, the new Mesa release manager, has issued a straw-man proposal to release Mesa 10.4 in early December.
In sticking to the three month release cadence of Mesa, Emil is proposing the Mesa 10.4 feature freeze and release candidate for 14 November with new release candidates to come weekly until the official release. Emil is tentatively thinking about the Mesa 10.4 release for 5 December.
With the Btrfs file-system continuing to stabilize while still adding more functionality and is generating continued interest from more Linux distributions and other open-source projects, I've found it time to run some fresh Btrfs RAID benchmarks to see how the next-generation Linux file-system is performing with its built-in RAID handling.
The Btrfs file-system offers built-in support for RAID 0, RAID 1, RAID 5, RAID 6, and RAID 10 (1+0) modes though the support for RAID 5/6 is still considered most experimental. Besides RAID, a single Btrfs file-system can span multiple devices to yield a larger file-system. In this basic Btrfs RAID benchmarking from Ubuntu 14.10, I used two identical hard drives and tested Btrfs without RAID and then in RAID 0 and RAID 1 modes. For a separate article I'm also working on a RAID 0/1/5/6/10 comparison using for solid-state drives, which should be much more interesting, so just take this article as an introduction.
Given yesterday's story about Ubuntu 16.04 LTS potentially being the last 32-bit release if that proposal goes through, and given the number of people still running 32-bit Linux distributions on Intel/AMD hardware that is 64-bit capable, here's some fresh x86 vs. x86_64 benchmarks using Ubuntu 14.10.
Using Ubuntu 14.10 with the latest daily development ISOs today -- just ahead of this week's official Utopic Unicorn debut -- I did clean installs of both x86 and x86_64 flavors and carried out the same benchmarks with the same settings. Using the open-source Phoronix Test Suite automated benchmark framework made all testing easy and effortless.
For those anxious to see how well the GeForce GTX 970, NVIDIA's new high-end, Maxwell-based graphics card will perform under Linux, here's some preview benchmarks.
As outlined yesterday, the GTX 970 Linux benchmarking is currently happening following my GeForce GTX 980 Linux review from a few weeks ago. The GTX 970 Linux testing is going well and the full review will be published in the next few days.
In complementing this week's Linux review of the AMD Radeon R9 285 and follow-up articles with some extra GPU scaling tests and Catalyst AI Linux benchmarks, here's some more OpenCL R9 285 "Tonga" performance numbers under Ubuntu compared to what was shared in the original Linux review.
For those interested I ran some more OpenCL benchmarks of the Radeon R9 285 with the fglrx 14.30 series driver on Ubuntu 14.04 LTS with the Linux 3.15 kernel against an assortment of other NVIDIA/AMD graphics cards as used in these recent Linux hardware reviews.
Along with today's R9 285 GPU scaling tests from Ubuntu, other Linux graphics tests I ran from the AMD Radeon R9 285 GCN 1.2 graphics card is a check whether to see Catalyst AI is doing much on Linux.
In the past I've found the Catalyst AI feature exposed by AMD's Catalyst Linux driver to be next to useless: AMD's Catalyst A.I. Is Good For Few Linux Games and AMD Catalyst A.I. Useless Under Linux?. But with this latest Catalyst Linux driver (fglrx 14.30 series) and newest graphics card (R9 285) I decided to run the OpenGL tests again on Ubuntu 14.04 LTS.