Lately we have talked a lot about the Intel Linux driver stack with their ongoing work of switching over to the Graphics Execution Manager for memory management, moving to kernel-based mode-setting, and migrating to the Direct Rendering Infrastructure 2. In the short term, this work has caused some nasty problems, but once the fallout has been addressed, the open-source Intel driver should be in a prime position to perform on all fronts.
Earlier this month Intel had pushed out the xf86-video-intel 2.7 graphics driver that brought a variety of new features and fixes. This 2.7 X.Org driver has been rolled into Intel's "2009Q1 Package" along with Mesa 7.4 plus a few more patches, the Linux 18.104.22.168 kernel plus a few more patches, and libdrm 2.4.9. It is also suggested that X Server 1.6.1 be used with this recommended Intel software configuration. The details regarding the 2009Q1 package can be found here.
As I alluded to earlier, I am out of the office this week. With me to Italy I took a Samsung NC10 that is loaded with an Intel Atom processor, Intel integrated graphics, an OCZ solid-state drive, and 2GB of DDR2 RAM. Prior to leaving I loaded it up with a clean install of the Ubuntu 9.04 release candidate, but what a mistake that was. As was pointed out in the days prior to that, there are a number of regressions present within the Intel Linux graphics stack due to the switch to the Graphics Execution Manager, DRI2, and other work. While the benchmarks showed there is indeed a drop (a large drop in some), this performance drop nearly renders the system useless.
Last night Intel pushed out the xf86-video-intel 2.7.0 driver, but still it does not correct some of the performance regressions that have been occurring within the Intel Linux graphics stack for the past few months. While there are UXA improvements in the 2.7 driver, the performance is still bad.
The day started out for us by looking at the performance impact of using greedy migration heuristics with Intel graphics as well as taking another look at their driver's UXA performance. After that article was published, we had benchmarks from the Linux 2.6.30-rc2 kernel that showed 2D/3D Intel performance improvements. The third Intel item today was then Intel announcing the release of the G45 hardware documentation that was four volumes and more than 1,000 pages of technical documentations. Well, the fourth and final item tonight is the release of the xf86-video-intel 2.7.0 driver.
Since earlier today when looking at whether greedy migration heuristics help in improving Intel's 2D driver performance and how it compares to using UXA acceleration, we have run a couple more benchmarks. This time Canonical's Bryce Harrington was interested whether a new kernel improves Intel's performance.
We are now into the second quarter of 2009, but Intel will soon release its X.Org driver for Q1'09. While late, the xf86-video-intel 2.7.0 driver update SDVO-TV support, basic SDVO-LVDS support, tear-free X-Video support, and many bugs have been fixed.
While it already seems like the Intel Atom processor family has been around for quite a while, it's really only been one year. Today Intel is celebrating the one year anniversary of the Atom processor from the Intel Developer Forum in Beijing.
In a rather interesting move, the Linux Foundation will now be the company behind Moblin. The Moblin project was started by Intel back in 2007 and has been used to push Linux onto their Atom processors, but it now appears they are handing the duties off to the Linux Foundation.
Back in January we shared that Intel's Poulsbo driver was a bloody mess. This Intel GMA 500 Chipset is used by a few select devices at the moment and its core is originally based upon a PowerVR ASIC. When it comes to the Linux support, the xf86-video-psb driver is ill maintained (it hasn't been updated in over a year), the 3D driver is binary-only, and the driver is in a badly broken state. The Intel Poulsbo driver and its Linux support is in a completely opposite state of where the xf86-video-intel driver is at and the Linux support for standard Intel IGPs. Today though there's a new part to the Poulsbo Linux story.
It has been nearly three months since the release of xf86-video-intel 2.6, which means it's almost time for their next open-source driver update to arrive (they abide by a quarterly release cycle). As a sign of this, Intel's Carl Worth is heading the xf86-video-intel 2.7 driver management and he has pushed out the first release candidate.
Just over a week ago Intel had come out with the xf86-video-intel 2.6.2 driver, but today that is already being replaced with xf86-video-intel 2.6.3. The 2.6.2 release was significant in that it presented some major fixes to the Graphics Execution Manager and kernel mode-setting. Along with that, Textured Video support when using kernel mode-setting was introduced along with rotation support.
We have already burned through more than half of Q1'09 and have yet to see Intel's quarterly X.Org Linux driver update approaching (or even any release candidate, although the 2.6 release was late). While xf86-video-intel 2.7.0 isn't ready yet, a fairly significant bug-fix release has been issued in the 2.6 series.
If you have been wanting to look at Linux performance numbers for the Intel Core i7 or AMD Phenom II processors, be sure to stop by Phoronix Global. While the PR departments at Intel and AMD seem to have no interest at all in their latest desktop processors on Linux, many users of this new hardware have provided their own benchmarks courtesy of the Phoronix Test Suite.
In addition to learning that Intel's UXA acceleration architecture will live on in its current form, there were a few other interesting bits of news reported as well during the X.Org talks at FOSDEM 2009. During Eric Anholt's talk some video playback/decoding improvements were discussed.
Back in August we talked about the UXA acceleration architecture, which was developed by Intel and based upon EXA but pixmaps were dealt with as GEM objects.
Most of the Intel Atom netbooks currently on the market (like the Lenovo IdeaPad S10, Samsung NC10, and ASUS Eee PC) use the GMA 950 Chipset for their integrated graphics, but some of the newer models are using the Intel GMA 500. The GMA 500 doesn't share many traits with other mobile Intel IGPs since much of the technology was licensed from PowerVR, which means a different X.Org display driver is required.
Intel's Eric Anholt has announced the immediate release of the xf86-video-intel 2.6.1 driver. This update coming just a week after the xf86-video-intel 2.6.0 release fixes a stability issue and kernel mode-setting fixes, More fixes are expected to land in the open-source Intel X.Org driver in the near future.
Intel generally updates their xf86-video-intel X.Org driver on a quarterly basis, but the 2.6 cycle has been running a bit behind schedule. Nevertheless, late last night the xf86-video-intel 2.6.0 driver was officially released.
Intel's Zhenyu Wang has announced the release of the xf86-video-intel 22.214.171.124 (2.6 RC2) driver. This latest open-source Intel driver depends upon libdrm 2.4.3 while introducing improvements for kernel mode-setting, UXA, and various other fixes. This driver will also now automatically build with support for kernel mode-setting if the X Server is new enough.
Intel's Zhenyu Wang is managing the release cycle for the xf86-video-intel 2.6 graphics driver and this evening he has announced the availability of its first release candidate. The Intel 2.6 X.Org driver brings DRI2 support, XvMC for the Intel GMA 4500 series, kernel mode-setting, RandR 1.3, UXA acceleration fixes, and HDMI audio support.
We have yet to deliver any Linux benchmarks of the Intel Core i7 processors, but there's a few Phoronix resources if you're interested in learning more about the Nehalem architecture.
Times are already exciting within the X.Org development community thanks to the Graphics Execution Manager entering Linux 2.6.28 and the release of X Server 1.6 quickly approaching, but still there's more good news to report this week. The Direct Rendering Infrastructure 2 support has entered the mainline xf86-video-intel driver and DRM mode-setting (commonly referred to as kernel mode-setting) has appeared in the drm-intel-next branch.
This week NVIDIA released their 177.82 Linux driver and AMD pushed out their Catalyst 8.11 Linux driver. Intel has now joined the party and pushed out not one but two new driver releases. Intel's Jesse Barnes has announced the release of the xf86-video-intel 2.4.3 and 2.5.1 DDX drivers.
Jesse Barnes, one of the developers within Intel's Open-Source Technology Center whose responsibilities largely deal with Intel's Linux graphics work, has shared some thoughts and statistics on past graphics-related releases. Jesse looked at the number of DRM commits to recent Linux kernel releases, a few thoughts on Mesa and xf86-video-intel releases, and the release history of the X Server.
Initially the ATI R600 GPUs and later with the available open-source drivers (xf86-video-ati and xf86-video-radeonhd) hadn't supported using the graphics card's integrated audio processor with HDMI output. An end-user though dissatisfied with this had went on to reverse-engineer the support for RadeonHD HDMI audio support. The audio patch was then merged to master within the last two weeks that now enable ATI HDMI audio output support with that open-source driver.
Last week was a major milestone for the Intel X.Org developers with GEM entering the mainline kernel and the xf86-video-intel 2.5.0 driver then subsequently being released on Monday. While these developers have accomplished a lot, there is still a lot of work ahead. Intel's Eric Anholt has commented on his blog about some of the recent work and what's still to come in the near future.
This weekend the xf86-video-intel 2.4.98 driver was released with hopes of releasing the Intel 2.5.0 driver on Monday. This driver is arriving a bit late compared to its original target of last month, but it's now available. This open-source Intel X.Org driver update features GEM integration (on supported systems), kernel mode-setting (on supported systems), improved EXA 2D performance, and quite a number of bug-fixes.
The xf86-video-intel 2.4.97 driver was released back in August as the first test-release for the 2.5.0 driver series. Some of the proposed changes for this quarterly update to Intel's open-source X.Org graphics driver include usable EXA support, GEM integration, kernel mode-setting support, video tearing fixes, and plenty of bug-fixes. It's been almost two months since this test driver was released, but Jesse Barnes has now pushed out the second and final test release for xf86-video-intel 2.5.
Since May when Intel first introduced their Graphics Execution Manager after X.Org developers were enraged over Tungsten's TTM memory manager and its development challenges, we've been talking about GEM several times since then.
998 Intel news articles published on Phoronix.