Well well, the day has finally come. It is now day number 50; the final day for this ATI Linux trial. With that said, it's about time for me to share some concluding remarks regarding ATI's recent Linux efforts with their monthly fglrx display drivers. On the first day, I didn't know whether I would be able to survive without going green (no pun intended), but towards the end of this experience, it was almost like achieving nirvana.
The trial revolved around me, Michael Larabel, exclusively trying out ATI graphics in my personal system as a replacement to the previous NVIDIA GeForce 7800GTX for 50 days. The main inspiration for this trial was to give the red side a chance to show me how they have progressed on the GNU/Linux front (outside of the numerous Phoronix tests). There isn't much disputing that in the past ATI has been known for their poor support -- poor control panel/configuration utility, lack of x86_64 support, and a host of other problems. However, ATI Technologies has truly performed a miracle turn-around with their Linux drivers, and a majority of these problems have been addressed in recent drivers. There does, however, remain the entire matter of the fglrx frame-rate performance.
The trial had begun on a negative note through a few problems with Fedora Core 5 (x86_64 SMP) on a dual LCD setup. The Fedora Core 5 problems were later fixed in the 8.26.18 drivers. ATI's Big Desktop is an excellent choice (as an alternative to using two X servers) for those using dual display heads. While ATI had not implemented Livna-like changes as planned with last month's 8.26.18 drivers, the installer does work accordingly on FC5 when using the automatic installation method. The Fedora Core packaging support will come in this month's 8.27.x driver release.
While discussing this trial with a colleague that was not familiar with the quality of ATI's Linux drivers (other then the benchmarks provided by Phoronix) he immediately classified ATI Technologies as attempting to fine-tune a hull on a ship while there is a giant hole in the side. However, is this truly the case? In my professional opinion, no, that is not entirely the case. In that analogy, the leaking hull would be ATI's current driver performance (or a few months back, the lack of Radeon X1000 support), while the patches are such things as logo.xbm/logo_mask.xbm and other odd ends. Outside of gaming, the drivers have become near-perfect. If you wanted to continue with that boating example, I would classify ATI's drivers as a broken bilge pump -- the ship is taking on some water, but as long as the issue is addressed in a timely matter, the ship can certainly be saved.
While the performance is one of the most important areas to avid gamers, ATI has taken great care in providing drivers that offer an arsenal of features for mobile and desktop users. Amongst the resources available to mobile users include PowerPlay, dynamic display management options, hotkey functionality (for Lenovo ThinkPad users), and the recent events daemon. The image quality for the ATI Radeon X1000 series on Linux has been phenomenal thus far. ATI Linux is also great for movie playback, and like tasks, especially with the current Avivo support and what will likely come down in the future. For non-X1000 users, the TexturedVideo option can be enabled manually from the xorg.conf.
However, if ATI wishes to remain competitive on Linux, they will certainly need to improve their performance. I have seen a few frames/second of improvement here and there due to indirect changes in the consequent release, but if ATI wishes to deliver a serious blow to NVIDIA, they must improve their fglrx frame-rate performance as it is the area where most gamers are concerned. When communicating with ATI on the matter, the stance has simply been that the frame-rate will improve with time. In my mind, based upon the growing number of people trying out Linux for their first time, and the number of Linux-native games increasing as well as the Cedega/WINE alternatives, the ATI Linux gaming performance is one of the leading problems with the fglrx drivers.
Could the ATI drivers also be adversely affecting the reputation of GNU/Linux? For instance, a computer gamer with ATI graphics is turned onto the open-source world and Linux, and ultimately gives it a shot, and with mediocre performance could his thoughts of GNU/Linux be indirectly tarnished? On the other hand, could ATI gamers resist trying out Linux once knowing the degraded performance they would likely face?
There is also the entire matter of ATI CrossFire support under Linux. NVIDIA Corporation had delivered its Linux SLI support 13 months after the Windows support. To this date, the quality and performance of Scalable Link Interface under GNU/Linux is almost nonexistent. There are barely any benefits to speak of when strapping in multiple GPUs and running them in SLI. In fact, there likely would not have been any Linux SLI support at all unless Hewlett-Packard had not helped. Frankly after seeing the SLI benefits under Linux being slaughtered, I do not miss the CrossFire support, and their time could be better spent by first off improving the driver performance.
Also among the features that have never made it into the fglrx drivers is support for overclocking. The Windows Catalyst suite offers ATI Overdrive for maximizing the visual performance through overclocking. In the 8.26.18 drivers, however, I was appalled to see that in the 8.26.18 drivers for the desktop X1800XT component listed with PowerPlay was a state for overdrive. Similarly, the Linux drivers do not support monitoring the onboard thermal sensors on capable cards. I am only left to wonder at this point whether someday in the near future those Linux users may see this overclocking support; similar to NVIDIA's CoolBits Linux implementation last year. Roderick Colenbrander has done a marvelous job for the NVIDIA Linux community with his work on NVClock. However, to this point there also remains no definitive third party ATI Linux utility. Rovclock is available, but that is primarily limited to R200 components.
One of the community benefits of using ATI Linux is the actively maintained Bugzilla. While the bug list is maintained by users, and is not officially supported by ATI Technologies, its developers do in fact stay tuned to the Bugzilla. Currently no NVIDIA Linux Bugzilla exists for public usage.
One of the other problems that some GNU/Linux users have ranted about has been the breaking of R200 support with the recent drivers. This was originally an OpenGL API Entry Point Issue, but R200 support was coincidentally removed from Windows Catalyst the following month. Even with this problem with the proprietary drivers, the open-source Radeon R200 drivers for Linux remain an excellent option.
Another problem that is an issue for both ATI and NVIDIA with their proprietary Linux drivers is the current lack of X.Org v7.1 support. X11R7.1 was released on May 22, but up to this point, neither ATI nor NVIDIA has yet to officially support this version. NVIDIA has continued to be closed-lip about their release cycle (the initial 1.0-9XXX Linux release will contain this support), but we can expect to see their next driver release to come in August or September. Meanwhile, ATI Technologies has not publicly stated when their X.Org v7.1 support will come, but I would expect the support as soon as this month's 8.27 release. However, ATI's initial X11R7.1 support will likely not contain support for the GLX_EXT_texture_from_pixmap extension.
To address user concerns about the ATI Linux Control Panel, their current graphical Control Panel (fireglcontrolpanel) is certainly lacking when compared to nvidia-xsettings. The Control Panel displays a bit of information, configuration for dual screens, and some minor adjustments. On the contrary, the text-based aticonfig is loaded to the brim in ATI-specific options. The aticonfig utility allows the automatic setup of the xorg.conf for the fglrx drivers: specifying AASF options, dynamic display management, and more. For what it's worth, the ATI Control Panel is open-source.
One of the areas I have thoroughly enjoyed about ATI's Linux drivers is their monthly driver releases. As the code base is shared between the Windows and Linux drivers, since last year the fglrx development team has been pushing out releases to accompany the monthly Windows Catalyst suite. While some monthly releases have had minimal changes, others have been hefty when it comes to the release notes.
Another benefit of ATI Linux is its installer. While NVIDIA continues to use its text-based installation mechanism, ATI does have a nice graphical installer. The installer had premiered with the 8.14.13 drivers, and it has continued to improve since. ATI's Linux department does also work closely with distribution vendors and volunteers for bundling packaging scripts for specific distributions. Similarly, ATI does have an open invitation for distribution vendors to participate in their Beta program.
With XGI Technology now seemingly out of the picture (after their GPL-ization of their Linux drivers seemed to have been canned), they are focusing on the niche market of servers and embedded clients. This of course all came after ATI Technologies had acquired Macrosynergy (an XGI alliance company) earlier this year. Meanwhile, NVIDIA does remain the popular option for Linux users, but ATI is quickly approaching the batting plate. NVIDIA does have the advantage of providing official support for Solaris and FreeBSD.
To sum up everything, ATI Linux is shaping up quite nicely except when it comes to their current level of performance. Will ATI remain in my system? Yes, ATI Radeon products have worked their way into both my desktop and notebook. These components will likely stay in there for some time unless the tides decide to turn. I am confident that in the coming months and the year ahead that ATI will be able to shine in the GNU/Linux spotlight, and that their drivers will continue to improve even more. Do not get me wrong, neither ATI nor NVIDIA is perfect, and in a perfect world, we would not have to deal with intellectual property preventing the drivers from being open-sourced. Who will ultimately win the race in the end? I do not honestly know. Even though the blog has reached the end of the road, I will continue to push along additional ATI Linux information at Phoronix, on the Phoronix Forums, and at additional places. Why not continue this blog? Well, a lot has been discussed over 50 days, and to prevent boring anyone, it is about time to move on. Some additional GNU/Linux crusades are also being worked on at this point. It was a great run, and thanks for reading.