“Open-source support has appeared by default in X.Org 7.1 for R300 generation GPUs. While ATI does not officially support these R300 open-source drivers, this alternative have been gaining momentum with users largely due to the lack of GLX_EXT_texture_from_pixmap support in fglrx. This extension is needed for the AIGLX desktop eye-candy. These drivers do lack TV-out support and many other features found within ATI’s fglrx drivers, but how does its performance compare?”
I for one congratulate the developers for coming so far by reverse engineering, and for working hard!
What you all seem to forget amid your bitching is that they ARE PEOPLE that are doing a VERY HARD THING and doing the best they can at it.
Please, respect a little more their work, and at least say “thanks, you’re on the right track”.
I for one thank them for keeping it up, and I might consider choosing ATI for my next purchase because of their drivers.
Thank you (and a few others I gave +1) for adding some actual enlightenment to the comments.
Reverse-engineering of hardware specs is a feat I could never hope to ever accomplish myself – the wizards who accomplish this are nothing less than amazing.
Rudolf C. from the BeOS/Haiku community was one of these such persons. He took the UtahGLX OSS 3d nvidia driver and turned it into an excellent work-in-progress hardware-3d driver for BeOS/Haiku.
These developers are currently the only hope for 3d hardware acceleration on alternative OSes with no commercial hardware support. Their work helps everyone.
If I were to choose between reliability or performance for my day-to-day computing, it would be reliability almost every time.
While the idea of a Free alternative to ATI’s drivers is a good idea, in some cases the drivers had 1/4 of the frame rate seen with ATI’s drivers. For most people I can’t imagine it’s worth it.
I can understand why ATI does not want to GPL their Linux drivers, but maybe a good middle ground would be to look at the OSS drivers out there and fix some of the bottlenecks where it wouldn’t reveal proprietary information.
Not everything used in ATI’s Drivers belongs to ATI, there are some things that are licensed to ATI that they cannot release as opensource.
you’ re rigth, but i hope ati can find a way to relase driver open without the licensed thing.
gpl is not the only model, bsd and cddl are open too and can fit the ati problem better.
“One of the other benefits from these drivers come in the form of supporting AIGLX with the GLX_EXT_texture_from_pixmap extension (not to mention the psychological benefits of being able to get by without using binary-only drivers).”
I could care less if I’m using open source drivers or not. What I want in drivers (of any kind) is performance and reliability. When there are differences in performance of 44fps to 131fps it’s no contest.
No, this guy is probably not on crack.
It is arrogant and ignorant to believe your own preferences are the only reasonable ones.
I also work on a PC with an ati r300 card in it. The proprietary driver sucks. I can’t change the modeline with xvidtune, and the default one makes my monitor do a high frequency noise. GLX_EXT_texture_from_pixmap is missing, as the author pointed out. Suspend to disk doesn’t work right.
It seems it is a big mess and the *only* thing working better would be OpenGL accelerated games. Even benchmarking the XRender speed revealed there was no big difference (in 2D accel. performance, that is). As I don’t want to play games, but instead to get work done, you’re right, there is no contest, as the proprietary driver just sucks and is kind of unusable for me.
As a consumer savvy enough to understand the issues, it pays to balance the factors against each other. Open drivers give you certain assurances and conveniences. Performance is also a factor… *if* you play games or do other things that require 3D. Otherwise, the performace differences count for zip. Reliability is quite significant.
Your mileage may vary.
I’m quite happy with the open drivers on my notebook with an ATI Express 200M. The performace in glxgears is about half what the fglrx drivers can do, but I don’t care. It’s less hassle not having to muck with the proprietary drivers.
I do use the proprietary NVidia drivers on my desktop machine.
I’d say more, but I think I’ll go back to trying to find a way out of my current level in ‘Rune’. 😉
I could care less if I’m using open source drivers or not. What I want in drivers (of any kind) is performance and reliability. When there are differences in performance of 44fps to 131fps it’s no contest.
Well, the problem is that the benchmark regarded the only closed source drivers advantage: 3D performance. Search for fglrx drivers on Google, or on the most important linux forums or mailing lists, you’ll get an infinite list of complains. If my memory does not fail me in the last years the closed source drivers released by ATI broke (or are still breaking) a number of things, including switching to a console from X and switching back. To not mention that ATI notoriously release the drivers seldom and untimely. This was not a problem some years ago, but now the X.org development has a different pace, and ATI users are forced to lag back for months with outdated software.
There’s also no contest when your ATI hardware is part of a PPC-based iBook and you want to run some version of Linux. In that case, performance for the binary drivers is a whopping 0 fps for everything.
I for one am very glad there are open source ATI drivers, and I hope development on them continues. Improvements in their performance is just gravy.
from wikipedia:
While the R100 and R200-series chipset drivers were written using specifications provided by ATI (r200 driver), the R300-R500 hardware acceleration was written through reverse engineering (r300 driver) the methods used by ATI’s proprietary driver. The reverse-engineered code is now in X.Org and Mesa, bringing experimental support for some of the current Radeon cards. All r3xx cards and all r4xx excluding the Xpress integrated chips should be supported by the new experimental r300 driver. With the 2.6.17 kernel, some cards from the r300 driver were added to the radeon driver
The incomplete and buggy r300 drivers dont even remotely compare to ATI’s latest proprietary drivers.
I managed to get DeusEx to play with r300 at an ok framerate but there were still pixel artifacts all over textures and models.
Nvidia users have no choice. The “nv” driver is very basic with no 2D/3D acceleration features whatsoever.
Now I understand the desire for open-source drivers but ATI/AMD and Nvidia arnt interested at the moment.
Maybe we should just be grateful that we even have drivers and leave the politics behind until Linux gains significant enough marketshare to push these companies around.
Edited 2006-10-06 15:36
“Nvidia users have no choice. The “nv” driver is very basic with no 2D/3D acceleration features whatsoever.”
Fortunately, the closed driver is excellent.
“The “nv” driver is very basic with no 2D/3D acceleration features whatsoever. ”
That’s actually not true, for 2D acceleration that is. And really, for many if not most things that folks think 3D is important, it’s actually just accelerated 2D that matters, not 3D.
http://xorg.freedesktop.org/releases/X11R7.0/doc/html/nv.4.html
Accelerated 3D on linux will get you the GL-based screensavers, and some select games (most use 2D however, not 3D.), not much more than that.
(well, ok if you really want to go the xgl/aiglx route I guess that’d be the case, but again, that’s mainly “wow cool” eye candy more than anything else (and there’s nothing wrong with that…))
Except NVidia’s proprietary drivers Just Work, and they are at least as fast as their Windows versions too. I believe that’s because of the same codebase, but whatever. I’m sticking to nVidia for my Linux/FreeBSD gaming. ATI products just suck too much.
Except NVidia’s proprietary drivers Just Work, and they are at least as fast as their Windows versions too.
Until you upgrade the kernel, anyway. And they still don’t support AIGLX unless you go to a beta version. Still, if all you care about is performance then going proprietary NVidia is the obvious route to take.
Until you upgrade the kernel, anyway
That’s not NVidia’s fault, that’s the kernel’s fault for not having a stable API.
And they still don’t support AIGLX unless you go to a beta version. Still, if all you care about is performance then going proprietary NVidia is the obvious route to take.
Oh, nice jab there. Soon that Beta driver will be the stable one so that point will be moot too.
The only reason to not use nVidia drivers is because you have some emotionally-based objection to them. Otherwise, any practical person that can use them will use them.
The only reason to not use nVidia drivers is because you have some emotionally-based objection to them. Otherwise, any practical person that can use them will use them.
Well, nVidia hasn’t published their drivers for ppc yet. We ppc linux users do not have an “emotional” problem with nVidia, we have a “pratical” problem.
Well, nVidia hasn’t published their drivers for ppc yet. We ppc linux users do not have an “emotional” problem with nVidia, we have a “pratical” problem.
Can you name *any* 3D video card manufacturer that makes drivers for a PowerPC based operating system?
Sounds more like an industry and PPC problem than an nVidia problem to me.
Besides, I’m only speaking of the groups of people that have an option of using the drivers and can’t. Not hardware that nVidia never advertised as supporting.
That’s not NVidia’s fault, that’s the kernel’s fault for not having a stable API.
OK, it’s obvious you are completely missing the point. It doesn’t matter whose fault it is. The point is, the problem exists. Even more importantly, it is unlikely to be fixed in the conceivable future.
Oh, nice jab there. Soon that Beta driver will be the stable one so that point will be moot too.
It wasn’t meant as a jab, just another example of why the NVidia drivers aren’t the perfect piece of software certain people seem to think they are. Trying to fit the proprietary drivers onto a constantly changing open source kernel/x.org is going to have additional problems that open source drivers won’t.
That’s not to say the proprietary drivers don’t have their uses. They are clearly superior in almost every respect and most people who care about performance or advanced features are going to need them.
The only reason to not use nVidia drivers is because you have some emotionally-based objection to them. Otherwise, any practical person that can use them will use them.
Here’s one – I have an ATI card and don’t want to spend $100 bucks just so I can use NVidia’s drivers. There’s also not any point. The only 3d stuff I do in linux is the Compiz stuff, which runs just as well on the open source drivers. If I want to do any gaming I just reboot into windows where I have greater compatibility anyway. Please explain to me why I should go to the trouble of installing proprietary drivers that may occasionally cause upgrade problems when I already have drivers that work fine for me and I would never use the advanced features anyway.
I hear the same argument all the time from the OO sucks crowd. How can you stand using OO? It sucks and MSO is so much better… Blah, blah, blah. Guess what – OO isn’t as good, but it works well enough for me. If I ever need something better then I’ll look around (or just use my work computer), but until then why would I bother?
It should be no surprise that FOSS drivers are worse than the proprietary ones provided by the hardware vendor, especially for 3d acceleration. It’s a miracle that they exist at all, since they won’t release the specifications needed to write a driver:
http://marc.theaimsgroup.com/?l=linux-kernel&m=109831011607347&w=2
http://hardware.newsforge.com/article.pl?sid=06/06/16/2110211&from=…
http://en.wikipedia.org/wiki/Open_source_hardware
Choose one. That’s ATI’s status with Linux at the moment. The free drivers are mostly stable (and predictable in their instability in the cases where they aren’t), and the closed drivers are faster for gaming, but may crash the X server for no apparent reason. I prefer the free drivers, and just choose 1280×1024 instead of 1600×1200 when playing Quake 3. Easier upgrades is a benefit as well.
I reboot to Windows for the more demanding games, like Oblivion (which, incidentally, makes Windows crash quite often).
I’ve noticed several people here that I think are looking at this backwards by asking why they should switch to these free drivers. That isn’t the real question, because these drivers will be installed by default and Just Work. The real question is whether the proprietary drivers are so much better that it is worth the hassles caused by getting and using them. This article seems to suggest that they aren’t unless you need performance for playing games that were made post Q3. Like IE on Windows, there are superior alternatives available if you want to go to the trouble of getting them, but most people are quite fine with what MS gives them.
“These drivers do lack TV-out support[…]”
Do they?
http://bugs.gentoo.org/show_bug.cgi?id=127642
EDIT: also, see http://www.thinkwiki.org/wiki/Additional_options_for_the_radeon_dri…
and http://www.thinkwiki.org/wiki/How_to_make_use_of_Graphics_Chips_Pow…
Edited 2006-10-06 22:24
on 9700pro fglrx isnt faster.
The main difference is that the open driver will improve, where the closed driver will fade away. When there is no more hope to get money from older hardware that is completely beaten by the competition, why make drivers for it ? For Open drivers, this does not work like this, because the objectives are not the same.
But for ATI, releasing full specs for their older models, could be an added value to their products : the insurance that if you buy a new 3D card, there will be (closed) drivers available for it, then this card will be fully supported by open drivers. Is there something wrong in my reasoning ?