Slashdot Mirror
ATI Updates Linux Drivers
GraWil writes "Famed graphics card maker and documented Linux supporter ATI has refreshed its proprietary Linux drivers (3.11.1) for the Radeon and FireGL series cards. Unfortunately, many of the previous comments still apply and it seems that ATI is not yet committed to supporting Linux well. The procedure for installing is now documented in a separate how-to but it seems that quite a few are stuck in an endless cycle of compiling kernels with/without DRI/AGPGART/RADEON/DBE (insert random module here). For those with strong enough feelings, ATI is seeking feedback on these drivers."
The article starts off: documented Linux supporter ATI. And then goes on to say: ATI is not yet committed to supporting Linux well.
So which is it?
When you get to hell -- tell 'em Itchy sent ya!
Maybe ATi just plain don't know how to make decent X/Linux drivers? A graphics card manufacturer like ATi would not traditionally hire people with relevent experience, and I doubt they can justify the expense of hiring a specialist to do nothing but create Linux drivers.
Of course, Open Source could help them here, but we all know the arguments for and against that.
I've been hoping that some of these companies would do similar to what Google did (before this TopCoder) thing and issue a bounty of sorts to get these done. Perhaps the winner/winning group could get the right to develop the *n?x driver and possibly have it made into a paid over time position of sorts. As long as they pay less than they would in house + paperwork it seems both parties would make out pretty well.
( o ) one could say I'm rather baked
If enough people leave the right kind of feedback, those drivers will be made open source.
There are just a few followers in management who think we need to follow NVidia's business model. They are wrong.
I am not the only one who is either thinking about or has already switched to Nvidia just because of the drivers.
I have been waiting for a year for proper drivers for Linux but as they still have not materialized the next card will be Nvidia, no question about that.
I got an idea: How about some 64-bit drivers. I'm sick and tired of my AMD64 3400+ having a GL refresh rate of a dead dog, or having to run it in 32-bit mode
(Which I refuse to do. I got 64-bits, I'm using them damnit. If I wanted to run a 32-bit OS, I'd run windows)
I can't be bothered buying their cards. I have used ATI boards since 1987. I have owned the EGA Wonder 800, VGA Wonder, Mach8 accelerator (a Win 3.1 accelerator!), Rage chipset boards, Radeons from 7000-9000. Since I ran into a Linux brick wall with them (no specs, no binary drivers) my last two purchases have been Nvidia. I recommend the same for you if you use Linux.
HBI's Law: Frequency of calling others Nazis is directly correlated with the likelihood of the accuser being Communist.
I'm not sure the problem is with ATI.
I see no reason why the drivers cannot be binary just like on Windows. There needs to be a pragmatic approach to this, one which lets binary drivers exist with an interface that doesn't change all the time.
GPL is perfect for GNU tools and the Linux kernel, but has no place for drivers. If always enforced for drivers, then manufacturers just will never support a Linux kernel.
For example, if glibc was change from LGPL to GPL, then Linux would die overnight for commerce, and commerce is what is driving Linux into the enterprise.
The installer software is not the major issue. It is the drivers themselves, and the about of engineering time ATI puts into their drivers.
ATI does beat nVidia for supporting the open source drivers, but that seems lost. Check the XFree86-devel mailing list archives.
Because they only support older models (pre-Radeon 9600 or so). While some work to add open source support to the R300 series is being done, it doesn't work yet. http://volodya-project.sourceforge.net/R300.php seems to be the site of the effort.
From what I've used the binary drivers, they're not _that_ hard to get running, on a friends fc2
laptop it was a matter of copying a few dri header files from the kernel sourcecode (the ATI drivers should be including a copy of these since there's no guarantee the kernel DRM stuff will continue
using it in the future)
Anyway, it was a matter of running make after that. Not as easy as the nvidia stuff, but still not that hard either.
And even then the process is prone to inexplicable failures.
I'm beginning to think the only way we'll see easy driver installation on Linux is if people fork the stable kernel series - while Linus and the gang make all the changes they like to the unstable series, a separate team is preserving ABI compatibility whilst backporting non breaking changes. This task wouldn't necessarily be a huge amount of work - the kernel is pretty mature these days, most of the user-visible work is on hardware support anyway. If users don't get kernel updates every other week, it's not such a big deal.
Seriously, I've been on the ATI beta testing team (although not anymore) and submitted feedback for every driver release to date.
I cannot get 3D working (2D works fine) with my 9800 pro - although exactly the same setup works fine on my old 8500 for 3D.
ATI have not responded to my emails, to my feedback, to any forum posts (although that isn't unexpected) - and this just plain sucks.
Please, if you want a 3D card in Linux, check people have the same hardware and it works if you're after an ATI card. Although only a small group of people have this issue, it is real and does exist.
Gentoo discussion
Rage 3D discussion
Quick Summary Enabling DRI causes X eat all my CPU and not start unless I have a working framebuffer.
With a working framebuffer I get screen corruption, menus and windows are not drawn properly and running any OpenGL application causes X to hang and eat all my CPU.
In both cases I can ssh into my box and kill X or the OpenGL app and I can use the box again.
The only common demoninator seems to be Asus motherboards with certain ATI cards - but the same hardware works fine for Windows XP!
um, why are you arguing that the driver is too proprietary when an actual (well purported...) employee is saying it is just an issue with management? In fact, I have not see any ATI employee argue that there are proprietary routines in the driver that prevent it from being open sourced. The only people that I have seen that make that argument is nvidia. And nvidia does it because they fix their badly designed hardware in the driver, the so called secret driver optimizations. If it's broken, fix the hardware!
I've been waiting for 2 years for decent Windows drivers and my patience is ALSO up.
I updated my old ATI 7500 All in Wonder to a 9600 AIW, thinking that "ATI tries to support the community - they are releasing some specs to the DRI developers, if not for the newest boards."
/. ATI into realising that they need to support us BETTER - after all, telling people "Sorry, our drivers don't work with DirectX 9.0, you have to downgrade to DirectX 8.0" would not fly, so why should we be told to downgrade from XFree80 4.4 or Xorg to XFree86 4.3?
/. story will be, as the bard put it, ".. a tale told by an idiot, full of sound and fury, signifying nothing".
First, the proprietary drivers do not work with Xorg - only XFree.
Second, they will lock up solid if you are running 4K kernel stacks - you need to have 8K stacks. Ven then, while their glxgears program runs, I cannot run UT2003 - as soon as I try to launch the game the monitor shuts down and the system locks.
Third, for reasons unknown I've lost all Xv support - so video playback sucks and I can no longer access my PCHDTV card.
Fourth, GATOS and the proprietary drivers don't mix - so you cannot use the tuner section at all.
I've asked one of the ATI developers who hangs out on the DRI mailing list to push for ATI deploying a Bugzilla-like tracking system, and to support the tuner in the proprietary drivers (since all they need to do is make the tuners an Xv subsystem).
So, let us all
Of course, past experience suggests that this
www.eFax.com are spammers
I'd be annoyed if the installer was an X Window System application. It being text mode is a good choice! Why set up X to use some shitty driver just to install a new driver and have to shut down and restart the X server again?
A GUI version of the installer would operate in exactly the same way, it'd just have pretty clicky buttons instead of a collection of character cells with a different background color representing a button.
Some people still despise ATI's Windows drivers. The latest version of the Catalyst Drivers for Windows XP broke multi-monitor support on my Radeon 9000. Rolled back the driver to an older version and it works fine.
C'mon, guys. You make great cards, how about some decent drivers?
My username does not make me Apathetic. It's irony, get it?
You bought a brand new and (I assume) expensive video card, intending to use it on an OS that's not a current version of Windows, and you didn't check to make sure the drivers worked well? I'm sorry, you get what you deserve.
NV's drivers are closed (as are ATI's later ones) for a very good reason - they contain code licensed from other companies. They can't open source the drivers without getting hit by numerous lawsuits.
The other suggestion, of course, is that they're more open with their hardware specs. That would be nice, but these companies make money off their engineering. If it were simple to kludge together a high-performance GPU, then S3 might actually sell more than 15 boards in a year. I wouldn't want completely open specs if I were them, either.
See, that's the failed logic. LINUX is not a primary market, period. WINDOWS is not a primary market, either. The primary market for computer graphics cards in general are:
1) Males
2) Ages 18 to 35
3) Disposable income
Linux users are definitely a significant segment of that market. I upgrade my video card about once every other year. I typically buy the "high midrange" shortly after the release of the new big dog card. For example, my recent card is an ATI Radeon 9700 Pro. It bothers me that it's sitting in a box and my GeForce Ti4200 is in my AGP slot.
Never confuse volume with power.
I had bought a 9600XT after reading several reviews that gave it outstanding marks for "fps/$." Some OpenGL apps were fine (and plenty fast), but others (notably Wine) crashed my box. Turns out the drivers were oopsing when running an SMP kernel on SMP hardware.
After reporting the (reproducable) kernel oops, I waited 7 months for the next driver release in the hopes it would be fixed. No such luck. I ditched my 9600XT and bought a GeForce 5700U - it just works.
What's sorta ironic is that the 5700U (a massive card, with a huge fan, several passive heatsyncs that requires it's own power input) is in the same "performance ballpark" on most tests (and significantly underperforms on some, like pixel shading, IIRC) as the 9600XT (a small card, with a small fan, and no passive heatsyncs).
It's a great contrast between design elegance and brute force. If ATI could write working drivers...
I'd have to disagree with your runs like a charm sentiment. I also run the DRI drivers. While I get 1100+ running a glxgears demo, the only game I actually use them for is an occasional game of Quake I.
In Windows, running Fuhquake with all the fancy GL extras (rocket trails, explosions, yadda) and 24-bit textures, my 1.2 Ghz Athon and Radeon 7200 (All in Wonder) is as ridiculously fast as you'd expect for a game this old. Under linux, the 24-bit textures alone are unusable. On large levels, things stutter all the way down into the 10-15 fps range. To really keep it smooth, I need to dump pretty much all the fancy GL effects and play what basically was vanilla GL quake in 1997.
I refuse to taint my kernel by using an NVidia card. On the other hand my ATI AIW 7500 still lacks functionality. The GATOS project is great but crippled and held back by lack of specs. I'm not expecting ATI to come out and GPL code for their drivers. All I ask is that the data sheets for the hardware be made available so drivers can be made. As things currently stand I will not buy any new ATI products. I'm not a gamer and what I have works. I'd like to buy a new card but what good would it be to have a card that isn't supported. If a friend wanted to play games on his box I'd have to recommend NVidia despite my dislike for kernel tainting.
Everytime you look at porn a devil gets their horns.
For Fedora Core users, the Nvidia graphics driver is already packaged, and soon ATI's driver will be too. Installation is one command:
/Peter Backlund
yum install nvidia-glx (or fglrx)
That's it. No configuration, no compilation, nothing. You don't even have to reboot. Even easier than Windows. The drivers are provided by the Livna.org repository (http://rpm.livna.org).
Progress on the ATI driver can be monitored here:
http://bugzilla.livna.org/show_bug.cgi?id=211
As of right now, the published version of the Nvidia driver is 1.0.6106, with 6111 coming out shortly.
Some of the improvments made by the livna re-packaging can be read about here:
http://rpm.livna.org/livna-switcher.html
The same applies to the ATI driver.
Note: an ATI employee (M Tippett) has been heavily involved in the packaging process, which shows real committment from ATI's side. Nvidia has not even bothered to answer a request to put a link on their driver download page to rpm.livna.org.
Hello, my name is Robert Lerner, and I pronounce Lernux as "99% cpu"
At least that's my theory.
I've got a Radeon 9800SE All-In-Wonder, which has the new(ish) Rage Theatre 200 chip. This isn't supported by GATOS. I should, of course, have checked this before buying the machine, but there you go. The reason it isn't supported is because it's really complicated and all though ATI have released some specs (under NDA), the GATOS developer(s) haven't gotten round to doing the huge amount of work involved in writing a driver.
I say developer(s), because I think the effort to support the Rage Theatre 200 actually consists of one bloke, called Vlad or something. I think he might be a student of some kind. This may be completely wrong, and I don't want to cause any offence, but that's the impression I've got - one single developer working on the Rage Theatre 200 driver, intermittently, as a hobby. There's been a "don't expect anything for at least 6 months" notice on the website for nearly a year.
The value of open source software is that if something is used by many people and has a long lifetime, the community can build that piece of software into something valuable for everyone, with minimal cost and maximum gain for the participants. This, at least to me, seems to be the key feature of open source.
ATI seem to have gotten the wrong end of the stick and decided that the value of the open source community is that a multi-million dollar corporation can print out a copy of it's specs, along with an NDA of some description, and as if by magic, some student, perhaps called Vlad, will appear out of thin air and do all it's work for it.
Some points for ATI:
Rant over. I make no claims as to the accuracy of the above. In fact I hereby certify that the above is guaranteed to be inaccurate in some way. Please correct me. The emotion is real, though. I'm just fed up with having to reboot into Windows to watch TV.
"The Milliard Gargantubrain? A mere abacus - mention it not."
I'm not going to mindlessly bash ATI but looking at their history and reading up on user reviews clearly shows ATI lacking when it comes to Linux and newer opengl games in general. If you use lots of directx stuff, get an ati card. If you use a lot of opengl stuff, get a nvidia card. That's what it comes down to until either company can get their stuff together and produce an all-around solid graphics card that doesn't cost thousands of dollars.
Main reason I bought an ATI is because they are a Canadian company based in the same City as I am - and its impossible to get a 6800 Ultra up here. But if they don't improve their Linux drivers I know my next purchase won't be an ATI card...
"Just let me know when I can use my ATI All-in-Wonder in a MythTV box... /hates ATI's MultiMedia Center software SO much..."
well... you could try something like snapstreams BeyondTV which supports the AIW and is leagues better than the bundled quasi-pvr applications...
Although even then I'm pretty "meh" on the AIW... I suggest picking up another tuner card that has a linux driver that uses a hardware encoder. (like the WinTV pvr250) You can use that with the OSS IVTV driver and MythTV and have a grand old time. =)
e.
Build Your Own PVR/HTPC news, reviews, &
I have a Radeon 9800 Pro in this machine, going utterly un-used. I've seen some kind of fucking voodoo that I had to pull to get the thing to work with Fedora Core 2, but didn't have the time to do all the tricks.
Honestly, I don't care for Nvidia. I don't like how they have to have their coders look through each game and change the drivers to intercept the game-code and change it to something that works better. Moving down to 16 bit precision wherever it seems "best" to them. It wasn't all that noticable with the Ti4200's, but with the Geforce5's and above, it's the entire method.
I'm getting to the point where I'm going to start swearing off of Ati in total. The Radeon 9800 Pro and the Geforce 5900XT cost around the same (in my part of Town), but the 9800 Pro pulls hela more performance on most of the games I've played with it. I've recommended the Radeon 9800 Pro's to everyone who could fit one in their computer.
I still think the Ati's make better cards for what you get out of them, but I think Nvidia plays the supporting game much better. To top everything off, the entire Radeon line is incompatible with the AMD 8151 "AGP Bridge" chipset.
So I'm converting back to Nvidia, strictly out of Necessity.
Well sure... if you've got a pre-8500 Radeon, thats great. For anything more recent you'll get zero acceleration (2D or 3D) with the XFree/Xorg drivers. Which is not to say that you get great acceleration with ATI's driver, but you get something.
I am not a video driver development, so I'm almost assuredly being naive here - but I would have thought most of ATI/NVIDIA's intellectual property would be invested within their GPU. As far as device drivers are concerned, aren't these just black boxes - eg, send opcode x, operand y, get output z? What's to protect here? Isn't the details of the engine that need to be protected (corporate-wise), rather than the programming interface?
sloth jr
I gotta ask... whatever happened to Matrox? Did they fall off the edge of the earth? What about Voodoo and the others? What is wrong with this industry that we've only got two viable choices left in video display cards when it comes time to buy a new PC? Even in the Windows world, that's a horrifying thought, that the video market is so close to becoming yet another monopoly that we'll have to deal with.
I have a dozen times more choices in what to wipe my butt with after a dump. I guess you're better off being an asshole than being in the computer biz....