Insight Into AMD's Linux Driver Development

Cowards Anonymous writes "It's no secret that ATI Technologies has had a rough time in the past delivering display drivers that met the expectations of their customers. When ATI started out producing a FireGL and Radeon Linux driver they for some time were greatly behind NVIDIA's feature-rich driver. The early ATI Linux driver had lacked essential functionality such as PCI Express and x86_64 architecture support and was also affected by stability and performance problems — not to mention a great deal of bugs."

Comments

rough start

[phrostie]

when i switched from NVidia to ATI, it was a rough start.
for the longest time i couldn't get the driver to build/install, then one day everything just worked!
i can't tell you which version it was, but from then on, i've had no problems or complaints.

an open driver would be nice, but even still, my compliments to them.

Re:rough start

[Chandon Seldon]

Everything works? So you can use Firefox at a reasonable speed when logged in as a second user now? You can use Beryl now? Those things sure don't work on the X1300 I bought (a horrible mistake) a couple months ago.

It's really absurd - if they'd just release the programming info for their hardware the X.org drivers would support this stuff inside a week.

Not really surprising

[Endo13]

It's also no secret that ATI has long had problems with their drivers for Windows too.

No, this is not a troll. I use ATI cards almost exclusively myself and I prefer them over NVidia, but I do have to admit that Nvidia's drivers as a general rule seem much better designed and simpler to install.

Re:Not really surprising

[Scoth]

Not sure what kind of "supporting evidence" you might want, but I once bought a Radeon once to replace my aging Voodoo3. I forget the exact model number, but at the time it was about $175 or $200. In some games it was *slower* than the Voodoo3. I gave up and reinstalled Windows clean, but still had the same trouble. Took it back, got a GeForce 3 Ti 200 for about the same price, and it worked beautifully out of the box. Fast forward to several years later and I was looking to replace that card, and I got another Radeon. Installed the drivers, slapped in the new card - poof, blue screen on boot in the video drivers. Stuck the GF3 back in, removed it from the DevMan, and manually installed the VGA driver. Rebooted with the ATI, installed the drivers - back to Bluescreen on Boot. So, another clean install of Windows, and still got bluescreens on boot. Took that Radeon back and got a GF 6600GT again for about the same price, slapped it in, and it's worked ever since. I have a hard time believing I had other bad hardware in there to cause the troubles since in both cases the GeForces worked perfectly. Not to mention the GeForces worked much more nicey in Linux than the Radeons ever did. I genuinely gave ATI two tries now, and both times I was hit with troublesome drivers. I doubt I'll be buying another Radeon anytime soon.

Re:Not really surprising

[Cal Paterson]

Sigh, yet another general statement without supporting evidence. I think your post is a sly bit of astroturfing for NVidia. ATI has had WHQL certified new driver releases for years now. NVidia has only recently been able to get their new releases WHQL qualified. Sure, there is more to drivers, but it indicates that ATI has had a solid development, testing, and qualification regimen in place for a long time.

This statement is utterly untrue. WHQL is pretty much just a case of;

1. Co-operate with MS on driver releases (and institute their "minimum-standard" level of QA)
2. Pay MS what I'm sure is a large some of money for the privilege

Lack of WHQL doesn't indicate anything about driver quality apart from that certain companies are co-operating with MS to institute a minimum standard; many, many third parties develop drivers over and above this already. That fact that companies do not do WHQL says less about their hackers' development style than it does about their executives attitudes towards unjustifiable costs.

So ?

[BESTouff]

When ATI started out producing a FireGL and Radeon Linux driver they for some time were greatly behind NVIDIA's feature-rich driver.

And they still are.

The best way...

[twoboxen]

to build a customer base is to alienate your existing customer base. I bought an R200-based laptop a couple years ago. ATI decided to just not support those cards in their fglrx driver package one day. Why would I buy from a company who won't continue support for their own products for more than a couple years? I will make every effort to never support them again until they get customer/product support in order. NVIDIA, bravo.

Re:The best way...

[the_humeister]

If you're using some sort of Linux/*BSD/etc, you shouldn't have to worry because X.org has had mostly full and useable R100, R200, and recently R300 open source drivers for quite some time now. They're decent. I've been playing Unreal Tournament (and variants) without problems. The only issue is visuals with Doom 3 do to S3 Texture compression being patented. If you're using Windows, well good luck!

what a joke

[radarsat1]

While ATI/AMD is working steadfast in addressing all of these issues and further enhancing their level of Linux support, many of their customers do not realize all of the work that goes into these drivers.

Whatever. They don't need to do any work. All they need to do is open up the specs, and people will do all the work for them. People aren't bitching that the drivers don't work, people are bitching because they aren't allowed to improve them.

There's a whole community out there willing to do all the software work from scratch, but they don't have the resources to create the hardware. The hardware developers somehow see this need to provide the software themselves, instead of taking advantage of the community, but then go and do a shoddy job of it. That's why people are annoyed by the whole thing. It could be so much better, with very little effort from ATI, but they steadfastly refuse to play nice, forcing developers to resort to reverse engineering. Same goes to Nvidia by the way, but at least they seem to be a bit more competent in Linux/X.org driver development.

This whole argument is just a big excuse. We don't want excuses, we want some damn drivers.

--- someone who's been buying Nvidia since he realized that ATI doesn't work as well on Linux.

Re:what a joke

[Kadin2048]

The hardware developers somehow see this need to provide the software themselves, instead of taking advantage of the community, but then go and do a shoddy job of it.

Bingo.

When hardware companies try to make software, the result is almost inevitably shit. There are some exceptions, but big hardware companies tend to see software development as a 'cost center,' an afterthought to be minimized as much as possible, rather than a critical and major part of their product.

Look at scanners if you want. I've used some great film scanners in the past; brilliant hardware engineering, but coupled with the absolutely shit software that came in the box with it, it was practically a doorstop. To get anything else done, you had to get VueScan or Silverfast -- addon software written by people for whom software is their primary focus.

Re:what a joke

[revengebomber]

The community COULD do their own drivers, but the specs aren't available. Everything about how to interface with the card would have to be found via reverse engineering.

I'm fairly sure the reason the specs aren't open, is because it would disclose some "secrets" about how the companies optimize their cards.

Re:what a joke

[jZnat]

It's also possible (and likely) that they are violating patents they don't have a license for, so giving the specs out might bring light to this. A stupid legal reason rather than a technical reason...

Nvidia is not the competition

[crush]

The article is a long excuse explaining why AMD/ATI are unable to release decent GNU/Linux drivers. That's interesting enough as far as it goes: AMD/ATI and Nvidia both have crap closed, proprietary drivers which don't work well, make kernel updgrading difficult and are unauditable for security. So why bother with them? Further ATI have a history of dragging their ass and blocking the release of Free drivers,

Why bother with this crap? Just get an Intel GMA X3000 integrated motherboard and save time, power, money and hassle due to Intel "getting it" and releasing Open Source drivers and full specs. (You'll probably also be able to benefit from their free wireless drivers.

If you're into hardcore gaming then you're probably running a PS3 or an Xbox on the side anyway.

Re:Nvidia is not the competition

[Mad Merlin]

You think you want a stable kernel ABI, but you really don't.

Re:Nvidia is not the competition

[kscguru]

That's why Linux hackers want all drivers in the kernel tree, so they can find anything which breaks due to an API change, and fix the problems.

And that's exactly why hardware manufacturers DON'T want their drivers to live in the kernel tree. They don't get:

• Bug fix backports: Kernel hackers apply a patch to the latest vanilla kernel, then say backports are the distro's problem. Two distros do, fifteen don't, users of those fifteen scream at the hardware companies for having buggy drivers. With an out-of-tree driver, you just need to update the driver. Windows and OSX are light-years ahead of Linux here.
• Community contributions: Kernel hackers only add GPLed code; very few even allow dual-license, and certain vocal hackers are zealously GPL-only (to the extent of rewriting code JUST to make it GPLed). Which means the hardware vendor can't take a fix from a linux driver and integrate it into a Windows driver without GPLing that too. The vendor has to either GPL all drivers (Windows included), or maintain two separate trees (GPL and non-GPL), or not open drivers at all. Guess which is easiest / cheapest?
• Driver API stability: a modern 3D graphics driver is a full OS in its own right, with internal threading models, schedulers, memory management, context switches, etc.; a modern driver needs more than just bugfixes. Every good developer knows the way to keep two large codebases manageable is a stable API between them; the only people who don't seem to get this are otherwise-intelligent Linux kernel hackers.
• Kernel API freedom: Kernel hackers like stable userspace APIs (for good reason). But hardware vendors don't need to provide stable APIs if they have a shim library that actually talks to the cards (e.g. atioglxx.dll, the ATI OpenGL implementation). It's a lot easier to let the API change rapidly and only commit to a stable API at the library interface (the OpenGL API).
• Easier work. The Linux kernel development process is optimized for making the kernel hacker's life easier at the expense of the driver developer's (hint: saying "we'll update your driver for you" clashes very badly when the HW vendor is simultaneously making changes). If kernel hackers want to see better device drivers, they need to stop treating drivers as second-class citizens. Microsoft is very good at courting driver developers; Linux is the definition of arrogance.

ATI and Fedora 7 / X.Org 1.3

[harrypelles]

I made the same mistake as many Fedora users - jumping (to Fedora 7) before looking. I'm not poking at Fedora here, on the contrary, I am a loyal Fedora user. It's ATI I'm upset with. ATI released a new fglrx driver (version 8.37) since Fedora 7's tests and final release that also does not work with X.Org 1.3. We're all sitting around waiting for the 8.38 which ATI claims will be compatible. And don't even get me started on ATI's absent AIGLX support for Linux. My next card will nVidia.

Re:ATI and Fedora 7 / X.Org 1.3

[drgonzo59]

The difference is that here Vista is to blame, while on Linuzz everyobody magically undarstand that this is a driver's problem.

Perhaps because 'Linuzz' is open it is easy to see where the problem lies. With Vista you get huge binary blob and if it's broken you don't know if it is the drivers or Vista -- you can't debug it and look at the source so you call MS tech support and wait 6 months for a service pack or MS tells you to call ATI/AMD and you wait 6 month for a fix. Binary drivers suck that's the problem here...

Nothing to see here.

[BESTouff]

Basically, TFA says that "ATI has a release cycle". They even have an unofficial bugzilla and an unofficial wiki. Oh, and they'll drop R200 support too. And all that's supposed to make better drivers for Linux one day. I really wish they'd go the Intel way: hire some top-notch developers, give them specs and make them do Free drivers.

Re:Misleading summary title?

[RingDev]

Yeah, in the same way that Mazda is a part of the big Ford company. But you wouldn't submit and article called "How Ford Builds Cars" and then cover only the Mazda factor's stereo installation.

-Rick

Translation: ATI fails to release OSS drivers

[dwheeler]

Wow, a lot of text to paper over the obvious problem: ATI is still failing to release OSS drivers for Linux. The paper describes all the ways that ATI tries to avoid releasing the drivers, and how they all fail to solve the problem. ATI has testing processes, etc., sure - but later on, when X.org and the Linux kernel change, there's no way for me to update the driver - so I have to hope that they will EVENTUALLY do so (leaving me vulnerable to any security problems) OR throw away the ATI card.

Dell has solved this problem by including the Intel stuff instead for their Linux offering. It's time for ATI to release their drivers as OSS.

Intel has not release docs

[dmoore]

> Intel "getting it" and releasing Open Source drivers and full specs.

Actually, Intel has not released docs for their GMA X3000. Their current stance is that the driver is the documentation. That's fine and good, except the driver is still very incomplete (missing OpenGL features, no XvMC, no tv-out, etc.). See here:

http://lists.freedesktop.org/archives/xorg/2007-Ma y/024582.html

nVidia blob isn't crap.

[gukin]

I know that lots of folks are vexed that nVidia won't open their 3D driver, saying "If nVidia stops supporting Linux you're all doomed." Well just who do you think supports the open source nv.c driver? How about the nVidia SATA driver? Yes, it's nVidia, so even though the 3D stuff is closed source, they're still supporting OSS.

Next about the "Crap" drivers from nVidia, I've ordered a bunch of new Linux PCs, each will have a low-end nVidia video card added when it arrives? Why? Well I need dual headed support and that can be spotty with other video card vendors. I also need to run them in 8-bit color (don't ask, I just need to.) and my experience with the glorious wonderful OSS Intel video drivers is a nasty little box that follows the cursor around on the screen. I don't have ANY issues with nVidia's "Crap" drivers, everything looks great and works great.

I applaud ANY vendor who makes efforts at supporting OSS but I buy stuff from vendors who support Linux. Every system I own is either an nForce mobo with an nVidia video card or is simply sporting an nVidia card. When asked about what to buy, I recommend nVidia products. They have the best quality 3D support and performance of ANY vendor (which isn't much), they make it possible for me to play games under Linux that I'd otherwise have to play under Windows and that is worth a lot to me.