"Ask ATI" dev thread

[Kano]

When you use newer kernels than 2.6.28 you get lots of errors in dmesg - even when you patch the kernel to support fglrx 9-3. This may run for a while but is not 100% stable. I did not test 9-1, but most likely it will run with Kanotix too when you downgrade to 2.6.28 kernel (use -v 9-1 option to the script). It's definitely not possible to patch the driver in a simple way to support newer X servers as you get unresolved symbols with newer X servers. Rendering issues with gl2benchmark have been fixed since 9-2 when i remember correctly but as you only want powermanagement that should not matter.

[piete]

When you use newer kernels than 2.6.28 you get lots of errors in dmesg - even when you patch the kernel to support fglrx 9-3. This may run for a while but is not 100% stable. I did not test 9-1, but most likely it will run with Kanotix too when you downgrade to 2.6.28 kernel (use -v 9-1 option to the script). It's definitely not possible to patch the driver in a simple way to support newer X servers as you get unresolved symbols with newer X servers. Rendering issues with gl2benchmark have been fixed since 9-2 when i remember correctly but as you only want powermanagement that should not matter.

Thanks Kano! Is the newest version the one available for download at kanotix.org?

I have been thinking about why the old binary drivers don't work with new kernels (in a very "high" level of course ), is it because they are so badly written that it is impossible for newer kernels/xorgs to advance and at the same time maintain compatibility with the old drivers, or is it the open source developer policy not to take into account the "requirements" of the binary blobs to favour the open drivers?

I hope it is not the latter case, since although I respect the open source driver development, I wouldn't see anything bad in having more choice, if it was easy to maintain compatibility. So far I am ok getting to use at least newer kernel than the enterprise linux distributions, but for how long? Soon I may want to be able to enjoy the new programming advances, but before open drivers mature to the (power saving) level that satisfies me, I'm not going to risk frying my laptop.

This may not be the right thread to write this, but let's face it, catalyst 9.1 (and/or 9.3) is the last driver to fully work with a big number of ATI cards. And it's a fact we can't do anything about. Knowing this, if it was possible for newer kernel/xorg versions to maintain the compatibility with the driver, doing otherwise would harm many linux users, be it closed driver or not. One reason I still stick with Windows is that my hardware works with it and it has passed numerous updates, even the kernel (although to a lesser extent).

Since March 2009 I've seen many angry messages towards ATI for not continuing support to legacy cards. But the cards still have a working driver. All kernel/xorg developers, forgive my ignorance if the old drivers really are impossible to support. If I was a programmer, I didn't have to ask my ignorant questions.

[Kano]

The newest link is what you get when you ask me in the IRC ( irc://irc.freenode.net#kanotix ).

[bridgman]

It's "the latter". Windows maintains back compatibility with older driver APIs, so drivers from a few years ago are normally still useable. Linux does not attempt to maintain a driver ABI, so drivers require frequent updates in order to keep working.

[Shining Arcanine]

i think they are reverse-engeenering the nvidia driver trying to tailor it for ati card, cause i seriously doubt they realy can come up with they own bugfixes ..

I suspect ATI is the reason why Nvidia will not open source their drivers. Nvidia's drivers are well written enough that they could probably support both Nvidia's and ATI's hardware with the right changes.

[piete]

It's "the latter". Windows maintains back compatibility with older driver APIs, so drivers from a few years ago are normally still useable. Linux does not attempt to maintain a driver ABI, so drivers require frequent updates in order to keep working.

So what's the reason for this? Are closed drivers the ultimate evil?

[bridgman]

Depends on who you ask

Most Linux kernel devs are OK with binary drivers but aren't willing to support a frozen ABI whose primary purpose is simplifying the use and maintenance of binary drivers.

It's probably safe to say that Windows design decisions are made with an eye to being effective in a closed source environment, while Linux design decisions are aimed at being as effective as possible in an open source environment (where kernel and driver can be changed together when necessary).

[bwat47]

Will fglrx EVER support compiz and video playback with compiz vsynced/without tearing? thats the reason Why I am using the open source driver on my laptop over fglrx (id love to be able to use flgrx cause the oss power management is god awful) but I cant stand the tearing in compiz and the tearing video playback. Compiz sync to vblank does nothing and neither does vsync in ccc.

[Fluggo]

@bridgman

Is there any support for rotating the screen coming anytime soon to linux? (when using pivot lcd's for example). Is it planned for or are ati at least working on it?

Can you explain why this support is still lacking in fglrx?

[Panix]

Why are Fermi/newest Nvidia cards supported but not Evergreen cards?

Fedora 13 users can't even use Evergreen cards? I realize FOSS drivers aren't ready yet but fglrx drivers, too?

I guess I can use Ubuntu 10.04 primarily but just curious why such a discrepancy? :-( Evergreen cards were released last Nov. ('09)?

Is there any process for driver support in Fedora 13 (and Ubuntu 10.04) such as various scripts, patches and/or anything else? I mean, if I bought an Evergreen card and installed F13, would there be any way to use the card?

Or do I have to use OpenSuse 11.3 or Ubuntu 10.04?