AMD Catalyst 12.3 For Linux Officially Released

[Kano]

Maybe you did not notice that I binary hacked fglrx to support Xorg 1.3 and to get rid of the watermark. But i absolutely hate that this is needed at all, that's complete nonsense to mark a driver that it was not tested on that specific hardware. You gain nothing with that info, if the driver works you see a stupid logo and you hate it. If it would not work at all then you don't see a logo, nice

[allquixotic]

You should watermark everyone's hardware with "Catalyst is broken messed up shit that will bring down your system and cause data loss. Don't install it on a system where you have anything important. In fact, just run rm-rf on random paths and files and you'll get the gist of the experience without having a 100 MB blob of who knows what running in your kernel." Catalyst is truly unsafe at any speed.

+1000. Having a watermark to specifically say that some hardware may be buggy is based upon the flawed assumption that if your hardware passed QA, then everything is fine and dandy and the driver works 100%.

But no, I don't think it's worth AMD's time to code up some way for people to disable the watermark. I already have code on the Phoronix forums for gutting it out of your driver. Just run the code and get on with your life.

On the other hand, I wouldn't be at all surprised if AMD did something tricky to deeply obfuscate their watermarking code so that you can't (easily) disable it with a binary hack anymore. They're fine with spending engineering effort to prevent people from doing things that help themselves, but if you ask them to spend 5 minutes to help the user, it's "you either let us work on the driver or we drop everything, no progress gets made, and we spend days of lost productivity just trying to satisfy you and your greedy need to have a driver that doesn't splat a big green watermark in the bottom-right corner".

I know I'm greatly exaggerating and grossly overstating the problem and dramatizing it, but I'm feeling abusive today. The world has been abusive to me, so I'm randomly lashing out. Disregard this post, bridgman -- I really like fglrx. With lolsed run over it, of course. :P

[DaemonFC]

Kano, I know you understand this and are just having fun, but in case anyone else was confused by this...

The PCI IDs in the code determine which hardware the driver will *try* to run on. The code might work perfectly, or be buggy, or fail to run at all.

The PCI IDs in the control file show which hardware the driver has passed QA on. The time and effort is not to add IDs to the file (that's all automated anyways) but to do the testing, bug fixing and re-testing required to pass QA on the hardware.

Once testing and bug fixing on specific hardware is completed, the IDs in the control file are updated to reflect that these binaries form a production driver for the hardware it's running on, whereas previously it displayed a "hey this might run but be aware it's not a production driver" message.

Is the watermark too intrusive ? In my personal opinion something like a one-time splash at startup would be better. The question though is whether the devs should be working on getting production support in place more quickly or spending that time on rewriting the watermark code so that running a pre-production driver isn't as annoying. I would put available people on speeding up production support myself, which is what seems to be happening, but I realize not everyone agrees with that and it's kind of academic because I'm not involved with fglrx development anyways.

As we get more consistent at posting SKU-specific launch drivers if a regular Catalyst release with support isn't coming out at the right time, I think you'll find the watermarks become less of an issue anyways.

You should watermark everyone's hardware with "Catalyst is broken messed up shit that will bring down your system and cause data loss. Don't install it on a system where you have anything important. In fact, just run rm-rf on random paths and files and you'll get the gist of the experience without having a 100 MB blob of who knows what running in your kernel." Catalyst is truly unsafe at any speed.

[zzarko]

It works somehow...

OK, creating deb files was a no-no... I installed the driver directly from the installer and now it works (after fiddling with xrandr to add 1920x1200 resolution to monitor). But, now I have another problem: my new monitor is 1920x1200, and my old projector can accepr resolutions up to 1920x540. Is there a way to scale screen in clone mode for displays with lower maximum resolution?

[nej_simon]

Wow, sleep/resume still doesn't work on my laptop! Only one more release and it'll have gone a year since they broke it! This issue is very common btw. My next laptop certainly won't be AMD based.

[zzarko]

Asus hd 7950

I just got new ASUS HD 7950, and I installed 12.3 on my Ubuntu 11.10 32-bit. After generating deb packages and installing them, I got 1024x768 on Dell U2412M (it should be 1900x1200). When I started CCC, my monitor was disabled, I enabled it, clicked on apply and got Segmentation Fault for CCC:

(process:3411): GLib-GObject-CRITICAL **: g_object_new: assertion `G_TYPE_IS_OBJECT (object_type)' failed
Segmentation fault

There were some errors in Xorg.0.log:
[ 39.063] (EE) fglrx(0): atiddxDriScreenInit failed, GPS not been initialized.
[ 39.441] (EE) fglrx(0): XMM failed to open CMMQS connection.(EE) fglrx(0):
[ 39.441] (EE) fglrx(0): XMM failed to initialize

When trying to play video from VLS, I get:
[0xa0a4e9c] xcb_xv generic error: no available XVideo adaptor
[0xaf6ac3e4] main video output error: video output creation failed
[0x9f1618c] main decoder error: failed to create video output

What now? OSS driver gives me 1900x1200, but it barely works (I cannot play any video, nor 3D).

EDIT: VLC info

[storma]

Even a simple checkbox within amdcccle to acknowledge that you are using an unsupported driver/hardware and that you wish to continue at your own risk, to turn off the watermark.

[Kano]

Nvidia has got a beta splash, add that if you want, but use all pci-ids from fglrx kernel module for aticonfig, everything else is stupid for a generic driver. No watermarks! Btw. i never saw a watermark with win drivers (as you have to use there always betas when you look what is recommended for gamers) just Linux users get that shit.

[grigi]

I think the issue with the watermarks is, that sometimes the driver works perfectly, but the watermark stays on to annoy users.
Fglrx is notorious for the annoying watermark, and I have had drivers where the watermark appears with an update.

I wonder if the watermark there causes less or more frustration? On this forums (and any other I have ever been on) it _seems_ more, but most people don't complain vocally.

[bridgman]

The supported pci-ids did not change at all from 12-2 to 12-3 when you look at the fglrx kernel module. from 12-1 to 12-2 only 2 ids where added (679e and 990a). So the only change could be in the control file to get rid of unsupported hardware. And the devs needed so long to do that, really really bad joke.

Kano, I know you understand this and are just having fun, but in case anyone else was confused by this...

The PCI IDs in the code determine which hardware the driver will *try* to run on. The code might work perfectly, or be buggy, or fail to run at all.

The PCI IDs in the control file show which hardware the driver has passed QA on. The time and effort is not to add IDs to the file (that's all automated anyways) but to do the testing, bug fixing and re-testing required to pass QA on the hardware.

Once testing and bug fixing on specific hardware is completed, the IDs in the control file are updated to reflect that these binaries form a production driver for the hardware it's running on, whereas previously it displayed a "hey this might run but be aware it's not a production driver" message.

Is the watermark too intrusive ? In my personal opinion something like a one-time splash at startup would be better. The question though is whether the devs should be working on getting production support in place more quickly or spending that time on rewriting the watermark code so that running a pre-production driver isn't as annoying. I would put available people on speeding up production support myself, which is what seems to be happening, but I realize not everyone agrees with that and it's kind of academic because I'm not involved with fglrx development anyways.

As we get more consistent at posting SKU-specific launch drivers if a regular Catalyst release with support isn't coming out at the right time, I think you'll find the watermarks become less of an issue anyways.