AMD Legacy Catalyst 13.1 for HD2000-4000 is out

[bridgman]

New Xserver/kernel support for a legacy driver has never happened and no one ever said it would (correct me if I'm wrong), so isn't that debate academic at this point?

I thought the first legacy driver added 1.12 support, didn't it ? Going from memory here so not 100% sure.

[TemplarGR]

Maybe a dumb question, but how are we "shooting ourselves in the foot" by making Valve Linux gaming fixes a priority for the legacy release ?



Definitely lots of unhappiness about not having the fixes, but your post suggests that prioritizing these fixes over (say) 1.13 X support was a Bad Thing. I'm not seeing that in the forums (other than this post, which is why I'm asking).

I think that you should stop pretending to be dumb Mr Bridgeman. We all know what's the problem here:

HD2000-4000 is hardware perfectly capable of handling ALL modern games. Maybe with a smaller graphical detail here and there, but overall for most gamers there is absolutely no need to upgrade to anything more powerful as long as games are mostly console ports of consoles 6-7 years old.

The reasoning behind dropping support for those is that they have reached their potential. Fine. It is perfectly reasonable. What is not, is that the legacy driver isn't at least maintained for bugs(like Steam) AND for later kernel/Xserver versions. I don't get why it is so difficult for AMD to at least add support for newer kernels/Xservers. If it costs too much on QA, just provide it as is, unofficially, as an "unsupported gamer version", whatever.

My HD3870 is perfectly capable to play almost all games on Windows 7 in maximum detailed 1680x1050 resolution with good framerates. Crysis with maximum settings at 1680x1050 is reasonably playable(although with severe frame rate drops). But Windows 7 is not my OS of choice, and right now is not installed on my PC at all. I would LOVE to game on Linux a bit, but i am not going to use an outdated distro in order to do so. Call me cheap, but i am not going to buy a new gpu to get driver support either. Especially since APUs are becoming so powerful and i will probably wait to get Trinity's succesor for my next system upgrade.

R600g is a great driver, and i salute AMD's efforts. I really do. But AMD shouldn't drop the Catalyst support for R600/700 until the R600g could at least have consistently 60-70% of the performance of the blob. There is a reason people complain Mr Bridgman. Not all of them have "lost it" all of a sudden...

[bridgman]

Huh ??

I was asking about this post :

Entropy and Rigaldo seem to be correct, the driver does not work with X.org 13.1, at least as far as I can tell. This was basically a rush-fix to make Team Fortress 2 (and perhaps others? not sure) work as it threw an error with 12.6. AMD is really shooting themselves in the foot with this, right as Valve pushes Linux towards being viable for gaming. Leaving a really sour taste and definitely making me second-guess buying another AMD card when I build my new rig, and judging by the reactions across various forums, I'm not the only one.

I read it focusing on the first two sentences, ie I thought LordSocky was complaining about the decision to prioritize TF2 over 1.13 support.

LordSocky's response was :

Getting that fix out was definitely the most important thing, I'm not downplaying that. It's just that it took so long to happen and didn't bring hardly anything else as far as I know. It doesn't help that there was a massive misconception around the release regarding whether or not it actually supported 1.13. That part's not necessarily your fault, but the version number initially confused a lot of people, including myself. There's also a lot of anger around the issue that our hardware, which is only a few years old and some of which is still very much capable of handling modern titles, has been left behind, at least for now. I don't want to dislike AMD, because I have a pretty strong dislike of nVidia already and that pretty much removes both of my choices, but my current situation has made me secondguess what I put in my next box if it's just going to be relegated to the legacy bin in a couple of years and cause more headaches down the line.

Basically saying that "shooting ourselves in the foot" was related to (a) time lag between getting Steam fixes out for new GPUs vs for legacy GPUs (consequence of the slower legacy release schedule) and use of same version number for main and legacy drivers when the content (specifically 1.13 support) was different. Both points seem reasonable, and my only response would be to say that I understand and to point out that :

1. We *are* doing legacy driver releases for Linux after the Windows support ends, which is something we have not done before.

2. The open source drivers are continuing to improve and apparently do support TF2 already, so there is a solution for users with newer X servers today albeit with a performance penalty relative to Catalyst.

I'm not suggesting that anyone has "lost it" and I don't understand why you are implying that I am.

I asked a specific question about a specific prioritization decision, nothing more. I can sometimes influence development priorities within the resources we have, but I don't control the allocation of resources to OSes, so my interaction tends to focus on areas where I can help.

For the record, I am quite capable of being dumb some days without pretending. Ask anyone here

[TemplarGR]

Huh ??

I was asking about this post :



I read it focusing on the first two sentences, ie I thought LordSocky was complaining about the decision to prioritize TF2 over 1.13 support.

LordSocky's response was :



Basically saying that "shooting ourselves in the foot" was related to (a) time lag between getting Steam fixes out for new GPUs vs for legacy GPUs (consequence of the slower legacy release schedule) and use of same version number for main and legacy drivers when the content (specifically 1.13 support) was different. Both points seem reasonable, and my only response would be to say that I understand and to point out that :

1. We *are* doing legacy driver releases for Linux after the Windows support ends, which is something we have not done before.

2. The open source drivers are continuing to improve and apparently do support TF2 already, so there is a solution for users with newer X servers today albeit with a performance penalty relative to Catalyst.

I'm not suggesting that anyone has "lost it" and I don't understand why you are implying that I am.

I asked a specific question about a specific prioritization decision, nothing more. I can sometimes influence development priorities within the resources we have, but I don't control the allocation of resources to OSes, so my interaction tends to focus on areas where I can help.

For the record, I am quite capable of being dumb some days without pretending. Ask anyone here

Basically, LordSocky wasn't clear enough, but from what i understood he implied that AMD dropped the ball by releasing 13.1 legacy WITHOUT Xserver 1.13. He didn't say that Steam fixes were unwelcome, just that by releasing the driver AMD could have supported the latest kernel/xorg. Making games work should be a priority, but making graphic cards work should rate higher than that... And that now that Linux gaming is becoming big and this "legacy" hardware is more than enough for it, AMD dropped support entirely. That is what i got from his message.

I may be a developer but have never programmed gpu drivers so i do not know much. But i simply cannot understand what makes it so difficult to bump the support for newer versions of kernel and xserver. Seriously. It is not like with every version the API is written from scratch... Nvidia does it for ages. Nvidia updates its legacy drivers for newer kernel/xorg support almost instantly. Why this is no problem for them?

The opensource drivers are indeed becoming really good, i can testify to this. Last Spring i was able to play Mass Effect 3 with Wine and Mesa 8 almost without issues(except less performance of course). I almost finished the game till a game stopping crash. I am on Archlinux with latest stable releases right now and performance is good, the driver is trouble free. I eagerly wait for the next MESA to try HyperZ and see how speedier r600g became(i could just install the latest development from git but that is not for me anymore LOL).

As i said in my previous post, i am really thankful for the AMD opensource support. The problem is that HD2000-4000 should not have been dropped last summer, r600g wasn't ready. This summer is a much more mature period and IMHO this is the period AMD should have picked to drop support.

And of course i don't say or imply that you can influence Catalyst decisions that much, or holding you responsible for anything . It just seemed to me that you tried to sidestep the issue LordSocky(and others of course) have with AMD's Catalyst policy.

[bridgman]

It just seemed to me that you tried to sidestep the issue LordSocky(and others of course) have with AMD's Catalyst policy.

Yep, that's fair.

Strictly speaking I guess I did, although in my mind I was "responding to the parts where I might be able to help" rather than "sidestepping the other parts"

[Vim_User]

1. We *are* doing legacy driver releases for Linux after the Windows support ends, which is something we have not done before.

So AMD declares my 3 years old laptop for obsolete, for "legacy" hardware, and they are boasting with supporting it with a legacy driver? You must be kidding.

About priorities: Priorities are not the same for anyone. I game on my desktop machine (sadly with "non-legacy" AMD videocard, next one will be Nvidia, since I will not accept AMD's crappy driver policy). My "legacy" hardware is in a laptop, not powerful for gaming anyways, so my priorities are different on this machine, I would prioritize newer versions of Xorg and especially official support for later kernels higher than gaming. Damn it, AMD the latest kernel supported by bothe 13.1 drivers is 3.5, which is EOL. So the latest really supported kernel is the 8 months old 3.4, althiugh patches for newer kernels are almost instantly available for later kernels. Why is it so hard to implement them in the official releases? Why do some distro developers have to do the job of AMD driver developers?
And no, the OSS drivers are not an option, in fact they are not an option for most laptop owners without working video-decoding and proper power-management.

AMD has to do much better than that to not loose their customers to Nvidia (they may not help with an OSS driver, but at least I get longer support and better support for old and new videocards from Nvidia) or Intel.

[pvautrin]

[...]the decision to prioritize TF2 over 1.13 support.[...]

Hi Bridgman,

As owner of a legacy HD 3650 AGP (1002x9596), I made the conservative choice of the 12.04 LTS. But while I can understand the prioritization decision between Steam fixes over new graphics stack, I fear AGP owners are being left out despite "AMD Radeon HD Series AGP" being listed on the download page.

Indeed, my installation attempts on both 12.6-legacy and 13.1-legacy all resulted in "No supported adapters detected" and restoring the backed-up /etc/ati/control file from the default 12.4 driver (which works fine) removes the watermark (but leads to another crash, so this is not the solution). I've been on the radeon driver for years, but I'm looking at using my system for Steam games and xbmc/xvba now.

Since hours of Google did not help finding a solution, could you please use your influence to clarify whether AGP cards (including 1002x9596) are supported by the legacy linux drivers?

Thanks in advance and apologies for asking you about fglrx, but I assume this is easier to answer than the ETA for UVD with radeon

[bridgman]

AFAIK the AGP cards are supposed to be supported by the legacy drivers as well.

Usual questions first, I guess -- are you sure that the previous driver was fully uninstalled before trying to install the newer drivers, and was uninstalled in the same way you installed it (ie if you did native install then run the uninstall script, if you built packages then uninstall the package) ? I don't think leftover bits from the previous driver install should cause a "no supported adapters found" message during install but you never know...

Guess I should check one other thing -- was the "no supported adapters" message from the installer or from the driver at startup ? If the latter, there should have been some kind of error message thrown during installation and that's the one we need to understand.

[Kano]

I doubt that it works. One Kanotix user had similar issues with that fglrx-legacy driver with this AGP HD 4 card.

[pvautrin]

AFAIK the AGP cards are supposed to be supported by the legacy drivers as well.

Usual questions first, I guess -- are you sure that the previous driver was fully uninstalled before trying to install the newer drivers, and was uninstalled in the same way you installed it (ie if you did native install then run the uninstall script, if you built packages then uninstall the package) ? I don't think leftover bits from the previous driver install should cause a "no supported adapters found" message during install but you never know...

Guess I should check one other thing -- was the "no supported adapters" message from the installer or from the driver at startup ? If the latter, there should have been some kind of error message thrown during installation and that's the one we need to understand.

The fglrx 12.4 were both installed and removed from Kmenu > Applications > System > Additional Drivers (jockey, I believe).
As for the legacy drivers, I've mostly tried the guide from this thread [1], where I included my logs (containing the "aticonfig: No supported adapters detected") and described symptoms. dmesg, Xorg.0.log and lspci from this Kubuntu 12.04 x64 are in [2].

Given your positive feedback, I decided to try this PPA [3] on a fresh Kubuntu 12.10 x64 (spare HDD), and this time 13.1-legacy installed fine! I still have the Unsupported hardware watermark but at least fgl_glxgears works (will try UVD soon). For reference, I provide Install log [3] (error gone), dmesg [4] and Xorg.0.log [5].

So it seems indeed this is another installation issue, sorry for the noise. If only a similar package/ppa existed for precise...

Kano, if you have some scripts to try to troubleshoot this issue, it could also help your users

[1] http://forum.xbmc.org/showthread.php...660#pid1219660
[2] http://forum.xbmc.org/showthread.php...970#pid1215970
[3] https://launchpad.net/~makson96/+archive/fglrx
[4] http://pastebin.com/AWAsmv3g
[5] http://pastebin.com/eEpBBRCf
[6] http://pastebin.com/QNZgzmEZ

Edit: CCC screenshot (12.10): http://picpaste.com/HD3650AGP_8.97.100.7-IpuhO0nk.png