8.41.7 Driver released!

[zlatan]

Driver:
https://a248.e.akamai.net/f/674/9206...x86.x86_64.run

Release notes:
https://a248.e.akamai.net/f/674/9206...ux_8.41.7.html

[Septor]

From what I know, 8.41 is recommended only for HD 2000 cards for now, although it may/may not work with lower cards, and not at all for Workstation cards (FireGL, FireMV).

The AMD homepage/RSS is not updated yet, so I take it you just guessed the URL for the driver

Everyone, please read the release notes thoroughly since they have many restrictions listed (hint: read all the CAUTION lines in the release notes).

[up_whatever]

custom Modelines will still crash xorg...

[KanoTheMaster]

I don't know which the problems with the FireGL cards, I only think that the OGL drivers are faster for games but not for apps that need "fully" working drivers. For standard users maybe a ChipID override would work, I just need one tester for that to prove my thoughts. My script has correct fallback of couse. I would not say it is much worse than former drivers, but exept more speed nothing really improved and some tiny bugs have even added. The "new" message from ATI that his monthly driver is only for a part of their cards instead of a unified driver for all newer cards is a bit weak. Of course the changes of this months driver are huger than before but does really anyone believe that ATI will beat NV in this field soon? The previews of that driver have been highly overrated in my opinion.

[zlatan]

kano can you please post the link of your script??

I haven't guessed the release name... it's posted on phoronix so I did a copy/paste

[KanoTheMaster]

The script can be used for Debian etch+ (just not with pure Debian 2.6.21 kernels). Also it works for Ubuntu when no lrm-manager was used before to activate fglrx. Has to be started from a text console (press ctrl-alt-f1).

http://kanotix.com/files/install-fglrx-debian.sh

Next (test) release of Kanotix will be able to start that script on the fly using "fglrx" as cheatcode, similar to the new "nvidia" one for cards which would only work with vesa otherwise. Of course downloading about 50 mb takes a while, nvidia drivers are much smaller.

[Stormking2004]

8.41.7 is the same crap as every month.

21 months ago, XVideo on TV-Out was broken, now OpenGL follows. The "totally new codebase" still has the same bugs as the old one, "known issues" in the realease notes still acknowledges only a fraction of the actual bugs and the driver is still getting worse with every release.

[up_whatever]

Originally Posted by Stormking2004

8.41.7 is the same crap as every month.

well, you have to acknowledge that opengl performance really has improved.

however, ati/amd also did a great job in preserving their great old bugs we got used to and they managed to include even more bugs...

[Stormking2004]

Originally Posted by up_whatever

well, you have to acknowledge that opengl performance really has improved.

I wouldn't notice because I use OpenGL only as a replacement for the broken XVideo-on-TV-Out. As with 8.41, OpenGL-on-TV-Out is also broken, there is absolutely zero progress for me.

Originally Posted by up_whatever

however, ati/amd also did a great job in preserving their great old bugs we got used to and they managed to include even more bugs...

In fact, they managed to break the workarounds ...

[JOPS]

No performance improvements here on my 9800Pro as far as I can see (tested with UT2004 + benchmark and Freespace 2), just more bugs. Every game I've tested under Wine that used to work great doesn't work at all after the upgrade. Maybe it's partly Wine's fault but it worked before so I assume FGLRX's new OpenGL codebase might have something to do with it.

[KanoTheMaster]

Maybe they added a sort of D3/D4 detection to get more speed only there In Win that is a common way to replace shaders and such things by optimised code to gain more speed for specific apps.

[morteoh]

This driver seems unstable on my mobility radeon 9700 (as was mentioned in the release notes).

But I did see a boost in opengl-performance (i know these are not benchmarks.. but still).

glxgears went from ~1600 => 3000
fgl_glxgears went from ~430 => 650

But I had random lock-ups and the installer messed up libGL.so.1 (had to manually symlink it) with the debian/lenny pkgs.
Lets just hope 8.42 fixes lock-ups and we finally get AIGLX

[Septor]

Debian users be warned... the symlink for libGL.so.1 is busted if you install with the package scripts. My fault, I haven't have time to update them in a while. If you just add the symlink from /usr/lib/libGL.so.1 -> /usr/lib/libGL.so.1.2 all should work fine.

As an alternative, you might be able to use the Ubuntu/feisty packages if you use a recent Debian.

[KanoTheMaster]

@Septor

How about removing the Debian target completely when it is more or less untested? My script uses the Ubuntu one as this usually doen't need on the fly patching anymore...

[Septor]

The Debian target has always been more-or-less untested. I threw it in since it was similar to the Ubuntu one, which I do also support and fairly actively test. The only reason that I haven't updated the scripts recently is because I just moved half way around the world and do not have my usual PCs to dev with, and won't for another month or more. I may hack it up on my laptop, but have been just too busy to even think about it.

Once I get settled in things should be fine. And anyways, AMD does not recommend 8.41 for general consumption unless you have a HD 2x00. The few lucky guys with those cards can use the Feisty packages or make the symlink themselves (or use your scripts)

P.S: Package script patches are always welcomed BTW... my email can be found in the package files for Ubuntu/Debian

[sc3252]

Originally Posted by Septor

The Debian target has always been more-or-less untested. I threw it in since it was similar to the Ubuntu one, which I do also support and fairly actively test. The only reason that I haven't updated the scripts recently is because I just moved half way around the world and do not have my usual PCs to dev with, and won't for another month or more. I may hack it up on my laptop, but have been just too busy to even think about it.

Once I get settled in things should be fine. And anyways, AMD does not recommend 8.41 for general consumption unless you have a HD 2x00. The few lucky guys with those cards can use the Feisty packages or make the symlink themselves (or use your scripts)

P.S: Package script patches are always welcomed BTW... my email can be found in the package files for Ubuntu/Debian

yeah, amd does not recommend using these drivers with cards below hd2000 series. Maybe someone who uses one of those can report how well the drivers work.

[sebazve]

does it work with xorg 7.3?

[JOPS]

Originally Posted by Septor

Debian users be warned... the symlink for libGL.so.1 is busted if you install with the package scripts. My fault, I haven't have time to update them in a while. If you just add the symlink from /usr/lib/libGL.so.1 -> /usr/lib/libGL.so.1.2 all should work fine.

As an alternative, you might be able to use the Ubuntu/feisty packages if you use a recent Debian.

Yepp, I noticed that but it was easy to fix.

IIRC Ubuntu/feisty packages are named differently than Debian's so any old ones should propably be removed first.

[JOPS]

Has anyone been able to run 3D apps in Wine with the new driver?

[JOPS]

Originally Posted by KanoTheMaster

@Septor

How about removing the Debian target completely when it is more or less untested? My script uses the Ubuntu one as this usually doen't need on the fly patching anymore...

It's still good enough and I think it should be kept. A missing symlink is not a big issue if it's a known problem.

[KanoTheMaster]

As Thunderbird found the wine issue the next .45 release should work with new driver (or pull cvs).

[JOPS]

Originally Posted by KanoTheMaster

As Thunderbird found the wine issue the next .45 release should work with new driver (or pull cvs).

Ah, there's already a workaround in the development version.

I'll see if I can get time to build some deb's from the git repo and do some tests after work.

[Mosga]

Tried it on my 64-bit FC7 system, and got awfuly slow 2D (X1900 card).

3D was fast, with minor corruptions in glxgears (which stayed even when glxgears window was closed), unfortunately vsync (swap_on_vblank option in .drirc) was ignored.

However, the driver was completely unuseable because 2D looked like it had no acceleration at all (i.e. what you get when you turn off any kind of acceleration) - any kind of scrolling, in browser or terminal, was very laggy, window movement was laggy too. No signs of problems in xdpyinfo output and Xorg.0.log, tried a few options, nothing changed, so I rolled back to 8.40.4.

Anyone got similar experience? According to google, most problems with new driver are either general ones or problems with 3D only, not this "normal 3D / broken 2D" stuff.

[KanoTheMaster]

Anybody can now even try the driver (+wine 0.9.45) in live mode with

screen=auto fglrx

cheatcode for Kanotix Thorhammer. You know it will download 50 mb, so fast connection is best

[1stFlight]

Originally Posted by Mosga

Tried it on my 64-bit FC7 system, and got awfuly slow 2D (X1900 card).

3D was fast, with minor corruptions in glxgears (which stayed even when glxgears window was closed), unfortunately vsync (swap_on_vblank option in .drirc) was ignored.

However, the driver was completely unuseable because 2D looked like it had no acceleration at all (i.e. what you get when you turn off any kind of acceleration) - any kind of scrolling, in browser or terminal, was very laggy, window movement was laggy too. No signs of problems in xdpyinfo output and Xorg.0.log, tried a few options, nothing changed, so I rolled back to 8.40.4.

Anyone got similar experience? According to google, most problems with new driver are either general ones or problems with 3D only, not this "normal 3D / broken 2D" stuff.

I noticed this as well in any video application, everything 2d is very slow on my X1950XT

[Mosga]

Well, in my case things that don't get accelerated anyway, like mplayer -vo x11 were pretty fast - well, not exactly fast, but just as fast as they should be. Regular mplayer -vo gl:yuv=2:lscale=1 was fast too. But window movement was horrible.

[KanoTheMaster]

@Septor

I would highly recommend to update the Debian scripts with current X Server 1.4 in mind. My script has a workaround for that, but it would really look bad when a new driver is out and it does not work.

[gamer101]

This driver release had issues playing video files (x264, xvid, divx) with mplayer & xine. Hope its fixed soon.

(on fresh opensuse 10.2 32bit running X.org & KDE on AMD 690G chipset, Athlon 64 x2, 2600xt

So once again we are left with a broken driver....thanks AMD/ATI.

[KanoTheMaster]

The driver itself works, you just need -ignoreABI. The install is a bit tricky, but you can use my script for that. Link is above.