AMD 8.41.7 Display Driver Released -- The Holy Crap Edition!

[andrig]

Shit, okay, that's a bug in the RPM SPEC file for x86_64 that I overlooked in the packaging script. You can correct it in the packages/Fedora/ SPEC file by adding the line for fglrx_dri.so or by deleting the fglrx_dri.so file after extracting it... I can release new packaging scripts here shortly. fglrx_dri.so was replaced in 8.41 as a dummy file late in the cycle.

I used --keep at the command line so I could look at the files, and modify as you say, but then I'm unsure of how to build the packages after. I have tried to run ati-installer.sh --buildpkg Fedora/F7, but that complains that Fedora/F7 is not supported. After either deleting the fgl_dri.so, or modifying the SPEC file, what script do you run with what arguments?

Thanks.

[Snake]

8.41 works for me

8.41 works for me (x1400, Core2Duo Lenovo Z61m, 2GB RAM, Gentoo 64-Bit, 2.6.22), and it IS faster with everything I tested so far:

- glxgears (~3100 FPS -> ~4200 FPS)
- fgl_fglxgears (~590 -> ~910 FPS)
- UT2004 Benchmark (50 FPS -> 75 FPS)
- blender (1300 iterations -> 1690 iterations)
- SecondLife ("KTris drawn/sec" ~2100 -> ~3300), plus VBO working, plus no more "1 frame per 5 seconds" when "GL mem" alloc grows bigger than half the phys. video ram size (128 MB, before SL had to be set to 64 MB to work at all), plus no more "turning lag"
- xv works (but not tested on TV-Out yet)

New problems so far:
- Had one "black screen" when switching consoles
- some glesx error when terminating X
- SL does not work in fullscreen anymore

I do not understand all that bashing done here. It was clear to me that 8.41 would be some kind of "beta" quality, because the code actually started to change. But the performance indicates those are *good* changes.

Sadly, setting up 3D on linux is not always easy currently (look at all those "No DRI here" threads everywhere), but there are many possible reasons other than the actual "driver" for this (e.g. with 8.40 I got only half of the "max LFB" as "free LFB" until I removed the "FSAASize"-setting from /etc/ati/admpcsdb)

So, I wonder if some problems discussed here are system / configuration related. I currently have NO special settings in xorg.conf, and a really "bare bones" amdpcsdb, and the system works (has bugs, but works).

So, for me 8.41 is BETTER than 8.40 (performance), and I think 8.42 will be BETTER than 8.41 (bugs, features). And there are open specs now. Things *are* getting better ;-)

[opera]

And you were planning to run Linux on it? knowing ATI's problems? Or did you find out after? (lack of research?)

Edit

Again, you pay for Windows with most laptops... No wonder pretty much every Linux laptop seller ships with nVidia + Intel (again, not meaning to bash AMD-ATI, just state of "business")... For several reasons, part the graphics support and performance, part for power consumption and battery life, etc.

OMG how much you and people like you disgust me. You nullify peoples issues because they "should have known" that ATI sucks? So then it's Ok to suck, and people should stop complaining?
No, whatever history a company has, doesn't make it Ok to give complete crap support to their customers, whatever they say "Linux is small" yada yada.

Fuck that BS. Linux is amazingly big for its enormously powerful enemies who actively make things impossible for open communities by locking in specs from hw providers, and ATI et al keep playing these major sw corps' game. And all you can come up with is "Your laptop was sold with Windows" yada yada...

Eat sh*t.

[opera]

I've done some benchmarks now, and it doesn't look good.
Here are the results:
8.40:
doom 3 = 35.7 fps (timedemo demo1 usecache)
nexuiz = (min/avg/max) 5/17/44 fps (-benchmark demos/demo1)

8.41:
doom 3 = 41.7 fps
nexuiz = 8/28/50

Quake 4 still runs in Ultra-Crappy mode, despite any settings that I choose, so a benchmark for it is pointless.

Every benchmark was ran twice, with maximum visual quality except for AA and vsync (ex: Ultra Quality in doom3) at 1280x1024. Nexuiz was ran without VBOs or Offset mapping, in order for it to work on 8.40.

The two games really are more playable when a lot of action is happening on screen, but the performance improvement is really not impressing. I wanted to do some more testing, but, after these two benchmarks, I'm really not going to put any more effort in proving that 8.41 is gold.

I hope x1k and HD users got a lot more from this new code base. I'll just go back to 8.40 (in order to play x2) and wait for the next release...

EDIT: at some point I got the screen "corruption" that a few people were talking about, but a quick switch to tty1 and back fixed it

Hi there Mr Bitch. Where's my hail? You told me you'd hail me, now since it's clear that the driver sucks...
And don't give me the only-r600 BS. No one was told that this driver would be r600-ONLY. Just that it would be "focused" on one thing, doesn't mean everything else will suck. At least Mr Michael didn't clarify that while praising and hyping this crap driver release.

I consider myself hailed.

[opera]

Just like every other version of the crap ATI driver with a modern kernel, and I'm getting used to it:

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x6d) [0x48586d]
1: /lib/libc.so.6 [0x2b1dbeae36a0]
2: /usr/bin/X(NumMotionEvents+0x12) [0x447832]
3: /usr/bin/X(CreateConnectionBlock+0x53) [0x439543]
4: /usr/bin/X(main+0x65a) [0x43a08a]
5: /lib/libc.so.6(__libc_start_main+0xf4) [0x2b1dbeacfb44]
6: /usr/bin/X(FontFileCompleteXLFD+0x229) [0x439189]

Fatal server error:
Caught signal 11.  Server aborting

[Xipeos]

Hi there Mr Bitch. Where's my hail? You told me you'd hail me, now since it's clear that the driver sucks...
And don't give me the only-r600 BS. No one was told that this driver would be r600-ONLY. Just that it would be "focused" on one thing, doesn't mean everything else will suck. At least Mr Michael didn't clarify that while praising and hyping this crap driver release.

I consider myself hailed.

You deserve no hail, as the driver is not even close to what you predicted. There's obvious improvement (did you miss the part on how "The two games really are more playable when a lot of action is happening on screen" ?).

[carpman]

Fuck that BS. Linux is amazingly big for its enormously powerful enemies who actively make things impossible for open communities by locking in specs from hw providers, and ATI et al keep playing these major sw corps' game. And all you can come up with is "Your laptop was sold with Windows" yada yada...

Eat sh*t.

None of that absolves you from doing research before you buy something. You can complain about it not working, but you also have to remember that you should have checked in the first place as well.

[zkyez]

Just like every other version of the crap ATI driver with a modern kernel, and I'm getting used to it:

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x6d) [0x48586d]
1: /lib/libc.so.6 [0x2b1dbeae36a0]
2: /usr/bin/X(NumMotionEvents+0x12) [0x447832]
3: /usr/bin/X(CreateConnectionBlock+0x53) [0x439543]
4: /usr/bin/X(main+0x65a) [0x43a08a]
5: /lib/libc.so.6(__libc_start_main+0xf4) [0x2b1dbeacfb44]
6: /usr/bin/X(FontFileCompleteXLFD+0x229) [0x439189]

Fatal server error:
Caught signal 11.  Server aborting

Linux florin-laptop 2.6.22.5-12-default #1 SMP 2007/09/04 12:37:49 UTC i686 i686 i386 GNU/Linux

i do not see the error you are reporting.

[carpman]

Just like every other version of the crap ATI driver with a modern kernel, and I'm getting used to it:

Backtrace:
0: /usr/bin/X(xf86SigHandler+0x6d) [0x48586d]
1: /lib/libc.so.6 [0x2b1dbeae36a0]
2: /usr/bin/X(NumMotionEvents+0x12) [0x447832]
3: /usr/bin/X(CreateConnectionBlock+0x53) [0x439543]
4: /usr/bin/X(main+0x65a) [0x43a08a]
5: /lib/libc.so.6(__libc_start_main+0xf4) [0x2b1dbeacfb44]
6: /usr/bin/X(FontFileCompleteXLFD+0x229) [0x439189]

Fatal server error:
Caught signal 11.  Server aborting

What in that backtrace makes you point a finger at AMD?

[Michael]

Opera, stop with your useless arguing with Xipeos and others... I've already warned you.