AMD Catalyst 8.5 For Linux

[ckxdce]

On my ATI Radeon X1800 XT, Catalyst 8.5 works as fine as its predecessors, but my two favorite issues still exist:

First, the GPU fan is running at full speed and neither aticonfig's power options nor atipower could help.

Second - and I seem to be the only one who cares about that one - the OpenGL pipeline of Java2D still doesn't work with this version of Catalyst so Java falls back to non-accelerated GUI rendering.

[oyvind]

Basically a solid release for me ..

I'm getting about a 6, on a scale from 1-10, on my very personal feel-good'o'meter on this release, after going through my usual rounds of "new fglrx release testing". To sum it up, apart from the slightly embarassing watermark-issue (thanks for the hotfix: http://www.phoronix.com/forums/showp...6&postcount=13), Kano !), it works just as well as 8.4 for me, but with no radically visible improvements.

2D
I'm curious about the 2D performance, what exactly has been improved, XRender-acceleration in general ? Or just some specific thing that only benefits a few scenarios, like the KDE translucency issue ? Can anyone shed some light on this ? Specifics, please .

Edit/added:
Actually, something has visibly improved. In a plain no-Compiz 2D desktop, I used to have problems with slow scrolling/response only on certain web sites in Firefox, but don't know the exact reason(s) for it. Typically it was sites with contents overlayed on a fixed position back-ground, using transparency effects and other fancy CSS-stuff, etc. Those problems seem to be gone, yay ! That's a very welcome improvement.

I have always used the "XaaNoOffscreenPixmaps" "1" option before, to fix some of the cases of slow scrolling, but I don't seem to need that anymore. Glad to get rid of it .

Hmm, Flash player performance seems better, but this may very well be Mr. Placebo-effect talking ..

Please share, if you find that you have similar experiences with this new release.

S2RAM
As a laptop-user, I'm still *very* disappointed that suspend to RAM still sucks and doesn't work at all on Ubuntu Hardy. Maybe I just *never* should have gotten this Lenovo laptop of mine; it has other quirks as well. Or maybe it's just that the new Ubuntu-Hardy kernel really dislikes the hardware ACPI-implementation. Da****. .

Edit/added:
Although Suspend to RAM is broken for me, hibernate/software suspend actually works. But I used to have problems with lowered 3D-performance after having resumed from this state (about 100 less FPS in glxgears). That is no longer the case. So another thing fixed in this release, it seems.

Compiz
The only thing keeping me from enabling Compiz is an overall sluggishness in almost all scrolling-situations, most notably Firefox. It *has* gotten better in the last fglrx-releases, but it's still painful ! However, as I understand, this is an architectural limitation of AIGLX (XCopyArea not accelerated, and heavy texture-copying going on when used on large rectangles), so I can only hope for better times in the future of Linux desktop compositing. Scrolling in Compiz also kinda sucks on another lower-spec Intel-based laptop of mine (though not as bad). I'm not blaming this solely on fglrx. I might be a demanding user, but as long as plain 2D is so incredibly fast compared to when running under Compiz, I simply cannot live with it, it's too vexing.

XVideo/TexturedVideo
TexturedVideo continues to work properly without any tearing (the day AMD breaks TexturedVideo/Xvideo again, is the day I'm selling my laptop). Still only works in full-screen when running under Compiz, though (with un-redirect full-screen windows option enabled).

VSYNC
I've got "Wait for vertial refresh" set to "Off, unless application specifies" in the AMD control panel and "Sync to VBLANK" enabled in Compiz, but tearing is still present.

The end
Although 8.5 is definitely not perfect, the trend has been positive for the last few releases for *me*, can't really speak for anyone else. A colleague of mine has an X1600-based laptop, I'll hear from him tomorrow on how the new driver works, but he has basically had less problems than me, in general (I hear my X1400 is a somewhat quirky beast, but can't remember where I got that from).

This little quote, from the Catalyst 8.5 Phoronix article, is also cool: "However, the best is yet to come so stay tuned to changes for the forthcoming Linux Catalyst releases.".

My hardware: Lenovo Z61m laptop, ATI Radeon Mobile X1400, 2GiB RAM, Intel Core Duo 2GHz, Ubuntu Hardy x86 32bit.

[Melcar]

The Vsync options don't work. The only choices seem to be on or off; the in between options do not do what the description says and either force it on or off globally. It's really annoying, but even the Windows driver suffer from a similar problem.
The tearing is annoying, since it happens in EVERYTHING rendering in Xv and/or OpenGL. Games, screensavers, movies, anything. It sucks.
Oh, and the signed release should be up now, so no need for the hotfix.

[chrisr]

Not perfect, but an improvement over 8.4

I finally have Warcraft running fullscreen on my laptop, at 1680x1050!!

However, the Fedora build .spec file template is broken. The latest changelog entry is incorrectly formatted, and the new dependency on compat-libstdc++ doesn't exist anyway! I had to hack it to say "compat-libstdc++-33" instead, which is probably specific to Fedora 8.

The .spec file also forgets to package the new libatiadlxx.so object.

I'm also getting some weirdness in my Xorg.0.log file:

(--) fglrx(0): Chipset: "ATI MOBILITY FireGL V5250" (Chipset = 0x71d4)
(--) fglrx(0): (PciSubVendor = 0x17aa, PciSubDevice = 0x20a4)
(--) fglrx(0): board vendor info: third party graphics adapter - NOT original AT
I
(--) fglrx(0): Linear framebuffer (phys) at 0xd0000000
(--) fglrx(0): MMIO registers at 0xee100000
(==) fglrx(0): ROM-BIOS at 0x000c0000
(II) fglrx(0): Primary V_BIOS segment is: 0xc000
(II) Loading sub module "vbe"
(II) LoadModule: "vbe"
(II) Reloading /usr/lib/xorg/modules//libvbe.so
(EE) fglrx(0): vm86() syscall generated signal 11.
(II) fglrx(0): EAX=0x0000004f, EBX=0x00000000, ECX=0x00000000, EDX=0x00000000
(II) fglrx(0): ESP=0x00001000, EBP=0x00000000, ESI=0x00000000, EDI=0x00002000
(II) fglrx(0): CS=0x0000, SS=0x0100, DS=0x0040, ES=0x0000, FS=0x0000, GS=0x0000
(II) fglrx(0): EIP=0x00000600, EFLAGS=0x00033202
(II) fglrx(0): code at 0x00000600:
f4 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
(II) fglrx(0): VESA BIOS detected
(II) fglrx(0): VESA VBE Version 3.0
(II) fglrx(0): VESA VBE Total Mem: 16384 kB
(II) fglrx(0): VESA VBE OEM: ATI ATOMBIOS
(II) fglrx(0): VESA VBE OEM Software Rev: 9.12
(II) fglrx(0): VESA VBE OEM Vendor: (C) 1988-2005, ATI Technologies Inc.
(II) fglrx(0): VESA VBE OEM Product: M66GL
(II) fglrx(0): VESA VBE OEM Product Rev: 01.00
(EE) fglrx(0): vm86() syscall generated signal 11.
(II) fglrx(0): EAX=0x0000004f, EBX=0x00004116, ECX=0x00000000, EDX=0x00000000
(II) fglrx(0): ESP=0x00001000, EBP=0x00000000, ESI=0x00000000, EDI=0x00002000
(II) fglrx(0): CS=0x0000, SS=0x0100, DS=0x0040, ES=0x0000, FS=0x0000, GS=0x0000
(II) fglrx(0): EIP=0x00000600, EFLAGS=0x00033202
(II) fglrx(0): code at 0x00000600:
f4 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
(II) fglrx(0): ATI Video BIOS revision 9 or later detected
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 8, (OK)
drmOpenByBusid: Searching for BusID PCI:1:0:0
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 8, (OK)
drmOpenByBusid: drmOpenMinor returns 8
drmOpenByBusid: drmGetBusid reports PCI:1:0:0
(II) Loading sub module "fglrxdrm"
(II) LoadModule: "fglrxdrm"
(II) Loading /usr/lib/xorg/modules/linux//libfglrxdrm.so
(II) Module fglrxdrm: vendor="FireGL - ATI Technologies Inc."
compiled for 7.1.0, module version = 8.49.7
ABI class: X.Org Server Extension, version 0.3
(II) fglrx(0): Using adapter: 1:0.0.
(II) fglrx(0): [FB] Find the MC FB aperturs range(MCFBBase = 0xc0000000, MCFBSize = 0x10000000)
(--) fglrx(0): VideoRAM: 262144 kByte, Type: DDR3
(II) fglrx(0): PCIE card detected
(WW) fglrx(0): board is an unknown third party board, chipset is supported
(EE) fglrx(0): vm86() syscall generated signal 11.
(II) fglrx(0): EAX=0x0000004f, EBX=0x00000003, ECX=0x00000000, EDX=0x00000000
(II) fglrx(0): ESP=0x00001000, EBP=0x00000000, ESI=0x00000000, EDI=0x00002000
(II) fglrx(0): CS=0x0000, SS=0x0100, DS=0x0040, ES=0x0000, FS=0x0000, GS=0x0000
(II) fglrx(0): EIP=0x00000600, EFLAGS=0x00033202
(II) fglrx(0): code at 0x00000600:
f4 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
(II) Loading sub module "ddc"
(II) LoadModule: "ddc"(II) Module already built-in
(II) fglrx(0): Connected Display1: LCD on internal LVDS [lvds]
(II) fglrx(0): Display1 EDID data ---------------------------
(II) fglrx(0): Manufacturer: IBM Model: 2887 Serial#: 0
(II) fglrx(0): Year: 2005 Week: 0
(II) fglrx(0): EDID Version: 1.3
(II) fglrx(0): Digital Display Input
(II) fglrx(0): Max H-Image Size [cm]: horiz.: 33 vert.: 21
(II) fglrx(0): Gamma: 2.20

Is anyone else getting those signal 11 errors?

This is on a Lenovo T60p, with a V5250 graphics chip. The laptop is running Fedora 8.

[linzerd]

MAchine Specs:
Intel D101GGC MOBO
1 Gb generic RAM Memory
Intel Pentium D 805 2.66 GHz
Integrated Xpress 200 Graphic Card (256 MB of RAM Allocated)
Seagate Barracuda 7200.10 160 GB Hard Drive
Distro: Ubuntu Hardy Heron

Main Issues with this release:

1- VLC still crashes with OpenGL output, throws this message:

[00000343] main private error: option glx-shm does not exist
libGL error: drmGetMagic failed
libGL error: reverting to (slow) indirect rendering
X Error of failed request: BadMatch (invalid parameter attributes)
Major opcode of failed request: 143 (GLX)
Minor opcode of failed request: 5 (X_GLXMakeCurrent)
Serial number of failed request: 58
Current serial number in output stream: 58

2- Scrambling video output bug with VideoOverlay option turned on in xorg.conf and XVideo output was partially resolved. Now it only happens rarely in fullscreen, but in windowed video playback, it is still seen.

[rbmorse]

I couldn't make it work on Ubuntu 8.04/32-bit. I either ended up with a "could not locate libGL.so.1" error or the FGLRX DRM wouldn't load and I ended up with MESA.

Got tired of fscking with it and went back to 8.03 from the Ubuntu repository via restricted driver manager.

[karthikrg]

ne ideas when fglrx will get around to supporting dri2 so that we can enjoy non-flickering videos with compiz???

[Moronix]

Hi,

for the people which are getting:

FATAL: modpost: GPL-incompatible module fglrx.ko uses GPL-only symbol '__rcu_read_lock'

the patch is:

237c237
< MODULE_LICENSE("Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY");
---
> MODULE_LICENSE("GPL");

for the file /lib/modules/fglrx/build_mod/firegl_public.c

then you have to rerun sh make.sh in /lib/modules/fglrx/build_mod/ and sh make_install.sh in /lib/modules/fglrx/

No, this doesn't work. When trying to install

FATAL: Error inserting fglrx (/lib/modules/2.6.25.4-r/kernel/drivers/char/drm/fglrx.ko): Unknown symbol in module, or unknown parameter (see dmesg)
failed.

WTF is up with this wonky driver anyways?

[Thetargos]

I finally have Warcraft running fullscreen on my laptop, at 1680x1050!!

However, the Fedora build .spec file template is broken. The latest changelog entry is incorrectly formatted, and the new dependency on compat-libstdc++ doesn't exist anyway! I had to hack it to say "compat-libstdc++-33" instead, which is probably specific to Fedora 8.

Hi, I also run F8 on my laptop and I'm intrigued for what exactly do you mean by Warcraft. Do you mean Warcraft III, Warcraft II, World of Warcraft or what Warcraft game?

I was about to ask if this driver resolves the issue of forcing Wine to do all 2D DirectX acceleration through OpenGL with the registry key HKCU/Wine/Direct3D/DirectDrawRenderer set to opengl (very useful for Starcraft and Warcraft II BNE as well as many other 2D DirectX 3 to 5 era games)

[oblivious_maximus]

Broke out my Sapphire HD3650 512MB to try the new driver.

The good:
-2d performance is definitely much improved. The visible delay (in KDE 3.5 for me) when drawing what was formerly underneath a moved window is no longer a visible delay. Moving windows seems much more fluid. 2d is now on par with my 8600GT & 169.12 imho, at least judging from the 2 minutes I used 8-5 before testing a logout.

The bad:
-At least with the packages I built for Debian testing from the .run archive(w/ --buildpkg Debian/testing), you still have to edit /etc/ati/authatieventsd.sh to point to /var/run/xauth yourself. Even so, for me this still doesn't make any difference, I can't logout of KDE without a massive lockup that even Raising Elephants with the Magic SysRq key cannot cure.

-Some weird graphical corruption during the Doom3 timedemo. A bunch of short vertical lines in a row near the bottom left of the screen, slightly obscuring the HUD, and a large black triangle where I think there should have been shotgun muzzle flash near the beginning of the demo.

The bad that probably isn't relevant to anyone but me:
-I'm still not using the ATI card I bought months ago because of fglrx And I realized with the last driver that the 3650 is much faster than the Nvidia card I'm using, at least as far as that doom3 timedemo is concerned. And it's a glorified paperweight right now

-Every time I plug this card in I notice the HDCP sticker on it and wonder "is this the reason why I still can't use this card??" Why must every card on the market be hamstrung like this? Another prime example of how Microsoft's monopoly is destroying society. OK so maybe I exaggerate a little. But ONLY A LITTLE. I want to see a line of cards whose selling point is that they inlude no protected video path BS and that they cannot decrypt "protected content". At least then we could directly see just how much trouble HDCP causes by comparing HDCP and non-HDCP cards of the otherwise same model. I'm basically ranting at this point... Sorry, I'm frustrated.

edit: oh yeah:
2.6.25-3-686 kernel, compiled from Debian's kernel source pakcage, 32 bit Debian testing. Built fglrx packages from the .run archive. No issues building or installing those packages, or the kernel module (using module-assistant). Asus M3A, X2 4400+, 4GB (2x2GB) DDR2-800.

Are there any specific kernel config options that are required (either builtin or as a module) for fglrx to properly function? Could there be something I've not compiled (I remove a lot of hardware I don't/won't have) or compiled as a module when it should be a builtin? I've pretty much ruled this out with previous drivers by trying to use stock Debian kernel images, but I always got the same logout issues as with the kernels I've built myself, but you never know.