AMD Catalyst 8.5 For Linux

[wojak]

hi,
I can't change refresh rate to 50 Hz. I have 60Hz and I don't have fullscreen on my TV connected dvi-hdmi. When xine finish play a film my X is broken.
Jun 4 20:46:58 myhost X[2932]: segfault at f000eeff eip 08192890 esp bfa79770 error 5
Jun 4 20:46:59 myhost kdm[2922]: X server for display :0 terminated unexpectedly
Jun 4 20:46:59 myhost [fglrx] PCIe has already been initialized. Reinitializing ...
Jun 4 20:46:59 myhost [fglrx] Reserve Block - 0 offset = 0Xfffb000 length = 0X5000
Jun 4 20:46:59 myhost [fglrx] Reserve Block - 1 offset = 0X0 length = 0X1000000
Jun 4 20:46:59 myhost [fglrx] Reserve Block - 2 offset = 0Xff7b000 length = 0X80000
Jun 4 20:46:59 myhost [fglrx] interrupt source 10000000 successfully enabled
Jun 4 20:46:59 myhost [fglrx] enable ID = 0x00000008
Jun 4 20:46:59 myhost [fglrx] Receive enable interrupt message with irqEnableMask: 10000000
Jun 4 20:48:59 myhost kdm: :0[3340]: Hung in XOpenDisplay(:0), aborting
Jun 4 20:48:59 myhost kdm: :0[3340]: Cannot connect to :0, giving up
Jun 4 20:48:59 myhost kdm[2922]: Display :0 cannot be opened

[Phlogiston]

Anyone able to use this driver with a recent .26 kernel?

[Redeeman]

#193 Kano:
nah, i didnt really know how good performance would be, though i was quite aware that a 3450 would not be highend

i just wanted to know if fglrx is working as it should.

to be honest i dont much care about that either, what i wanted from the 3450 was something that could be used right away with fglrx, and then test radeonhd progress on

[Kano]

I guess he thought that any new card must be faster than an old one, but that was a older midrange NV card you compare against an entry level card (at least powerfull for h264 with Win). That's really a but unfair *g*

[bridgman]

performance isnt terribly good, alot worse on this c2d 3ghz box than my amd64 2ghz with 6600gt agp. I didnt expect HD 3450 to be fast, but can you tell me if im hitting around the correct performance? (im getting ~8-10fps in wow @ 1024x768 almost full details).

I haven't found an apples-to-apples benchmark but at first glance I would expect the 6600gt to be faster than a 3450. The 6600gt has 8 pixel pipes and 3 vertex pipes, with a 128-bit memory bus, while the 3450 is more like a 4-pipe chip in terms of ROPs and shaders and has a 64-bit memory bus AFAIK. Clocks are probably a bit higher on 3450 but not enough to make up for 2x the hardware in the 6600GT.

Not sure what numbers to expect in wow on a 3450 but if you're seeing roughly 1/2 the performance that would sound about right. Again, this is a guess based on my understanding of hw internals, not based on benchies or real world experience.

[yorik]

Hi guys,
I just reinstalled a clean ubuntu hardy on my computer (laptop, 64bit os, with an xpress 200 chip) and catalyst 8.5 (was running gutsy with fglrx 8.40 before)... Everything seems fine, everything running normally, but blender is incredibly slow... Not really what's pure 3D (moving big quantity of objects on screen, etc...) but everytime something like menus, texts, dotted lines appear on screen, or if you change your selection of objects, it gets veeery slow.

Has anyone experienced something like this?

Running the same blender version with the mesa opengl lib statically linked runs fine, so it doesn't seem to be a problem with blender... The same blender works still perfectly on gutsy with fglrx 8.40...

Everything fglrx-related seems perfectly correct, aiglx & composite are disabled, direct rendering is enabled, fglrx xorg driver loads perfectly, as well as the module kernel, both are the correct 8.49 version, i tried to switch on/off a couple of xorg.conf options, doesn't seem to change anything... Other 3d apps seem to run fine too...

Anyone has an idea? Something I could try?

Thanks

Yorik

[Extreme Coder]

The issue is that everything *below* the DRM stuff needs to be closed source and difficult to hack (you need to protect the decrypted stuff) so you end up with most of the stack being closed source.

Even seemingly good targets for open source like display don't really work out because robust output protection is an important part of the DRM solution.

Just so it's clear, we don't enable any DRM stuff in our Linux Catalyst drivers today.



I think we have some Mandriva users in the beta test program; will check.

Ok, in any case, I'm available for testing at any time

[bridgman]

I've never understood why the opensource drivers couldn't be written to accept the equivalent of a closed-source DRM module.

The issue is that everything *below* the DRM stuff needs to be closed source and difficult to hack (you need to protect the decrypted stuff) so you end up with most of the stack being closed source. Even seemingly good targets for open source like display don't really work out because robust output protection is an important part of the DRM solution.

There's no question that "just a few bits" are affected by DRM, but "all the stuff in between" needs to be secure as well and you very quickly end up with a 95% closed source driver. I'm guessing at the 95%, but it's closer to 95 than 50.

Just so it's clear, we don't enable any DRM stuff in our Linux Catalyst drivers today.

So no Mandriva? Let me test Mandriva for ya

I think we have some Mandriva users in the beta test program; will check.

[Extreme Coder]

I don't think we test on Debian today -- mostly RHEL, OpenSuSE (SLED upstream) and we're starting to get some coverage on Ubuntu, which should help a bit with Debian.

So no Mandriva?
Let me test Mandriva for ya :P

[NaterGator]

I've never understood why the opensource drivers couldn't be written to accept the equivalent of a closed-source DRM module.

For the most part the entire rest of the linux architecture is open source and it's not really like DRM is going to ever be able to prevent unlicensed use (look at the FairUse4WM on closed source windows as a prime example). That doesn't mean AMD/ATI can't appease content owners though and at least provide them a 'secure' closed-source branch of the open source driver.