Catalyst 9.2 Released, Still Fails To Deliver XvBA

[Odin]

If you don't remove the old one, you don't have to execute aticonfig...at least I did it like that and haven't got any problems. :/

[elsie]

Has anyone else experienced X crashing/system freezes with XV video playback?
Crashes computer everytime I go fullscreen. This happened to me on 9.1 and 9.2.
Doesn't make a difference if compiz is enabled, either way computer crashes every time I play a video.
I'm using ubuntu 8.04 and x1600 mobile.

[oyvind]

Has anyone else experienced X crashing/system freezes with XV video playback?
Crashes computer everytime I go fullscreen. This happened to me on 9.1 and 9.2.
Doesn't make a difference if compiz is enabled, either way computer crashes every time I play a video.
I'm using ubuntu 8.04 and x1600 mobile.

Same here. X1400 mobile. It seems AMD is forgetting about older/weaker cards wrt. video playback. They should provide a cheaper XV playback method in their driver for older gen cards. I really don't care if it's simpler and looks marginally worse than their fancy adaptive filtering, I just want it to work. Period. Also, I find it really funny to read this:
http://ati.amd.com/products/Mobility...400/index.html

Especially the section about "Avivo™ video and display perfection" .. Ah, the irony.

[bridgman]

With respect, I don't think that page talks about running with the additional GPU overhead of a compositor layered on top of the video acceleration. Realistically, I think the open source drivers will probably become that "simpler option" for older/weaker cards running under a compositor.

I *think* the "Unredirect Fullscreen Windows" option should let the video accel code go straight to the front buffer and skip the extra overhead.

[Moronix]

With respect, I don't think that page talks about running with the additional GPU overhead of a compositor layered on top of the video acceleration. Realistically, I think the open source drivers will probably become that "simpler option" for older/weaker cards running under a compositor.

I *think* the "Unredirect Fullscreen Windows" option should let the video accel code go straight to the front buffer and skip the extra overhead.

There is Xv flicker without composite on older cards. It's unusual and happens with cpu spikes more. Xv implementation is extremely wobbly on that hardware. In fact, with composite on said cards, Xv won't even show so it's moot. I'm talking about things like the older Xpress series which worse yet doesn't seem like radeonhd will support properly.

[dungeon]

Why just ATI don't do one good drop on support for pre HD cards, that way fglrx could be much easier to maintain/develop?

[oyvind]

With respect, I don't think that page talks about running with the additional GPU overhead of a compositor layered on top of the video acceleration.

I *think* the "Unredirect Fullscreen Windows" option should let the video accel code go straight to the front buffer and skip the extra overhead.

Yes. That page talks about plain Windows XP, I would assume. Though the words are really shiny and HD is definitely part of them. I'm just venting and being bitter because of all the troubles I've had with Linux+fglrx over the last couple of years, most of the time in a non-composited environment .

Unredirect-fullscreen is very quirky and moody, and does not always work (i.e. many things simply go fullscreen without being unredirected). This might be a Compiz-problem, but it's been like that for a long time.

[Melcar]

Why just ATI don't do one good drop on support for pre HD cards, that way fglrx could be much easier to maintain/develop?

Then people would complain why their +5yr. card is not supported anymore . What I think they should do is offer support for pre-HD cards up to a certain release and then stop, with future driver release dropping said support and focusing on newer cards. That "legacy" driver should then only be updated so that it keeps working with newer kernels and xservers. If you still want new features on your ancient card, then look at the open source drivers or something (which is what I think people should do anyway).
Unfortunately, people are likely to also complain if such a thing were done, like when active support for pre-9500 cards was dropped... "why is my card not actively supported anymore?". People will never be happy with the driver.

[oyvind]

Why just ATI don't do one good drop on support for pre HD cards, that way fglrx could be much easier to maintain/develop?

Even though I am an owner of a pre-HD-era card, I would tend to agree with you. If support becomes more theoretical than real/practical, it would be better to officially drop it. And since AMD is doing many good things for open source development of drivers for their hardware, they can actually pull that off without too much grief. Give the open alternatives some more time to mature, and fglrx won't be necessary for older cards (hopefully).

[Moronix]

Then people would complain why their +5yr. card is not supported anymore . What I think they should do is offer support for pre-HD cards up to a certain release and then stop, with future driver release dropping said support and focusing on newer cards. That "legacy" driver should then only be updated so that it keeps working with newer kernels and xservers. If you still want new features on your ancient card, then look at the open source drivers or something (which is what I think people should do anyway).
Unfortunately, people are likely to also complain if such a thing were done, like when active support for pre-9500 cards was dropped... "why is my card not actively supported anymore?". People will never be happy with the driver.

What a load of bollocks. It would be perfectly sufficient to release a single driver that runs all the capabilities of the card and the feature software that existed at it's release, this that you propose Nvidia has done for a long time already. This is not something that has been done ever in fglrx. So don't try to blame the users for having bought them at one point. They're not the ones who implied a fully working driver would exist either.