AMD Catalyst 8.12 Linux Driver

[energyman]

no tool. just let it run hot, cool it down (shut off), run hot, shut down .... after a while either your PSU fails or the card

[RealNC]

You can just stick something in the GPU's fan to stop it and then boot Vista. It will explode in no time.

Note: Works with all chips, not just NVidia

[makomk]

Apparently, it still sucks, and more video acceleration standards is kind of silly, I wish they'd just stick to one good open standard if possible and be done with it but oh well. Actually interested in Intel's video efforts for a good HTPC solution (since Intel sucks at 3D still). They were also the ones really "going after" tearing, or so they said, but of course it's one thing to talk about it...

Actually, I've been looking at the XvBA libraries, and it looks like it's intended to be a reasonably sensible extension to XvMC, loosely based on Microsoft's DXVA. I'm guessing, from glancing over the disassembled code, the idea is that you create an XvMC context with XvMCCreateContext and surfaces with XvMCCreateSurface as usual, then use XvBA. More specifically, XvBAStartDecodePicture, XvBADecodePicture, and XvBAEndDecodePicture appear to be roughly equivalent to IDirectXVideoDecoder::BeginFrame, Execute, and EndFrame under Windows, but with XvMC surfaces. (Not sure where XvBACreateDecodeBuffers and XvBADestroyDecodeBuffers fit in.)

Of course, I can't really test this - aside from the lack of headers, my Radeon HD 2600 XT only supports UVD 1. (MPlayer won't even try, and when I attempt to call XvMCCreateContext my test app segfaults.) Should hopefully be fairly easy to use once the headers are out, though.

Also, you've got to remember that when it was created, VAAPI was essentially vaporware (it still is) and VDAPU hadn't been released yet. This probably made the most sense at the time.

Edit: Hmmm... looks like this changed in 8.12, which seems to seperate XvBA and XvMC. (Oh, and removes XvMC from libXvBAW.so.) I wonder when they'll actually finalise the spec?

[RealNC]

Btw, 8.12 is still slow as molasses with ANY compositing manager (Compiz, KWin, xfwm, just anything.)

I get the feeling they're never going to fix that one.

[pedepy]

well this is just a bad driver ... you could never sell a single ati card with that kind of performance.

every release seems to break something that once worked to 'introduce' a shaky 'new feature'.. That much constant breakage is a sure sign of an *utterly* bad coding style.

And at least with open source projects, when it breaks, you usually know why and what to expect. fglrx is just retarded..

[Yfrwlf]

And at least with open source projects, when it breaks, you usually know why and what to expect. fglrx is just retarded..

If they really wanted to work with the community, they'd abandon it or open source it, but of course both are "difficult" to do from a CEO's perspective. :P

Oh well, let the competition continue, but still, them along with Intel, are more "open source friendly" than Nvidia, but that may largely be because of competition, and this open sourcing effort is to try to get community help so they can hopefully catch up to them. It's a smart card to play for that reason, as well as all the other basic advantages of open source of course.

It's an effort they've helped convince me to support even if their documents haven't really created any super capable open source driver yet and even though they don't seem to be really interested in working on the actual open source code. Why would they when they still have their closed driver around?

[PuckPoltergeist]

If they really wanted to work with the community, they'd abandon it or open source it, but of course both are "difficult" to do from a CEO's perspective. :P

As long as the closed source drivers shows the better performance, supports more features (OpenGL > 1.4, powermanagement) and supports the existing cards better, abandoning would be a really bad decision. And opening isn't a opionon cause of IP reasons, I believe.

[cruiseoveride]

what IP? AMD is worth nothing now, they might as well start selling pizzas.

[cruiseoveride]

wait... second that.. if AMD sold pizzas it would be red, look really good, taste like crap, and need 14.345mm long by 7.23mm wide, 3 prong forks, otherwise you cant eat it at all.

ATi should just come suck my balls.

[PuckPoltergeist]

wait... second that.. if AMD sold pizzas it would be red, look really good, taste like crap, and need 14.345mm long by 7.23mm wide, 3 prong forks, otherwise you cant eat it at all.

ATi should just come suck my balls.

No one forces you to buy ATI-Hardware. So don't whine here cause you bought 'crap'.