Tearing on ATI 5xxx cards

[emergence]

Hey I'm using a 5770 under the same os and driver version as emergence, and I can confirm that I have been suffering from basically the same tearing problems you have, and believe me I feel your pain....RealNC's explanation of the indirect rendering done by compiz, and fglrx's lack of support for vsync with indirect rendering is like a revelation to me, I finally now understand what it is that is causing my desktop to look worse than my 4 year old integrated intel graphics on my dell inspiron.....however I would rather have it fixed...(on that note, bridgman, is there any kind of news on when exactly we might be getting evergreen support on the opensource drivers? When I bought the cardit never occured to me that I would still be waiting on the the opensource drivers like 7 months later).

The unredirect fullscreen windows is is the only way I can get tear free movies using opengl video out (which makes complete sense to me now after RealNC's explanation)...the new 2d acceleration has reduced some of the tearing for me...but only be a very minor amount, it still gives me hope for future releases though.

Thanks for the confirmation. I'm glad I'm not alone. The tearing creates a very bad taste in my mouth when using linux. It's just so annoying that everyday I try to search and see if I can find a solution to the problem. Googling "ATI linux tearing" clicking on "past 24 hours", just to find that post that will solve my problems.

[bridgman]

Talking about chronniff's suspend to ram regression, not tearing. Are you seeing the same STR problem ?

[chronniff]

Yeah, turns out it happens on the standard kernel for me too......in case anyone is curious the first line of the error I get for the preemptable kernel in dmesg is:
BUG: using smp_processor_id() in preemptible [00000000] code: Xorg/1273

[chronniff]

ok, I just tested the suspend on the genric kernel, and am having the same behavior I was getting in 10.5, x is slowing down, and cairo-dock starts behaving fine, but eventually clicking on it doesn't do anything, and when I kill it, and then try to launch it again, it doesn't even show up, although it does have a pid....I think there is a problem with opengl, not cairo dock...I just tried to launch a video with mplayer gl video out, and it doesn't play, just a black mplayer screen, but using xv works (although with awful tearing as usual)

[chronniff]

the freezing on wakeup seems to be a problem with the preemptable kernel...although eventually I think generic kernel will freeze, but I am too lazy to stay in the generic kernel and wait....also the preemptable kernel doesn't freeze everytime....sometimes it does exactly what the generic kernel is doing right now, and eventually it freezes

[chronniff]

after testing out all the kernels I have a feeling the longer I leave my box suspended to ram the more likely it is to freeze on wakeup, because they all wake up ( with gl issues ) when I let it sit for a few minutes...but when I wake it up in the morning after is suspended to ram for the night, it almost always freezes before I can login...and that is my final analysis....don't know if it helps anyone, but it satisfied my curiosity....

[BlackStar]

Thanks for the confirmation. I'm glad I'm not alone. The tearing creates a very bad taste in my mouth when using linux. It's just so annoying that everyday I try to search and see if I can find a solution to the problem. Googling "ATI linux tearing" clicking on "past 24 hours", just to find that post that will solve my problems.

For what it's worth, this is a general limitation that affects all drivers, not just fglrx (there is no way to get a reliably tear-free desktop under compiz, even with nvidia). AFAIK, this requires a new X11 protocol which is still being worked on.

[RealNC]

For what it's worth, this is a general limitation that affects all drivers, not just fglrx (there is no way to get a reliably tear-free desktop under compiz, even with nvidia). AFAIK, this requires a new X11 protocol which is still being worked on.

Nope. It works OK with the open source Radeon drivers when using KMS/DRI2. And it works with the proprietary NVidia drivers. I'm not sure about Intel and Nouveau.

[BlackStar]

Nope. It works OK with the open source Radeon drivers when using KMS/DRI2. And it works with the proprietary NVidia drivers. I'm not sure about Intel and Nouveau.

Tear-free video most certainly doesn't work with the prioprietary nvidia drivers when compiz is enabled. I'm reproducing this as I write.

Disable compiz and video aligns itself, while the X server starts tearing instead.

Radeon KMS seems to work better on my R500 but it's still not 100% tear-free.

[Kano]

I have got a GeForce 8800 GTS 512, that has got a G92 chip and VP2, what you mean are the G80 variants GeForce 8800 Ultra, 8800 GTX, 8800 GTS (320/640 MB). Those are really rare as the 9800 GTX replace those. The latest VP4 engine is in DX10.1/11 cards. Don't use the term UVD for Nv cards, that's incorrect.

And if you really want to try oss video accelleration and you are happy with mplayer vaapi then get something with intel hd graphics. I have got no doubts that it will improve, h264 is working today better with it than with xvba. xbmc is just a bit choppy and vlc green, lets see when those problems are solved. But xvba is even worse. intel hd graphics can decode h264 l5.1.

So ati managed to fall behind a pure oss solution, what do you say now?