xf86-video-ati 6.10.0 Released

[phoronix]

Phoronix: xf86-video-ati 6.10.0 Released

AMD's Alex Deucher has announced the release of the xf86-video-ati 6.10.0 driver. This open-source ATI graphics driver update brings forth bi-cubic scaling on R300/400/500/690 chipsets, new ASICs are supported by this DDX driver, reduced X-Video tearing, and quite a few bug-fixes. The whole list of changes found in xf86-video-ati 6.10.0 can be read on the X.Org mailing list...

http://www.phoronix.com/vr.php?view=Njk3NA

[bulletxt]

can someone explain what is forth bi-cubic scaling?

thanks

ps: actually, what does this support mean in a normal day usage?

[bridgman]

I think this is talking about bicubic filtering in Xv. Whenever the driver displays video at a different resolution from the recording (eg when you are playing full screen or a big window) it uses the GPU to scale up the frames.

If you don't filter while scaling you get big scaled up pixels in the displayed video and it looks crappy, so you have to filter. Bilinear filtering (the default for most HW, I think) gives you a slightly blurry image; bicubic filtering takes more processing but gives you a sharper image while still getting rid of the jaggy scaled up pixels.

[NeoBrain]

can someone explain what is forth bi-cubic scaling?

thanks

ps: actually, what does this support mean in a normal day usage?

"to bring forth" is an English phrase that means something about the same as only "to bring" :P
(Well, the explanation could be a bit better, but you get the point)

bi-cubic scaling improves the picture quality when e.g. playing movies by using mathematical algorithms.
http://en.wikipedia.org/wiki/Bicubic_interpolation

EDIT: ok, bridgman was faster :(

[ayumu]

Did they make xv work at all in r700 (HD4850)?

Still tearing with the privative drivers :(.

[bridgman]

The Xv code is next on the list now that Alex has the 780 IGP parts working. We're still trying to figure out why the YUV-to-RGB shader for 6xx/7xx works in the test program but not in the driver code.

[ayumu]

The Xv code is next on the list now that Alex has the 780 IGP parts working. We're still trying to figure out why the YUV-to-RGB shader for 6xx/7xx works in the test program but not in the driver code.
Thanks for the prompt answer :)

[curaga]

Compared to rc1 I have some flicker when X starts and it also starts slower. Haven't yet tried to watch videos.

X1300 Pro (Rv516)

[Kano]

Run

xrandr

to see it flicker ;)

[curaga]

Should I do something else than just xrandr -q? 'cause that does not flicker one bit on this card..

[agd5f]

The r6xx/r7xx Xv code in the radeonhd r6xx-r7xx-support branch works now for planar formats.

[Stormking]

System still freezes after 20-30 minutes while using Firefox. This happens only when MigrationHeuristics greedy is not set. Hardware is a Mobility Radeon 9700.

[Porter]

The new driver works very well, though I'm getting some odd tearing in Xv.

Instead of the old diagonal line or triangle shape (which seems to be fixed!), now when there is motion in the image I get a repeated series of very shallow diagonal tears across the middle of the image, in a band. These seem almost horizontal until you look closely, but they are actually a series of very small diagonal tears.

This is on an X1550 PCIe. Xv tested in MythTV and in SMPlayer (mplayer).

[calica]

I'm getting some strange behavior with 6.10.0. When I come back to the machine after a few hours (over night, etc) and move the mouse to wakeup the display, it wakes up for a second then blanks for a few secs and repeats a few more times. Total time varies between 10-30 secs (rough guess, seems a lot longer but I'm sure I'm just impatient)

My log contains a bunch of the following entries:

Disable CRTC 0 success
Blank CRTC 1 success
Disable CRTC 1 success
Enable CRTC 0 success
Unblank CRTC 0 success
Output 66 disable success
Enable CRTC 0 success
Unblank CRTC 0 success
Output 66 enable success
Output 66 disable success
Blank CRTC 0 success


My card is a X1650Pro. Running "xrandr -q" doesn't reproduce it. And it was just fine in 6.9.0. I'd like to submit a real bug report but not sure of the URL of the tracker. I'd also like to know what other info would be useful.

Thanks.

[Porter]

The new driver works very well, though I'm getting some odd tearing in Xv.

Instead of the old diagonal line or triangle shape (which seems to be fixed!), now when there is motion in the image I get a repeated series of very shallow diagonal tears across the middle of the image, in a band. These seem almost horizontal until you look closely, but they are actually a series of very small diagonal tears.

This is on an X1550 PCIe. Xv tested in MythTV and in SMPlayer (mplayer).

^ Is anyone else having this issue?

[oibaf]

^ Is anyone else having this issue?

Try latest git (post 6.10.0), there is a possible fix to that problem:
http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=b437da34cb4f7c637674fb432ad81b9bf32244a3

[oibaf]

I'm getting some strange behavior with 6.10.0. When I come back to the machine after a few hours (over night, etc) and move the mouse to wakeup the display, it wakes up for a second then blanks for a few secs and repeats a few more times. Total time varies between 10-30 secs (rough guess, seems a lot longer but I'm sure I'm just impatient)

My log contains a bunch of the following entries:

Disable CRTC 0 success
Blank CRTC 1 success
Disable CRTC 1 success
Enable CRTC 0 success
Unblank CRTC 0 success
Output 66 disable success
Enable CRTC 0 success
Unblank CRTC 0 success
Output 66 enable success
Output 66 disable success
Blank CRTC 0 success


My card is a X1650Pro. Running "xrandr -q" doesn't reproduce it. And it was just fine in 6.9.0. I'd like to submit a real bug report but not sure of the URL of the tracker. I'd also like to know what other info would be useful.

Thanks.

https://bugs.freedesktop.org/enter_bug.cgi?product=xorg&component=Driver/Radeon

[Porter]

Try latest git (post 6.10.0), there is a possible fix to that problem:
http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=b437da34cb4f7c637674fb432ad81b9bf32244a3

Weird missing link issue on the tracker, that's odd. I'll try the latest build from Tormod Volden's PPA. Thanks!

[nightmorph]

Just reinstalled my X1950 Pro. xf86-video-ati 6.10 seems to work nicely. I do have a few extra things enabled in xorg.conf, as per the suggestions on the radeon wiki:

Section "Device"
Identifier "Card0"
Driver "radeon"
Option "DPMS" "true"
## Use EXA (faster) or XAA (merely fast)
Option "AccelMethod" "EXA"
## For PCIe. Use accelerated EXA DownloadFromScreen hook.
Option "AccelDFS" "1"
## Increases 3D performance substantially
## possibly in XAA mode only
Option "EnablePageFlip"
## 0 - 64 MB of gart (system) memory used.
## Wrongly defaults to 8MB sometimes, so check logs.
#Option "GARTSize" "64"
## Increases 3D performance substantially, supposedly
Option "ColorTiling" "1"
Option "DynamicClocks" "on"
EndSection
I do notice that according to my logs, EnablePageFlip is disabled in R500 and newer chips, so it seems that this option is definitely not needed when running in EXA mode on this card. Though I stuck the GARTSize option in there in case I needed it, I couldn't find any log hits that reported GART size, so it doesn't seem to be needed.

A quick run of glxgears shows that it's still reporting about 5500 FPS, same as with xf86-video-ati 6.9. (Edit: with mesa-7.3_rc2, it's now 5700 FPS, about 150 FPS faster.)

The last time I tried UT2004 with the radeon driver, I had to turn all settings down to minimum. There was still a fair amount of occasional slowdown in online matches. The card wasn't as good as my nVidia 7600GT running with the proprietary nvidia driver. Too bad -- the X1950 has at least twice the specs of the nVidia card, yet the FOSS drivers can't seem to fully utilize the hardware.

I fired up UT2004 again, and on servers with few to no players, I was able to play at almost the same settings. I usually play with max resolution and everything else at max on the nVidia card. I just had to turn off a few of the highest settings and reduce the detail distance for my X1950. Works for mostly empty maps. Maps full of players reduced the game to a 9FPS stutter-fest. Completely unworkable, so I'll have to do more tweaking.

More disturbing is that the new driver hard-locks my box sometimes! If I pull up the in-game menu with Escape, sometimes that hardlocks the box. The screen gets filled with graphical corruption -- triangles, colors, lines -- and doesn't respond.

I can't SSH into the box, so I have to hit the reset button. Not good.

Overall, the new driver is okay -- a bit faster for 2D work than even 6.9, and even a bit faster for 3D. But it's much, much more unstable for 3D usage. Kind of a mixed bag, I'm sorry to say.