Discussion: the ati/radeon open source drivers & radeon repository

handy · 2010-07-22 08:03:03

ijanos wrote:

handy wrote:
On another note, the OpenGL speed of this testing kernel is dreadful on my machine:
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: Mesa DRI R600 (RV630 9583) 20090101 TCL DRI2
OpenGL version string: 2.0 Mesa 7.9-devel
OpenGL shading language version string: 1.10
OpenGL extensions:
Running synchronized to the vertical refresh. The framerate should be
approximately the same as the monitor refresh rate.
304 frames in 5.0 seconds = 60.778 FPS
glxgears is NOT a benchmark and these value are perfectly fine.

I know it is NOT a benchmark, I am just stating that glxgears has dropped from 1260~ to 60 with these last two kernel26-radeon-testing releases that Perry3D has so very kindly put together for us here.

Last edited by handy (2010-07-22 08:04:07)

ijanos · 2010-07-22 08:37:38

handy wrote:

I know it is NOT a benchmark, I am just stating that glxgears has dropped from 1260~ to 60 with these last two kernel26-radeon-testing releases that Perry3D has so very kindly put together for us here.

And by deducing that this drop has something to do with opengl percomance you are using it as a benchmark
What happend is that the new driver has vsync turned on by default, which will sync your framerate with the monitor's refresh rate (nowdays all monitors run on 60Hz). glgxgears told you that too.

handy · 2010-07-22 09:07:29

ijanos wrote:

handy wrote:
I know it is NOT a benchmark, I am just stating that glxgears has dropped from 1260~ to 60 with these last two kernel26-radeon-testing releases that Perry3D has so very kindly put together for us here.
And by deducing that this drop has something to do with opengl percomance you are using it as a benchmark
What happend is that the new driver has vsync turned on by default, which will sync your framerate with the monitor's refresh rate (nowdays all monitors run on 60Hz). glgxgears told you that too.

Now you have posted in an educational way I can thank you.

Before you were posting in an egotistical way that just bought a somewhat defensive explanatory response from me.

Due to your most recent response I can now understand the end of ViOLO's response, though it had kind of rung a bell due to the 60 number.

It always helps us all, if we consider the possible ignorance of others.

Not all of us can read everything, understand everything & then remember all that we have taken in, let alone synthesise new understanding from what we have taken in. As I get older I do all of these things less efficiently (this is the down side of the hill).

Vi0L0 · 2010-07-22 15:05:28

Vi0L0 wrote:

handy wrote:
On another note, the OpenGL speed of this testing kernel is dreadful on my machine:
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: Mesa DRI R600 (RV630 9583) 20090101 TCL DRI2
OpenGL version string: 2.0 Mesa 7.9-devel
OpenGL shading language version string: 1.10
OpenGL extensions:
Running synchronized to the vertical refresh. The framerate should be
approximately the same as the monitor refresh rate.
304 frames in 5.0 seconds = 60.778 FPS
302 frames in 5.0 seconds = 60.324 FPS
302 frames in 5.0 seconds = 60.324 FPS
302 frames in 5.0 seconds = 60.324 FPS
302 frames in 5.0 seconds = 60.324 FPS
302 frames in 5.0 seconds = 60.324 FPS
Looks like vsync to vblank works good :-)
BTW Is there finally a switch to disable it?

Ha! There is a way to disable it, i asked on irc, and heres the answer:

agd5f wrote:

grab mesa master, and set vblank_mode=0

i guess vblank_mode=0 need to be set as environmental variable.
Edit: agd5f confirm its need to be set as env var, so if its not working - you should update mesa.
You may set it in ie. /etc/profile.d/custom.sh (or /etc/profile.d/radeon.sh) :

export vblank_mode=0

or (maybe) by running application with it, ie.:

vblank_mode=0 glxgears

2nd tip is guessing only, and i can't actually test any of them cuz im not using oss radeon atm.

Edit: if its working i think its a good idea to add this info to wiki...

Last edited by Vi0L0 (2010-07-22 15:41:17)

megawebmaster · 2010-07-23 10:03:37

This last tip is working.

roughL · 2010-07-24 10:53:10

Perry3D wrote:

Install radeon_ucode from AUR or gpu acceleration will be disabled (even with the official stable packages). Instead of radeon_ucode you can use linux-firmware (now in Testing).

This isn't true anymore. radeon_ucode doesn't exist in AUR anymore and linux-firmware moved to core. I think you should update the howto.

handy · 2010-07-25 09:53:49

Thanks ViOLO, your last tip was very easy to test & as previously stated it worked.

Sorry for the late reply, I have been away for a few days.

On another note, is there a file or a place where I could put a file that automatically runs the command:

sudo dhcpcd eth0

on boot, so as to automatically enable my network when using the radeon-testing kernel?

Last edited by handy (2010-07-25 10:01:32)

Labello · 2010-07-25 10:31:32

rc.local:

dhcpcd eth0

no need for sudo since everything is run by root anyway!

handy · 2010-07-25 11:34:46

Labello wrote:

rc.local:
dhcpcd eth0
no need for sudo since everything is run by root anyway!

Thanks for refreshing my memory.

I've only ever used rc.local for one thing, which was running backlight, which allows me to control my iMac screen brightness, & that was over 2 years ago, so I had forgotten what I did (not that it takes 2 years for me to do that usually).

handy · 2010-07-25 11:54:00

Vi0L0 wrote:

Edit: agd5f confirm its need to be set as env var, so if its not working - you should update mesa.
You may set it in ie. /etc/profile.d/custom.sh (or /etc/profile.d/radeon.sh) :
export vblank_mode=0
or (maybe) by running application with it, ie.:
vblank_mode=0 glxgears
2nd tip is guessing only, and i can't actually test any of them cuz im not using oss radeon atm.
Edit: if its working i think its a good idea to add this info to wiki...

I just made /etc/profile.d/radeon.sh to hold export vblank_mode=0 & it works fine:

handy ~ $ gl
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: Mesa DRI R600 (RV630 9583) 20090101 TCL DRI2
OpenGL version string: 2.0 Mesa 7.9-devel
OpenGL shading language version string: 1.10
OpenGL extensions:
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
6311 frames in 5.0 seconds = 1262.168 FPS
6326 frames in 5.0 seconds = 1265.139 FPS

I'll have a look tomorrow & see if I can put something worthwhile at the end of the troubleshooting section of the ATi wiki page.

Last edited by handy (2010-07-25 12:20:41)

ijanos · 2010-07-25 12:25:56

handy wrote:

I just made /etc/profile.d/radeon.sh to hold export vblank_mode=0 & it works fine:
handy ~ $ gl
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: Mesa DRI R600 (RV630 9583) 20090101 TCL DRI2
OpenGL version string: 2.0 Mesa 7.9-devel
OpenGL shading language version string: 1.10
OpenGL extensions:
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
6311 frames in 5.0 seconds = 1262.168 FPS

Are you aware of that the vsync is a feature to reduce tearing?

handy · 2010-07-25 12:33:36

ijanos wrote:

handy wrote:
I just made /etc/profile.d/radeon.sh to hold export vblank_mode=0 & it works fine:
handy ~ $ gl
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: Mesa DRI R600 (RV630 9583) 20090101 TCL DRI2
OpenGL version string: 2.0 Mesa 7.9-devel
OpenGL shading language version string: 1.10
OpenGL extensions:
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
6311 frames in 5.0 seconds = 1262.168 FPS
Are you aware of that the vsync is a feature to reduce tearing?

I don't get tearing?

ijanos · 2010-07-25 12:47:53

handy wrote:

I don't get tearing?

http://en.wikipedia.org/wiki/Screen_tearing

It is not annoying or most people doesnt even noticie it but my point is that vsync will not hurt.

Vi0L0 · 2010-07-25 13:15:25

handy wrote:

handy ~ $ gl
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: Mesa DRI R600 (RV630 9583) 20090101 TCL DRI2
OpenGL version string: 2.0 Mesa 7.9-devel
OpenGL shading language version string: 1.10
OpenGL extensions:
ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
6311 frames in 5.0 seconds = 1262.168 FPS
6326 frames in 5.0 seconds = 1265.139 FPS

And what a nice ATTENTION

ijanos wrote:

handy wrote:
I don't get tearing?
http://en.wikipedia.org/wiki/Screen_tearing
It is not annoying or most people doesnt even noticie it but my point is that vsync will not hurt.

Ofcourse vsync is good, or even kinda needed if you are watching movies, but generally vsync is seriously busticating performance of any game or opengl application (like ie googleearth) - so it's very good that this vblank switch has been finally made.

Last edited by Vi0L0 (2010-07-25 13:17:42)

ijanos · 2010-07-25 13:27:34

Vi0L0 wrote:

Ofcourse vsync is good, or even kinda needed if you are watching movies, but generally vsync is seriously busticating performance of any game or opengl application (like ie googleearth) - so it's very good that this vblank switch has been finally made.

It will limit the maximum framrate to 60, it will not ther performance at all.

Vi0L0 · 2010-07-25 13:36:45

ijanos wrote:

Vi0L0 wrote:
Ofcourse vsync is good, or even kinda needed if you are watching movies, but generally vsync is seriously busticating performance of any game or opengl application (like ie googleearth) - so it's very good that this vblank switch has been finally made.
It will limit the maximum framrate to 60, it will not ther performance at all.

Oh it will touch performance - i still remember that before vsync_on_by_default i could easily own in q3a game, and after enabling that functionality there was a huuuge difference. This performance problem has been discussed on many forums and bugzilla, thats why this vblank switch has been made .

Last edited by Vi0L0 (2010-07-25 13:39:03)

ijanos · 2010-07-25 13:43:08

Vi0L0 wrote:

ijanos wrote:
Vi0L0 wrote:
Ofcourse vsync is good, or even kinda needed if you are watching movies, but generally vsync is seriously busticating performance of any game or opengl application (like ie googleearth) - so it's very good that this vblank switch has been finally made.
It will limit the maximum framrate to 60, it will not ther performance at all.
Oh it will touch performance - i still remember that before vsync_on_by_default i could easily own in q3a game, and after enabling that functionality there was a huuuge difference. This performance problem has been discussed on many forums and bugzilla, thats why this vblank switch has been made .

Then it is a bug which sould be fixed. Also, Q3 is a special case beacuse the ingame physics depends on the framerate.

Perry3D · 2010-07-25 13:52:47

The only problem with vsync is when your framerate drops under 60fps. Then you only have 30, 20, 15 fps ...

Vi0L0 · 2010-07-25 13:55:11

ijanos wrote:

Vi0L0 wrote:
ijanos wrote:
It will limit the maximum framrate to 60, it will not ther performance at all.
Oh it will touch performance - i still remember that before vsync_on_by_default i could easily own in q3a game, and after enabling that functionality there was a huuuge difference. This performance problem has been discussed on many forums and bugzilla, thats why this vblank switch has been made .
Then it is a bug which sould be fixed.

So maybe this vblank switch is a temporary solution, although it is still useful. I also heard that mentioned by me googleearth is also working better without this vsync.

ijanos wrote:

Also, Q3 is a special case beacuse the ingame physics depends on the framerate.

Could be, but still i can play in it with vsync on at catalyst.

Perry3D wrote:

The only problem with vsync is when your framerate drops under 60fps. Then you only have 30, 20, 15 fps ...

This could happen often with oss ati 3D performance...

Last edited by Vi0L0 (2010-07-25 13:58:34)

handy · 2010-07-25 14:18:44

Thanks guys, this run of posts has been very educational for me.

andre.vmatos · 2010-07-28 15:18:39

Hi, people. I've a radeon 3200 HD and had installed the opensource drivers, that worked fine, until some days ago, when I updated to GIT, and installed the testing kde 4.5 rc3. from there, the composite no longer working on kwin+opengl. This is strange, because it works with others windows managers, like xfwm4 and enlightenment, and 3D is working, with 1200+ FPS on glxgears (the normal average), but kwin don't allow to start with opengl composite, saying the opengl drivers is unsafe to use, or that the system doesn't support compositing. Someone can help me? I tryed to recompile the mesa-full git, xf86-video-ati-git, downgraded the kde back to 4.4.5, etc, and nothing worked. Thnkx for the help

Lothium · 2010-07-28 15:35:51

I had the same problem here with my ATI X700 and the only solution for me was to delete the whole .kde folder and start kde again.

pbnan · 2010-07-28 20:01:22

Well, using mesa-full-20100721 from radeon repository, I have issue (same as andre.vmatos' one, except I use KDE4.4) with kwin and compositing. However, everything's working fine with mesa-full-20100712.

Last edited by pbnan (2010-07-28 20:01:58)

Perry3D · 2010-07-29 06:43:49

I updated mesa in the radeon repo and did a short test with KDE: no problems so far.

pbnan · 2010-07-30 13:28:43

Thank you, it's working fine now with mesa-full-20100728.

Arch Linux

#1476 2010-07-22 08:03:03

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1477 2010-07-22 08:37:38

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1478 2010-07-22 09:07:29

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1479 2010-07-22 15:05:28

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1480 2010-07-23 10:03:37

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1481 2010-07-24 10:53:10

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1482 2010-07-25 09:53:49

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1483 2010-07-25 10:31:32

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1484 2010-07-25 11:34:46

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1485 2010-07-25 11:54:00

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1486 2010-07-25 12:25:56

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1487 2010-07-25 12:33:36

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1488 2010-07-25 12:47:53

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1489 2010-07-25 13:15:25

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1490 2010-07-25 13:27:34

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1491 2010-07-25 13:36:45

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1492 2010-07-25 13:43:08

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1493 2010-07-25 13:52:47

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1494 2010-07-25 13:55:11

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1495 2010-07-25 14:18:44

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1496 2010-07-28 15:18:39

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1497 2010-07-28 15:35:51

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1498 2010-07-28 20:01:22

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1499 2010-07-29 06:43:49

Re: Discussion: the ati/radeon open source drivers & radeon repository

#1500 2010-07-30 13:28:43

Re: Discussion: the ati/radeon open source drivers & radeon repository

Board footer