Mobility HD2400 + ubuntu 10.10

[Kjella]

When I got in fglrx, the score point is 173 and with the radeon score point (with drm KMS and compiz off) are 1058 (yeah, almost 10 times greater than fglrx)
But now, with radeon driver default, the benchmark score is 178 (it drops a lot).

My bet would be that if the open source driver is 10x faster than the closed source driver, it's probably not doing what it should do. Maybe some sort of stub function that doesn't actually compute what the benchmark is trying to measure? The other number seems reasonable, if you match fglrx in performance that is *good*.

[Niccola]

My bet would be that if the open source driver is 10x faster than the closed source driver, it's probably not doing what it should do. Maybe some sort of stub function that doesn't actually compute what the benchmark is trying to measure? The other number seems reasonable, if you match fglrx in performance that is *good*.

I've thought about that! I saw a better performance for 3D render with fglrx driver. But 2D of open-source, I think, is better.
One thing that fglrx driver don't "make me happy" is the bad and bad and bad and worst vsync that I ever seen.

To see just a video, or moving window tearing appears and "burn my eyes hardly"

If I could remove the tearing of fglrx, maybe I use it in my desktop PC (with a HD 3870).

But in my notebook, its garanteed the use of radeon driver. Because I just use it for college usage. The hardest thing that I use in this is compiz.

Another reason to try the open driver is the fact that open-source. I think being the free-software community is extremely good

[nanonyme]

My bet would be that if the open source driver is 10x faster than the closed source driver, it's probably not doing what it should do. Maybe some sort of stub function that doesn't actually compute what the benchmark is trying to measure? The other number seems reasonable, if you match fglrx in performance that is *good*.

Agreed, those numbers don't sound credible. Even the open one matching fglrx in performance doesn't. Then again, fglrx having the 1k fps and radeon being 175-178 fps would be very much more credible. Re-testing warranted.

[Niccola]

Agreed, those numbers don't sound credible. Even the open one matching fglrx in performance doesn't. Then again, fglrx having the 1k fps and radeon being 175-178 fps would be very much more credible. Re-testing warranted.

These are output I got: (Using glmark2 benchmark)
radeon driver first:
1- without drm and no compiz effects:
http://dl.dropbox.com/u/13977411/glm..._drm_no_compiz

2- With DRM but no compiz effects:
http://dl.dropbox.com/u/13977411/glm..._drm_no_compiz

3- Without DRM with compiz effects:
http://dl.dropbox.com/u/13977411/glm..._no_drm_compiz

4- With DRM and compiz effects:
http://dl.dropbox.com/u/13977411/glm...eon_drm_compiz

FGLRX driver now:
1- without drm and no compiz effects:
http://dl.dropbox.com/u/13977411/glm..._drm_no_compiz

2- With DRM but no compiz effects:
http://dl.dropbox.com/u/13977411/glm..._drm_no_compiz

3- Without DRM with compiz effects:
http://dl.dropbox.com/u/13977411/glm..._no_drm_compiz

4- With DRM and compiz effects:
http://dl.dropbox.com/u/13977411/glm...lrx_drm_compiz

5- FGLRX with all options activated:
http://dl.dropbox.com/u/13977411/glmark2_fglrx_with_all

Compare yourself the output...

I've considered the output that fglrx is with all options activated.
This options are the quality options in CCC and following options in xorg.conf:

Section "Module"
   ...
   Load "GLcore"
   Load "glx"
   Load "dri"
   ...
 EndSection

Section "Device"
	Identifier	"ATI Technologies Inc ATI Default Card"
	Driver      	"fglrx"
	Option	    	"VideoOverlay" "on"
	Option	    	"OpenGLOverlay" "on"
        Option          "Capabilities"  "0x00000800"
        Option          "FSAAEnable"    "yes"
        Option          "FSAAScale"     "4"
	BusID		"PCI:4:0:0"
EndSection

The fglrx fps only falls when I set these options. But quality increased

[Kjella]

You're comparing the open driver with vsync off versus the catalyst driver with vsync on. The card can generate hundreds of the frames per second, but your screen operates at 60 Hz so it can not display more than 60 frames per second.

With vsynch on it'll only measure the 60 fps actually rendered to your monitor and give you a low score, if you turn it off the fps and score will be much higher. Vsync should be off for performance benchmarks, and particular not compare scores with it on and off...

[Niccola]

You're comparing the open driver with vsync off versus the catalyst driver with vsync on. The card can generate hundreds of the frames per second, but your screen operates at 60 Hz so it can not display more than 60 frames per second.

With vsynch on it'll only measure the 60 fps actually rendered to your monitor and give you a low score, if you turn it off the fps and score will be much higher. Vsync should be off for performance benchmarks, and particular not compare scores with it on and off...

Alright! I understand that!
Well, and what would I do to have more than 60 fps and NO video/rendering/desktop tearing?

[agd5f]

Alright! I understand that!
Well, and what would I do to have more than 60 fps and NO video/rendering/desktop tearing?

You don't. If you update the screen faster than the refresh rate, you get tearing. That's what tearing is. You could do some things with multi-buffering, but the extra frames will just be discarded.

[Kjella]

You don't. If you update the screen faster than the refresh rate, you get tearing.

Agd5f knows this better than me, so I'll assume he's just inaccurate. Technically, you get tearing whenever there the rendering is trying to update the screen at the same time as the monitor is reading it, no matter the frame rate. Shortly said vsync means screen updates and monitor reads are timed so only full frames will be displayed. As a side effect, it doesn't update the screen more often than is necessary.

If you're hitting 60 fps with vsync on you are already seeing the best your monitor can display. If you want anything better you'd need a faster monitor, though I doubt you'll notice the difference between 60 fps on a 120 Hz display and 120 fps on a 120 Hz display. It is mostly important for benchmarks and scaling, if we can render 300 fps at this resolution, you can render 100 fps at this resolution and 50 fps at this resolution. In itself it makes no sense to render 300 fps of anything, your mind can't keep up anyway.

[Kjella]

In itself it makes no sense to render 300 fps of anything, your mind can't keep up anyway.

Your eyes, I should say. I think the cones in your eye is the limit, though biology is not my field. Blame the edit limit...

[Niccola]

Your eyes, I should say. I think the cones in your eye is the limit, though biology is not my field. Blame the edit limit...

I know that! our eyes recognize a movement when frames are up to 12fps at least. But, the maximum frames per second that we are able to recognize.

Look, I've another question. Some people says about a Compiz Benchmark that automatically adjust the best settings to performance.
I've googled it but without results... You guys can tell me how to run and apply the compiz benchmark in my distro??
thank you guys!