kwin performance gets worse and worse with every release

Duncan 1i5t5.duncan at
Thu Jun 30 21:20:57 BST 2011

Nikos Chantziaras posted on Thu, 30 Jun 2011 12:45:54 +0300 as excerpted:

> On 06/30/2011 06:28 AM, Duncan wrote:
>> Nikos Chantziaras posted on Wed, 29 Jun 2011 23:22:36 +0300 as
>> excerpted:
>>> (I'm using a Radeon HD4870 with the open source drivers.)
>>> I was able to find a tweak that makes kwin usable again though.  I had
>>> to create a ~/.drirc file

>>> Also, the "VSync" option in System Settings must be unchecked.  Both
>>> of these things must be performed

>>> It should be noted that the desktop is still tear-free after doing the
>>> above.

>> I'd be interested in your locked and unlocked glxgears framerate
>> stats.  [Y]our hd4870/rv780 should in theory be getting better
>> stats than my hd4650/rv730.
> I'm getting about 7000 FPS windowed, and 4000FPS maximized (yes, that's
> thousands, not hundreds.)  But I think glxgears is not pretty good as a
> benchmark these days.

You are correct, but it's still useful at an extremely basic opengl 
functionality level (besides being interesting eye candy, like the cube, 
flip-windows effect, etc), and can be useful for comparisons within the 
same family (r7xx Radeon chips, in our case).

And yes, your results DO show it as being vastly better than mine.  Part 
of that may be that the chip is better, but I believe part of it is the 
bus, as well.  AGP bus simply can't compare to 16x PCIE, no matter HOW 
you slice it!

But I've long suspected that like a purebred race horse, the really top 
performing graphics hardware is more finicky and requires more careful 
handling than ordinary stock.  So it may be that your card, being tuned 
for higher performance, was having trouble with a force-refresh-locked 
"de-tuned" configuration.  (Caveat: My way of simplifying a subject that 
in detail is certainly well above my head.)

>> Next thing to try is OpenGL video playback.  I wonder if the unlocked
>> framerate will affect it any...
> If you have SwapbuffersWait disabled, then the effect will be ugly
> tearing and higher GPU temps.

I do, and I believe I saw some corruption on playback as a result (hard 
to tell without going back and playing the same video without 
swapbufferswait and/or without refresh-unlocking), but it wasn't as bad 
as I suspected it might be.

Part of that, however, was probably because of the low-quality video I 
was testing with.  Youtube higher (but not highest, generally only 
available on a few videos, often paid or at least interstitial-ad-
supported, which I have trouble with as I won't do proprietary flash) 
quality, which might almost match SD, but is still a LONG way from 
hardware-challenging HD, even 720p.

I'll have to try it with a DVD image or the like...

Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

This message is from the kde mailing list.
Account management:
More info:

More information about the kde mailing list