[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Poor Xorg performance in Etch



Andrei Popescu wrote:
On Sat, Aug 18, 2007 at 09:10:32PM -0400, Marty wrote:
Andrei Popescu wrote:

For 2D apps it should be very fast. Its responsiveness seems normal under moderate graphics load, only a few percent of the CPU, but jumps to 100% under heavy load, such as scolling up a large web page (i.e. a blog),

Can you give an example?

http://www.dailytech.com/PC+Manufacturers+Finally+Embracing+Linux+Sort+Of/article8326.htm

My crude test is to resize the browser to fit one half screen, open a bash
terminal window on the other half and run top -d .5 (.5s refresh interval).

Watching top, I scroll through the page with the scroll slider.  At low or
medium scroll rates I see about 20% CPU utilization or less. While increasing the scroll rate I seem to cross a threshhold causing the CPU to peg, likewise by
dragging the top window quickly over the browser window.  Finally, I see it when
I open several such pages in different windows and cycle through them.  The
problem does not appear during similar tests on a Sarge system.

opening a web page, starting or stopping an X app. When I enabled on

It's normal to use 100% CPU, you do want the app started as soon as possible, don't you? ;) The question is for how long. Iceweasel takes about 4 seconds (counted) to start on my P4 1.6, Intel on-board, 512 MB RAM.

I'm not sure about it.  Now I think it's limited to opening and closing certain
web pages. not applications.

If I enable the frame buffer (which I did since my last post):

(EE) RADEON(0): MergedFB does not work with Option UseFBDev, MergedFB mode is disabled

I also get the following warnings:

(WW) RADEON(0): DRI init changed memory map, adjusting ...
(WW) RADEON(0):   MC_FB_LOCATION  was: 0x1fff0000 is: 0x1fff0000
(WW) RADEON(0):   MC_AGP_LOCATION was: 0xffffffc0 is: 0xe07fe000
(WW) AIGLX: 3D driver claims to not support visual 0x23

The last one is followed by several similar lines. The framebuffer does not seem to affect the freeze issue.

I'm not familiar with radeon cards, but if you suspect a driver issue you could try the generic vesa driver.

I think the evidence of this problem while using the onboard graphics and one
other Etch system tends to point away from the Radeon card.



Reply to: