View Single Post
Old 03-06-12, 09:34 AM   #1
Registered User
Join Date: Mar 2012
Posts: 7
Default Quadro 4000M: Internal display flickers with drivers from 290.10 on

Since I mailed this to <> on the 1st of March and did not receive any kind of reply, I'll try again here.

I have a HP8760w laptop with an nVidia Quadro 4000M graphics controller.

The system mainly runs gentoo Linux, with the drivers provided as "stable" by gentoo.

When I got the system in October 2011, everything worked perfectly.

In the course of normal updates I was recently updated to nVidia drivers 290.10. With these drivers I observe the following strange problem:

After about half an hour of use, the (normal desktop, no video etc., but '3D' window manager, KDE4) image will start to flicker, especially in the mid-tone areas. I.e. very bright/white areas flicker much less noticably than mid-blue areas common in my desktop background.

This flickering gets progressively worse over time, and within about another quarter of an hour I can see alternating light/dark horizontal stripes sort-of-like interlace stripes in bright areas.

I immediately suspected the video driver to be the cause of this and downgraded to 275.09.07.

During the necessary reboot to switch the driver I notice that the BIOS screens flicker in much the same way as the desktop. This flickering is preserved even when powering the machine off for ~ 1/2 minute.

Nevertheless, I eventually let it boot and run the 275.09.07 driver; and over the course of a few minutes the flickering did gradually subside until it was completely gone again.

At that time, I thought it was freak behaviour of the 290.10 drivers.

When 295.20 became available, I tried again; and much to my dismay I can observe the same flickering / horizontal lines / weird "gradually worsening" effect as with 290.10.

Reducing the display backlight brightness appears (subjectively) to lessen the flickering.

Due to the gradually changing nature, I had a look at system and GPU temperatures insofar as they are available to me on the 8760w, nothing seemed out of the ordinary (GPU at ~45 Celsius, CPU ~50 Celsius).

Since the 275.09.07 drivers don't have support for 1.11, it's a question of time for me before I'm left stranded with no working display.

I'm including nvidia-bug-report logs for 275.09.07 (working correctly, with 1.10.4), the broken 290.10 (with 1.10.4) and the broken 295.20 (with 1.11.2, nvidia-bug-report generated before and after the flickering became noticable).

I seem to remember reading in the Release Notes of newer nVidia drivers something to the effect of "fix flickering in some display/GPU combinations"; possibly this fix "unfixed" it for me.

Is there a configuration parameter to tune what the "fix flickering" change does?

When using an external (Samsung SyncMaster 2693HM, attached via a DisplayPort
to HDMI cable, 1920x1200) display, I did not see a problem with either 290.10 (although the problem was not apparent to me at that time) or 295.20.

While reproducing the problem for the bug report logs, I stumbled over another factor:

Apparently, there is some correlation between the programs displaying windows and the flicker.

A) I could not reliably reproduce the flickering without having firefox at least opened at some point during the X session.

B) One constellation of windows on my desktop allowed me to directly influence the amount/visibility of the flicker; two screenshots are attached. Whenever I move the transparent window to cover the temperature gauge in nvidia-settings, the flickering worsens.

I dug still further around the "firefox connection", and in one of the flickering constellations I was able to see a "shadow" of some off-screen text of the page I was viewing in firefox in the portion of the desktop background that flickered worst.

Trying to screenshot this 'shadow' yields a 'clean' screenshot.

I cannot say whether the flickering in conjunction with the shadowing points to a GPU memory usage / mapping problem; or whether it rather is an "analogue" problem caused by mis-driving the panel (with the shadow being an after-image). Wrong common-electrode voltage comes to mind, see:
(under "Inversion" / "Extreme tech")

Obviously, I cannot exclude that another component of my installation is at fault here, but the problem reliably disappears when downgrading to 275.09.07.

I'll be happy to help with further information / debugging steps.
Attached Thumbnails
Click image for larger version

Name:	LessFlicker.jpg
Views:	204
Size:	292.2 KB
ID:	43706  Click image for larger version

Name:	MoreFlicker.jpg
Views:	206
Size:	292.6 KB
ID:	43707  
Attached Files
File Type: gz nvidia-bug-report_275.09.07_WORKS.log.gz (48.2 KB, 115 views)
File Type: gz nvidia-bug-report_290.10_X10.4_FLICKERS.log.gz (43.3 KB, 120 views)
File Type: gz nvidia-bug-report_295.20_X11.2_FLICKERS.log.gz (45.5 KB, 102 views)
count0 is offline   Reply With Quote