Updating from 8776 to 9631 results in flickering panel using twinview

[macdonalder]

Just updated from 8776 to 9631 and my setup that used to work without a hitch does no longer. I have two panels (BenQ FP931s) in twinview, but since the update the signal to my right panel has changed in that it falls out of the range the display is capable of and it flickers.

I can play with refresh settings in my xorg.conf and make it a little better/worse but it seems to do it all the time regardless. Of the times it's better, it seems as though whenever I have darker content on the panel it's alright, but when I drag a window with a lot of white content the panel starts flickering and finally goes blank. Again, this didn't seem to be a problem with 8776.

I've swapped the panels to see if it was one in particular, maybe one is more tempremental than the other but they both exhibit the same problem.

Attached is the nvidia bug report output. The refresh settings in the config are exactly those specified by BenQ for my panels. One thing I've noticed that's curious (and maybe someone can expand on this, I'm not familiar with it) but there seems to be what X refers to as an internel and external clock, and their speeds differs? BenQ lists the 'video bandwidth' of the monitors to be between 25 - 135MHz, should that match the clock specifications in X (which are listed at 155 and 165MHz)?

Heh, as you can see, I need all the help I can get

[netllama]

If this problem persists with 1.0-9742, yet without the Coolbits, SecondMonitorHorizSync & SecondMonitorVertRefresh options, please start X with the following command:
startx -- -logverbose 5

and then generate and attach a new bug report.

Thanks,
Lonni

[macdonalder]

Thanks for the quick reply.

I installed the new driver and disabled those options as you've requested and I still have the same problem; lines through the display as it flickers trying to get a picture. The same thing happens with the lighter and darker colors, dark colors seem to stabilize the display while lighter ones cause the display to flicker and fall out of range.

Attached is a bug report with the increased verbosity you requested.

[netllama]

Please try replacing all instances of "1280x1024" with "nvidia-auto-select" in xorg.conf, and see if the flickering problem persists. If it does not, then I believe I know what the problem is.

Thanks,
Lonni

[macdonalder]

I've made those changes and for the few minutes it's been running it's been perfect!

I've attached a bug report for educational purposes but after taking a cursory look I can't tell what that actually changed. Care to share your secret?

[netllama]

I believe that the issue here is that you were running up against the pixelclock limit of the panel, which was resulting in signal degradation. The default mode that you were using in 1.0-9631 was a 76Hz refresh rate mode. By using nvidia-auto-select, you are getting a 60Hz refresh rate, which has a lower pixelclock.

In order to confirm all of the above, I'd need you to start X with the following command:
startx -- -logverbose 6

and then generate an nvidia-bug-report.log, once against 1.0-9631, and again for 1.0-8776 (a total of two bug reports).

Thanks,
Lonni

[macdonalder]

Ahhhh ok. See attached.

I see what you mean now, 1280x1024@76Hz was running at 141Mhz which was just above the capabilities of the panel. 1280x1024@60Hz runs at 108Mhz which seems a much better solution.

Thanks so much for the help, I would never have guessed to use 'nvidia-auto-select'.

[feffer]

My system also shows flickering after updating from 8776 to 9631, but only in VMware Workstation virtual machines. I'm using an older chip; Nvidia GeForce 420. My temporary solution was to downgrade back to 8776. I'm wondering if this issue will get addressed in future Nvidia driver upgrades? Or should I apply the fix above? If so will this be an issue for each new driver?

Thanks,
Ron

[netllama]

feffer,
The solution above is only applicable to the BenQ DFPs that macdonalder is using.
The issue with VMWare in full screen mode was actually discussed in a few other threads, and appears to potentially be a VMWare bug where its changing to an interlaced mode for unknown reasons. It might be best if you started a new thread along with an nvidia-bug-report.log generated after starting X with the following command:
startx -- -logverbose 6

thanks,
Lonni