The bug is present already at boot (even from a USB stick boot without installing) in all Ubuntu (and variants) relying on Linux kernels 3.10 to 3.12.
Because it's so easy to boot via USB, I tried it on 7 computers (all of them Dell Optiplex 9010). The upshot is that old distros boot correctly, with working displays. The new 13.10 distros doesn't work (i.e. the display is distorted) on 3 computers.
On my computer (which is one of the 3 malfunctioning), recovery mode also always display correctly. Playing around with xrandr and the xorg.conf turoned out to be of no avail. I tried changing the Linux kernels (after indications from this Ubuntu forums thread: http://ubuntuforums.org/showthread.php?t=2186873 ). It turns out that kernels of version <= 3.9 are OK. Kernels >= 3.10.19 are not OK. I'm willing to bisect in the 3.10.x range if deemed of interest.
I kept the lshw, lspci, xdpyinfo, ddcprobe and Xorg.0.log of all these 7 boots from USB stick. I put all those log files in a zip file availiable at http://dfiles.eu/files/njfpo8b4k
Here are the main differences I found in those files (except the Xorg.0.log which are not so handy to compare as the beginning of lines are always different).
I guess that the most telling is the following line from ddcprobe. For non-correctly-displaying computers: dtiming: 1920x1080@68
For correctly-displaying computers: dtiming: 1920x1080@60
For the record, there is no other ctiming or dtiming line in the output of ddcprobe.
The configuration are otherwise strikingly similar. The other differences which seemed of significance (i.e. not mentionning hard drives partitions and serial numbers) is that uncorrectly working computers have the following extra line in lspci
00:16.3 Serial controller: Intel Corporation 7 Series/C210 Series Chipset Family KT Controller (rev 04)
Correspondingly, the lshw of these computers have the following extra section:
Lastly, in the *-pci section. Also, the *-memory *-bank:n (for n an integer) are not of the same vendor (Samsung or Nanya technologies for correctly-displaying computers and Hynix/Hyundai for non-correctly-displaying ones).
The visual distortion is roughly as follows: every line gets horizontally enlarged by a factor t. This factor varies with time and with the height of the line to make nice sine curves out of straight vertical lines. Actually there is a phase difference between even and odd lines. The leftmost part of the screen remains legible as the distortion effect is smaller there but the rightmost part is impossible to read. A straight vertical line looks like sine wave, the time frequency of this wave varies with time, the amplitude is always increasing linearly as one moves to the right of the screen. See also the first post in http://ubuntuforums.org/showthread.php?t=2186040
Do not hesitate to contact me if further testing is of interest.
The bug is present already at boot (even from a USB stick boot without installing) in all Ubuntu (and variants) relying on Linux kernels 3.10 to 3.12.
Because it's so easy to boot via USB, I tried it on 7 computers (all of them Dell Optiplex 9010). The upshot is that old distros boot correctly, with working displays. The new 13.10 distros doesn't work (i.e. the display is distorted) on 3 computers.
On my computer (which is one of the 3 malfunctioning), recovery mode also always display correctly. Playing around with xrandr and the xorg.conf turoned out to be of no avail. I tried changing the Linux kernels (after indications from this Ubuntu forums thread: http:// ubuntuforums. org/showthread. php?t=2186873 ). It turns out that kernels of version <= 3.9 are OK. Kernels >= 3.10.19 are not OK. I'm willing to bisect in the 3.10.x range if deemed of interest.
I kept the lshw, lspci, xdpyinfo, ddcprobe and Xorg.0.log of all these 7 boots from USB stick. I put all those log files in a zip file availiable at http:// dfiles. eu/files/ njfpo8b4k
Here are the main differences I found in those files (except the Xorg.0.log which are not so handy to compare as the beginning of lines are always different).
I guess that the most telling is the following line from ddcprobe. For non-correctly- displaying computers: dtiming: 1920x1080@68
For correctly- displaying computers: dtiming: 1920x1080@60
For the record, there is no other ctiming or dtiming line in the output of ddcprobe.
The configuration are otherwise strikingly similar. The other differences which seemed of significance (i.e. not mentionning hard drives partitions and serial numbers) is that uncorrectly working computers have the following extra line in lspci
00:16.3 Serial controller: Intel Corporation 7 Series/C210 Series Chipset Family KT Controller (rev 04)
Correspondingly, the lshw of these computers have the following extra section:
bus info: pci@0000:00:16.3
width: 32 bits
clock: 66MHz
Lastly, in the *-pci section. Also, the *-memory *-bank:n (for n an integer) are not of the same vendor (Samsung or Nanya technologies for correctly- displaying computers and Hynix/Hyundai for non-correctly- displaying ones).
The visual distortion is roughly as follows: every line gets horizontally enlarged by a factor t. This factor varies with time and with the height of the line to make nice sine curves out of straight vertical lines. Actually there is a phase difference between even and odd lines. The leftmost part of the screen remains legible as the distortion effect is smaller there but the rightmost part is impossible to read. A straight vertical line looks like sine wave, the time frequency of this wave varies with time, the amplitude is always increasing linearly as one moves to the right of the screen. See also the first post in http:// ubuntuforums. org/showthread. php?t=2186040
Do not hesitate to contact me if further testing is of interest.
Cheers,
X.