I have posted this bug (FS#31573), but in case it is just some peculiarity of my system, I am wondering if anyone else is seeing the same or similar issues.
After upgrade to kerneL 3.5.3-1 (i686) and running my GeForce 8400 GS under xf86-video-nouveau (both 1.0.1-1 from Extra and 1.0.2-1 from Testing), I experience screen corruption - apparently only under heavy load. Using the icewm WM this manifests as progressive corruption of: (1) gkrellm display (2) window borders / widgets (3) program icons. This first occured after browsing back and forth between a few high definition photos on Facebook (which reliably reproduces the crash). I have also run the ungine-heaven benchmark program from AUR and this very quickly produces the same outcome. Reverting to kernel 3.4.9-1 fixes the issue. There is nothing in my logs.
I don't know about other people but my experiences with nouveau are just bad. Is there any particular reason why you don't ride nvidia-blob-wagon? My experiences with nvidia (especially under Arch) are very good. Support for your GPU is not dropped with 304.xx drivers, 3D performance is top notch (under nouveau 3D not so much...), 8400GS is not mobile card so you don't need to worry about Optimus... If there is no serious problem for you with nvidia I suggest to go with it, especially if you plan to use anything openGL (compositions, games)
Last edited by masteryod (2012-09-17 17:44:13)
I don't know about other people but my experiences with nouveau are just bad. Is there any particular reason why you don't ride nvidia-blob-wagon?)
I used the nVidia binary for years and was happy with it, but there was an issue round about v.295 which affected me (http://www.nvnews.net/vbulletin/showthread.php?t=173214) so I switched to nouveau. On the whole I have been happy with it, but I am not a gamer and my only gripe is that FlightGear will not run using it (or would not last time I tried). I know that I could try switching back to the current binary to see if my issue has been fixed, but I thought I would persevere for a while with my bug report and this thread before doing that.
I'm experiencing similar problems with 7600GT. Corruption affects kde plasma and contents of windows. If multiple windows are open, only one of them shows somewhat uncorrupted when focused. Corruption appears randomly, no relation to load. First I thought my card has broken memory, as that is what this corruption looks like, but I have no such issues with nvidia binary blob.
Thanks for the response Skry. I have also wondered whether there is some fault on my card. Perhaps there is a defect which is exposed by recent changes in the kernel. My card is quite old and has only 256 MB of memory - maybe that is it. I have done my best to confirm the bug by swapping kernel versions a few times, and the problem inevitably and only occurs in 3.5.3-1. I have fallen back on the nv driver, which does work (the nvidia blob still does not). As I understand it, the nv driver runs in user space - which may be a clue to what is going wrong - but it is not a clue that I am capable of interpreting. Although I can easily trigger the issue by imposing heavy load, I find that it occurs after a few hours use as well. Yesterday's update installed kernel 3.5.5-1. I will give that a try with nouveau later today.
Last edited by perseus (2012-10-07 08:26:04)
Try linux-3.6 from testing, which should include a fix for a similar bug that affects a number of chipsets.
Mobo: ASUS P8Z77-V PRO // Processor: Intel Core i7-3770K 3.4GHz // GFX: nVidia GeForce GTX 970 Ti // RAM: 32GB (4x 8GB) Corsair DDR3 (@ 2133MHz) // Storage: 1x 3TB Seagate SATAII 5x 1TB Samsung SATAII, 2x 120GB Corsair SSD
Making lemonade from lemons since 2015.
WorMzy - many thanks. I will - and report back later.
I tried both 3.5.5-1 and 3.6.1 from testing using both nouveau and the nvidia binary. The same problems persist. Thanks anyway.
Issue is gone with the 3.5.6, for me that is.
Thanks Skry. It has not gone for me, but I do think I have made some progress in understanding my own issue. I just added a comment to my bug report FS#31573 which deals with it.