I recently applied all outstanding updates to a new 11.4 system (a Phonom 1090T with a Nvidia GT 240). Previously the system was completely stable, but now I’m experiencing hangs (not even the reset button works) and total power-off’s. I’ve noticed reports of similar problems with openSUSE 12, so this may be related.
For me this is happening from within chrome. By looking at what was upgraded, I’ve narrowed the problem down to flash-player-11. My standard test for this is speedtest.net - just do tests and fool around with it for a few minutes and boom! But eventually any kind of flash content in pages will cause a hang, it might take an hour or two if it’s less intense than speedtest.net.
I tried both the x86_64 and i586 versions of flash-player-11 and both cause hangs or power-offs. I’ve backed out to flash-player-10 and so far things have been stable.
Initially I suspected the nvidia driver, and it may be involved because other reports seem to be for nvidia based systems. But downgrading from 290.10 to the old nvidia 270.41 driver from my previously stable system didn’t help - so if the nvidia driver is at fault, it’s a long outstanding issue.
Obviously there may be user-mode software other than flash that could trigger the issue.
On 11/30/2011 09:36 AM, mchnz wrote:
> I’ve backed out to flash-player-10 and so
> far things have been stable.
i’d suggest you stay there with version 10…and make a backup copy of
/usr/lib/browser-plugins/libflashplayer.so…
that file is the working part of the browser’s Flash…with it safely
stored you can then allow the next Flash update to install through YaST
Online Update, and if it also is broken, then you can just overwrite the
updated file with the old one and be back in business…
wait! you are not using packagekit, are you? if you are do NOT, it is
broken and it will mess up your system, use only YaST Online Update or
“zypper patch” to roll in your updates…
After finding that the nouveau drivers were even less stable, sometimes falling over during kdm initialisation, I installed 12.1 into another spare partition and found that it too was just as unstable (except with the frame buffer driver, which was stable). I decided that things couldn’t be that broken without it being a hardware issue (even though my older 11.4 in another partition is 100% stable).
The Nvidia card was old and proven in a previous system. So looking at what was new… I reset the motherboard to defaults. I started pulling and testing RAM modules with memtest86 - RAM tested OK, but the problem persisted. The ASUS M5A97 motherboard came with TPU/EPU switches on by default. Disabling TPU/EPU seems to have done the trick. I now run speedtest.net several times with Flash 11 and the machine is still up. I suspect some new driver/library optimisations must be stressing voltage levels more than before.
Neither had I. As the system was previously stable, I hadn’t really given them much thought. They are physical switches. TPU is supposed to auto-tune the board. And EPU is supposed to save power by dynamically altering supplied power according to load. But a harder read of what TPU does and some Google searching, it’s clear that TPU is “safe” over-clocking that some have found to be unsafe. And EPU has been reported to cause problems as well. The funny thing is my MB manual says the board’s default is for them to be off, but they did not arrive that way - with hindsight I should have checked. Still stable today.