when I enable C1E in the BIOS my cpu temps get a few degrees lower with Phenom II X2, but the problem is that my system time goes wrong now.
The mainboard is a GA-MA78GM-UD2H from Gigabyte with 780G.
It's an offline PC so otherwise you could sync more often with NTP I guess. Any idea how to solve or workaround this and keep the benfits of C1E (energy saving, lower noise, more potential longevity of chips and fans)?
Very strange. I wonder if C1E is tied into the same timer as the CMOS? Otherwise I have no idea. I have exactly the same board as you, so I'm gonna take a look into this and see if I can duplicate the clock problem. I dont have the ability to fix it, but maybe we can troubleshoot the problem together.
Check which clock is used in /sys/devices/system/clocksource/clocksource0/current_clocksource. If it's not "hpet", make sure it's enabled in the bios, If it is, you could try one of the other clocks listed in /sys/devices/system/clocksource/clocksource0/available_clocksource.
You could also try to periodically sync your system clock to the hardware clock (with hwclock -s) as a workaround.
Surprisingly, the clock seems to work perfectly now after setting the correct time again with linux.
I remember that linux can correct systematic time drift with software and tries to detect that drift automatically when the time is set manually, at least this was so in a Debian release a few years ago.
I checked syslog which shows that the clocksource had been changing several times to hpet without me noticing. It said "clocksource tsc unstable" and eg. delta = 300014646454 ns.
Everything is still working fine at the moment. I think the automatic linux drift correction I've mentioned has nothing to do with it because the /etc/adjtime file is not present. (maybe something changed here in recent Ubuntu releases though)
But I've read TSC access is the fastest. I don't know how relevant this is for a desktop system, though.