specs: alienware m18x2 with 680m (single card)
tested on basic updated installs (ACPId) present: all 64 bit os’es used with ia32libs present and i386 architecture available.
Opensuse 12.2, Ustudio 12.10, Crunchbang Waldorf, Mint 14
also on Ustudio 12.10 with 3.7.0(xorg-edgers) and 3.8.0 kernels.(mainline)
nvidia driver versions tested:
driver versions 304, 310.14 310.19, (from .deb maintainers) 310.3x (.run blob install) and 313 (.deb maintainer)
ACPI does report AC plug unplug to OS, desktop shows on battery, charging, fully charged, proc changes and events reported to acpi. Kernels confirmed compiled with AC_apater as module, checked with a loaded .config into make menuconfig with kernel sources.
nvidia-settings shows powermizer power source as battery, no matter what the condition, if plugged in, adaptive mode cycles as needed from 0-2 states, each with a 324 clock and 1800,900, and 324 memory clocks.
If unplugged, and actually on battery: level 1 appears used as max level in adaptive, so its detecting an event.
if laptop_mode-tools is never installed, the clock of memory reports 1800 fro 0-2 states, if ever installed it updates to 324,900, and 1800 in memory clock only with GPU clock remaining at 324, however many distributions install this initially so i was only able to notice this testing on #!.
Concerns: level 3 is never available, nor is AC adapter present ever achieved:
i understand the Fermi GPU clock may not be being reported as to what it it running at actual clock.
performance difference noted: opengl with 0 tesselation twinview on linux 43fps, same heaven settings on windows7/64 with latest nvidia release stable 60+ fps. Opengl opensim viewer in linux approx 100fps, windows 7/64 approx 150fps, temps of GPU in linux remain low as if card is never achieving highest clock.