Hi, question, does any of you think that
- VBE/PM aware screensavers save more energy than classic VGA compatible
  "put monitor into standby / suspend / poweroff" ones? I mean, is there
  a chance that VBE/PM puts not only the screen but also the graphics
  card into some energy saving mode?
- PCI / AGP power management D0/D1/D2 modes could be used to save notice-
  able amounts of energy in DOS in standby / suspend? Do you think that
  some (which?) of the states needs VGA BIOS re-initialization after wakeup?

Next question: Does any of you know if
- your SCSI BIOS supports int 4f? If not, what other interfaces apart
  from int 13 are provided by your BIOS?
- your PCI or AGP cards report (via PCI Power Management Capabilities
  Status Register, Capabilities List and data register, RBIL tables
  00884..00888) the power consumption for D0..D3 state, and if so, how
  much is the difference between them? Should be visible with help of
  Linux, Windows, and maybe some good PCI system information tools.
  Try Linux lspci -vv or similar...

Background: I would like to add SCSI disk spin down (is the spin down
timer configurable by SCSI BIOS setup for you?), either "at once" or
"configure timer" style. I would also like to add - IF that makes a
real difference - something like "make all PCI/AGP cards doze", but I
would not want to have to save/restore a lot of state info for that.
Basically if some extreme 3d graphics card uses 100 Watts when busy with
a lot of 3d, 20 Watts in DOS text mode, and 15 Watts in doze mode, we
would have an example of doze mode not being worth the effort for FDAPM.

Other pending problems: I would like to add ACPI or direct hardware
access power off to FDAPM, for those stupid BIOSes which think that
APM is legacy and only ACPI should be supported. Maybe even non-APM
CPU doze beyond the possible energy saving of HLT would be possible
with help of ACPI. Finally, I have to fix SUSPEND mode of FDAPM: It
has to fix the timer tick count (int 1a.0/1) by reading the RTC time
(int 1a.2/3) after waking up. I assume that no date-related int 1a
functions need refreshing, only time-related ones (1a.0/1, also known
as the 40:6c timer tick count).

Thanks for the feedback - any feedback welcome!


The SF.Net email is sponsored by: Beat the post-holiday blues
Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek.
It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt
Freedos-user mailing list

Reply via email to