Re: KB835732 breaks S3 standby timer

From: Rock (rock1_at_hotmail.com)
Date: 04/23/04 

Date: Fri, 23 Apr 2004 18:51:06 GMT

I would like to clarify a couple of Andrew's comments below: (probably
more than you ever wanted to know about Power Management)

Since ACPI, the BIOS simply tells Microsoft what methods to run or can
be run to support power management. ACPI was Microsoft's attempt to
take over power management so it was uniform for all PCs and basically
didn't permit one vendor's BIOS to work better than another for
purposes of battery life benchmarks. (of course their reasoning was
that it gave users a better experience....)

Now with ACPI, the BIOS should not be in control. If it's trying to
do power management, and ACPI drivers are trying to do it too, it's
likely you will see bizzare problems. Consider when BIOS turns off a
device and a driver goes out to wait for a particular bit of a
register to be set, and that chip's turned off. What will the driver
do? Having written drivers, I'll tell you we don't set timeouts on
every Windows function call or every register bit tested. So it can
hang.

Recommendation: if you're using Windows power management, disable
BIOS power management in your hardware setup.

I have seen BIOSes that have the ability to turn-off ACPI. Basically,
there's a bunch of tables that the BIOS puts in a certain spot in
memory so Microsoft ACPI drivers can find the info. (Typically
hardware vendor ACPI drivers don't use this info). There are BIOSes
that can disable this feature, so if you want to run Windows power
management, you have to enable this ACPI. (This iis *not* the same as
PnP aware). In the Device manager (see below) you can tell if you're
running ACPI, there will be a system device that's functioning
normally (I hope).

If you are hibernating, and resuming, you are very fortunate indeed.
Power Management is absolutely the last thing that goes into a driver
development, because the developers are trying to make the device
work when it's powered on, not figure out how to turn it of and
restore its state when it comes back up. I have seen drivers that
have had to be completely restructured to add PM, and management (of
course) decided not to invest the money to do it. Since it's
typically added as an afterthout, you'll find controls scattered
around. so here's where to look

1. You found Control Panel/Power Options
2. Check device manager for each device: rt-click My Computer,
select properties, click the hardware tab, push the device manager
button. Then open each device and look around for power management
options. I have powr management options on my net card and monitor
devices.
3. Check your network cards. Rt-click My Network Places select
properties, rt-click the entries that have a device name on the
right-hand side, and look at the device configuration (button near the
top). In my case I have the same power management options there that
appear in DeviceManager.
4. Check the setup configuration software for installed devices. I
saw power management options in a Dlink wireless LAN card's special
control panel.

Regarding what's left-on in hibernate: that's up to the PC designer.
Hibernate is S4 . The system's state is saved to a file, and
everything's typically turned off except for the things that can bring
the PC out of hibernation: typically "Wake-On-Lan" special packets.
You don't move the mouse to come out of hibernation. This is the
thing that's last done in an ACPI driver, so if it's working for you,
you are lucky. And you buy good hardware. Anyway, it's up to the
hardware vendor to write the ACPI hibernate stuff to permit that
driver to save the state on ACPI power-down messages, and reset the
device and put that state back in. Consider how hard it is to restore
a VPN after you've saved to file: the keys are gone, the sequences
that make the VPN have completely disappeared. So it usually just
doesn't work after a hibernate.

There can be2 versions of S4: turn off the power to the PCI bus (so
your LAN card can't wake the machine ever) or leave power to certain
devices on (like the PCI or Southbridge, PCI bus, PCMCIA copntroller,
etc.) Again, up to the discretion of the manufacturer.

S3 is deep sleep. HAving designed PCs, the only thing we could safely
turn-off was the CPU power. We could spin-down the hard drive, but we
couldn't turn it off (typically). The RAM is left on, and goes to
slow-refresh mode or for very good hardware, self-refresh mode.
Typically, all PLLs are left-on (they are the clocks inside the chips
and I have seen where they can consume 50% of the power dissipated by
a given chip).

S2 is typically not used. It's defined as a lowr-power S1.

S1 is standby. In this state, different devices may enter
power-saving mode. And maybe not. I have seen CPU clocks slowed, and
hard drives spun down. The rules are different and depend on your
hardware supplier. Microsoft simply issues the ACPI message, then the
video driver may determine to blank your CRT. But maybe not. because
that'sure to generate support calls (Why did my screen suddenly go
blank when I was running that presentation!!!????!!!??) I think
Microsoft's driver sends the ATAPI command to spin-down the drive, but
you might not be using Microsoft's driver.

There are utilities out there that can set device timers to enter the
standby state after receiving the ACPI message. Phoenix had a good
one.

So if you want to save more power in standby, you'll have to get one
of those utilities. There used to be a way to set power profiles in
Windows, but I just looked and it seems it's gone (that is, you could
set the timers and stuff.)

HTH

Rock

On Wed, 21 Apr 2004 18:32:34 +0200, Andrew Aronoff
<NOSPAM_WRONG.ADDRESS@yahoo.com> wrote:

>Hi, Peter.
>
>>Standby didn't seem to do anything on my machine
>
>Probably because it wasn't configured in the BIOS. Windows will ask
>the BIOS to put the PC on standby. If the BIOS hasn't been configured,
>it simply ignores the request.
>
>There are two different modes of standby typically available, S1 and
>S3.
>
>S1 (standard) standby is less advantageous, IMHO, because it leaves
>the fans running. Hibernation, at least, shuts down all the noise
>sources, but it also disables Scheduled Tasks. IOW, when the PC
>hibernates, it's completely shut down. It just starts up a bit faster
>and puts the O/S back to the same operating point.
>
>S3 standby has significant advantages: all the noise sources are
>stopped and the PC restarts quickly, much quicker than with
>hibernation. The only thing that's left powered is the RAM and, at a
>diminished level, the processor. Since the processor and RAM stay
>powered on, scheduled programs can still run. The disadvantage of S3
>is that power is still consumed and, if it's lost (a portable's
>batteries die, for example), any unsaved data in programs that were
>open when the standby started will be lost. If a user is careful to
>hit the save button often when working, the issue of data loss should
>be minimal. The exact power consumed by S3 standby is variable. You'd
>need to test it out to see if it meets your needs.
>
>I know of no PC that's delivered with S3 standby enabled. I don't
>really know why. Since it _must_ be configured in the BIOS, most
>people don't even know it exists. If it _was_ typically configured,
>its proper function could not be ignored by Microsoft.
>
>If you look in your PC's owners manual, you may see an explanation of
>the suspend modes that are available in the BIOS. Look for "S3" or
>"STR" or "Suspend To RAM". If your PC has it, you'll probably have an
>option to set a "password" for wake-up. In fact, it's not really a
>password at all, but a handy keyboard shortcut. If I want the PC to
>resume, I can just hit a few letters on the keyboard. If your PC's so
>equipped, you might want to give S3 a try.
>
>regards, Andy