CpuIdle better than Win98SE's power management?

Archived from groups: alt.comp.hardware.homebuilt (More info?)

I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
running Win98SE and CpuIdle, and I'm monitoring power consumption
using a wattmeter on the AC mains input. I notice that the system
consumes about 106W in the idle state with CpuIdle disabled. This
drops to about 75W with CpuIdle enabled. By my calculations, a power
reduction of 30W at the mains equates to about 21W at the PSU's
output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
means a reduction in CPU current of ~13A, assuming the CPU accounts
for the majority of the savings.

If I allow Windows' power management to switch the system to standby,
the power consumption is around 90W, some 15W *higher* than if the
system were idling. As expected, enabling or disabling CpuIdle does
not affect this figure. While it seems counterintuitive, I believe I'd
be better off disabling power management and allowing CpuIdle to
retain control of the CPU. Not what I expected at all ...

For those that don't have access to a wattmeter, the differences in
power consumption for the various states are reflected in the CPU
temperatures as reported by MBM, for example. In fact, I notice that
when my system returns from standby mode, CpuIdle believes it is
enabled when in fact the CPU temp indicates that it is disabled.
Correct operation, as evidenced by a drop in CPU temp, is restored
only after I toggle CpuIdle off and then on again.


- Franc Zabkar
--
Please remove one 's' from my address when replying by email.
6 answers Last reply
More about cpuidle win98se power management
  1. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    Franc Zabkar wrote:
    > I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
    > running Win98SE and CpuIdle, and I'm monitoring power consumption
    > using a wattmeter on the AC mains input. I notice that the system
    > consumes about 106W in the idle state with CpuIdle disabled. This
    > drops to about 75W with CpuIdle enabled. By my calculations, a power
    > reduction of 30W at the mains equates to about 21W at the PSU's
    > output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
    > means a reduction in CPU current of ~13A, assuming the CPU accounts
    > for the majority of the savings.
    >
    > If I allow Windows' power management to switch the system to standby,
    > the power consumption is around 90W, some 15W *higher* than if the
    > system were idling. As expected, enabling or disabling CpuIdle does
    > not affect this figure. While it seems counterintuitive, I believe I'd
    > be better off disabling power management and allowing CpuIdle to
    > retain control of the CPU. Not what I expected at all ...
    >
    > For those that don't have access to a wattmeter, the differences in
    > power consumption for the various states are reflected in the CPU
    > temperatures as reported by MBM, for example. In fact, I notice that
    > when my system returns from standby mode, CpuIdle believes it is
    > enabled when in fact the CPU temp indicates that it is disabled.
    > Correct operation, as evidenced by a drop in CPU temp, is restored
    > only after I toggle CpuIdle off and then on again.
    >
    >
    > - Franc Zabkar

    Interesting. Maybe Windows is using a simple HALT while CPUIdle uses a
    deeper sleep HALT?
  2. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <i0td51hthql7qk64lklgpfevulj5vcielc@4ax.com>, Franc Zabkar
    says...
    > I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
    > running Win98SE and CpuIdle, and I'm monitoring power consumption
    > using a wattmeter on the AC mains input. I notice that the system
    > consumes about 106W in the idle state with CpuIdle disabled. This
    > drops to about 75W with CpuIdle enabled. By my calculations, a power
    > reduction of 30W at the mains equates to about 21W at the PSU's
    > output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
    > means a reduction in CPU current of ~13A, assuming the CPU accounts
    > for the majority of the savings.
    >
    > If I allow Windows' power management to switch the system to standby,
    > the power consumption is around 90W, some 15W *higher* than if the
    > system were idling. As expected, enabling or disabling CpuIdle does
    > not affect this figure. While it seems counterintuitive, I believe I'd
    > be better off disabling power management and allowing CpuIdle to
    > retain control of the CPU. Not what I expected at all ...
    >
    Absolutely spot on. The differences between CPUIdle and Windows
    NT4/2000/XP are very minimal though. Win9x and WinNT use different
    methods.


    --
    Conor

    Windows & Outlook/OE in particular, shipped with settings making them
    as open to entry as a starlet in a porno. Steve B
  3. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <115e430j4eblj1f@corp.supernews.com>, David Maynard says...
    > Franc Zabkar wrote:
    > > I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
    > > running Win98SE and CpuIdle, and I'm monitoring power consumption
    > > using a wattmeter on the AC mains input. I notice that the system
    > > consumes about 106W in the idle state with CpuIdle disabled. This
    > > drops to about 75W with CpuIdle enabled. By my calculations, a power
    > > reduction of 30W at the mains equates to about 21W at the PSU's
    > > output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
    > > means a reduction in CPU current of ~13A, assuming the CPU accounts
    > > for the majority of the savings.
    > >
    > > If I allow Windows' power management to switch the system to standby,
    > > the power consumption is around 90W, some 15W *higher* than if the
    > > system were idling. As expected, enabling or disabling CpuIdle does
    > > not affect this figure. While it seems counterintuitive, I believe I'd
    > > be better off disabling power management and allowing CpuIdle to
    > > retain control of the CPU. Not what I expected at all ...
    > >
    > Interesting. Maybe Windows is using a simple HALT while CPUIdle uses a
    > deeper sleep HALT?
    >
    Windows 9x implements it differently to Windows NT. Win9x does
    basically a half assed attempt at it. You benefit greatly from using
    RAIN/CPUIdle on Win9x but the benefits aren't as great on Win NT as it
    does the job properly.

    --
    Conor

    Windows & Outlook/OE in particular, shipped with settings making them
    as open to entry as a starlet in a porno. Steve B
  4. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    Conor Turton wrote:

    > In article <115e430j4eblj1f@corp.supernews.com>, David Maynard says...
    >
    >>Franc Zabkar wrote:
    >>
    >>>I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
    >>>running Win98SE and CpuIdle, and I'm monitoring power consumption
    >>>using a wattmeter on the AC mains input. I notice that the system
    >>>consumes about 106W in the idle state with CpuIdle disabled. This
    >>>drops to about 75W with CpuIdle enabled. By my calculations, a power
    >>>reduction of 30W at the mains equates to about 21W at the PSU's
    >>>output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
    >>>means a reduction in CPU current of ~13A, assuming the CPU accounts
    >>>for the majority of the savings.
    >>>
    >>>If I allow Windows' power management to switch the system to standby,
    >>>the power consumption is around 90W, some 15W *higher* than if the
    >>>system were idling. As expected, enabling or disabling CpuIdle does
    >>>not affect this figure. While it seems counterintuitive, I believe I'd
    >>>be better off disabling power management and allowing CpuIdle to
    >>>retain control of the CPU. Not what I expected at all ...
    >>>
    >>
    >>Interesting. Maybe Windows is using a simple HALT while CPUIdle uses a
    >>deeper sleep HALT?
    >>
    >
    > Windows 9x implements it differently to Windows NT. Win9x does
    > basically a half assed attempt at it. You benefit greatly from using
    > RAIN/CPUIdle on Win9x but the benefits aren't as great on Win NT as it
    > does the job properly.
    >

    Well, I was trying to figure out what "half assed" means.
  5. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Sat, 9 Apr 2005 00:23:08 +0100, Conor Turton
    <conor@conorturton.com> wrote:

    >In article <115e430j4eblj1f@corp.supernews.com>, David Maynard says...
    >> Franc Zabkar wrote:
    >> > I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
    >> > running Win98SE and CpuIdle, and I'm monitoring power consumption
    >> > using a wattmeter on the AC mains input. I notice that the system
    >> > consumes about 106W in the idle state with CpuIdle disabled. This
    >> > drops to about 75W with CpuIdle enabled. By my calculations, a power
    >> > reduction of 30W at the mains equates to about 21W at the PSU's
    >> > output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
    >> > means a reduction in CPU current of ~13A, assuming the CPU accounts
    >> > for the majority of the savings.
    >> >
    >> > If I allow Windows' power management to switch the system to standby,
    >> > the power consumption is around 90W, some 15W *higher* than if the
    >> > system were idling. As expected, enabling or disabling CpuIdle does
    >> > not affect this figure. While it seems counterintuitive, I believe I'd
    >> > be better off disabling power management and allowing CpuIdle to
    >> > retain control of the CPU. Not what I expected at all ...
    >> >
    >> Interesting. Maybe Windows is using a simple HALT while CPUIdle uses a
    >> deeper sleep HALT?
    >>
    >Windows 9x implements it differently to Windows NT. Win9x does
    >basically a half assed attempt at it. You benefit greatly from using
    >RAIN/CPUIdle on Win9x but the benefits aren't as great on Win NT as it
    >does the job properly.

    Do you have any evidence to support this, ie- specifics?
    IMO, whether you like how Win9x does it or not, it does do
    it properly and has same end result on power consumption
    (and of course CPU temp).

    More often the issue is whether the motherboard bios has the
    proper chipset registers set for the bus-disconnect, and if
    not, whether the CPU idling software added on top of the OS
    recognizes that chipset and can toggle the chipset
    registers. Without the registers flipped all the HLT
    commands won't accomplish anything. Then again, some
    boards' bios have a setting to enable or disable this
    feature. Don't know about Sis chipset socket A, I do try to
    avoid Sis based boards for any platform.
  6. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Sat, 9 Apr 2005 00:23:59 +0100, Conor Turton
    <conor@conorturton.com> put finger to keyboard and composed:

    >In article <i0td51hthql7qk64lklgpfevulj5vcielc@4ax.com>, Franc Zabkar
    >says...
    >> I have an AMD Athlon XP 2500+ CPU on an ECS L7S7A2 motherboard. I'm
    >> running Win98SE and CpuIdle, and I'm monitoring power consumption
    >> using a wattmeter on the AC mains input. I notice that the system
    >> consumes about 106W in the idle state with CpuIdle disabled. This
    >> drops to about 75W with CpuIdle enabled. By my calculations, a power
    >> reduction of 30W at the mains equates to about 21W at the PSU's
    >> output, assuming an efficiency of ~ 70%. For a Vcore of 1.65V, this
    >> means a reduction in CPU current of ~13A, assuming the CPU accounts
    >> for the majority of the savings.
    >>
    >> If I allow Windows' power management to switch the system to standby,
    >> the power consumption is around 90W, some 15W *higher* than if the
    >> system were idling. As expected, enabling or disabling CpuIdle does
    >> not affect this figure. While it seems counterintuitive, I believe I'd
    >> be better off disabling power management and allowing CpuIdle to
    >> retain control of the CPU. Not what I expected at all ...
    >>
    >Absolutely spot on. The differences between CPUIdle and Windows
    >NT4/2000/XP are very minimal though. Win9x and WinNT use different
    >methods.

    AFAIK, one difference is that Windows versions 98 and above are ACPI
    aware (or at least partially so), whereas Win95 and NT are not.
    Another difference is that Win98 does not support the S3 state, but
    Win98SE does.

    Anyway, I tried the same comparison on a system running Win98SE, with
    an AMD K6-2/450MHz CPU, SiS 5597 chipset, and AMIBIOS. Enabling or
    disabling BIOS power management/APM did not affect the results for
    either system.

    K6-2 XP 2500+
    ---- --------
    Idling with CpuIdle disabled ---- 71W 106W
    Idling with CpuIdle enabled ---- 52W 75W
    Standby (S1, stopgrant mode) ---- 41W 90W
    Standby (S3 mode, suspend to RAM) -------- 5W

    ACPI was enabled in the Athlon system but was not available in the
    K6-2 machine. The former's BIOS power management mode has three
    options, S1, S3, or auto. CpuIdle defaults to S1 mode.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
Ask a new question

Read More

Homebuilt Management CPUs Power Systems