Laptop ACPI question

[Kevin Oberman]

> From: "Markie" <mark.cullen at dsl.pipex.com>
> Date: Sat, 1 May 2004 22:28:55 +0100
> 
> 
> ----- Original Message ----- 
> From: "Kevin Oberman" <oberman at es.net>
> To: "Markie" <mark.cullen at dsl.pipex.com>
> Cc: <freebsd-mobile at freebsd.org>
> Sent: Saturday, May 01, 2004 10:05 PM
> Subject: Re: Laptop ACPI question
> 
> 
> | > From: "Markie" <mark.cullen at dsl.pipex.com>
> | > Date: Sat, 1 May 2004 21:29:56 +0100
> | > Sender: owner-freebsd-mobile at freebsd.org
> | >
> | > Hello,
> | >
> | > Just a quick question... would having ACPI working on a laptop increase
> | > battery life at all? I just left it sat idle without ACPI and it got to
> an
> | > hour and 30 minutes... I went to start xchat and it just switched off
> | > straight away after it started loading :o) if I leave it doing `cat
> | > /dev/random > /dev/null` it only lasts 45 minutes :o(
> | >
> | > I am just wondering if it's a naffed battery... or... there's something
> a
> | > bit wrong with the laptop or... I just need ACPI? I don't really know
> what
> | > ACPI does, so.... :o)
> |
> | Actually, ACPI will greatly improve battery life soon, but not yet. The
> | bits and pieces are being fed into CURRENT and I suspect that SpeedStep
> | support will be coming soon.
> | In the meantime, you can use sysctls to manually adjust CPU performance
> | to enhance battery life.
> |
> | Look at:
> | hw.acpi.cpu.throttle_max: 8
> | hw.acpi.cpu.throttle_state: 8
> | hw.acpi.cpu.cx_supported: C1/0 C2/1 C3/85
> | hw.acpi.cpu.cx_lowest: 0
> | hw.acpi.cpu.cx_history: 1453705/0 0/0 0/0
> |
> | Reducing the hw.acpi.cpu.throttle_state will increase battery life by
> | effectively reducing CPU speed. The reduction is linear and a setting of
> | 1 makes my system crawl.
> |
> | Setting hw.acpi.cpu.cx_lowest will reduce the responsiveness of P4-M or
> | Centrino system by putting the system in a "deeper sleep" than just
> | halting the CPU. The cost is that it takes longer for the system to
> | start processing again. Its effect can appear as jerkiness in some
> | operations. It may be set to as many values (higher is slower) as are
> | listed in cx_history. (In the example, there are three, 0, 1, and 2
> | available.)  Depending on hardware connected, some levels may not be
> | available. On my laptop, 2 is not available if the USB driver is loaded.
> |
> 
> Thanks for both your replies! I am not too sure I have speed step or
> anything. It's a fairly old CPU, a mobile celeron 800MHz, 100MHz FSB
> (anyone know anything about these at all? are they not that good? seems
> nice and quick to me). I guess I am pretty buggered then. There's not alot
> of info I can find about the laptop either. In any case, it wouldn't boot
> with ACPI enabled anyway... so I think I am just double buggered :o)

Make sure that your BIOS is te latest available. That can make a huge
difference. The CPU is not the critical issue; it's the BIOS
support. Older systems may simply not run well with ACPI and APM is the
best way to go on these systems. (Have you tried APM?)

> All I know is some site said that the battery was supposed to last 2 hours
> ish, which would be cool. No doubt this was taken with the laptop sitting
> totally idle though.

Battery benchmarking is tricky and highly variable. And, as batteries
age, they go downhill. Try the command 'acpiconf -i 0' to get a lot of
detailed battery information on current. I think it works on 5.2.1, but
may have to run as root. I may tell you things like how much of a
charge the battery is taking and the type and manufacturer.

> I will have to try Linux or Windows if I can't get FreeBSD ACPI working and
> see if it gets me any more battery life I guess! Such a shame, it doesn't
> seem that bad of a laptop for what I want... but if I have to use it on
> mains it's not as useful as it could be!

Most ACPI code is shared by Linux and FreeBSD as it is based closely on
the Intel supplied sources. It is possible that Linux support for
control of different parts of the ACPI is more advanced. but I suspect
it's pretty close.

Since every vendor makes sure that their Windows code plays well with
their hardware, it's unlikely that open source software will ever quite
catch up with it, but it is rapidly closing the gap. It's just that,
until a few more pieces are in place, it's not obvious.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634