Compaq evo n800v

Compaq evo n800v

[Anders Persson]

I have problems with CPU-throttling on my Compaq Evo n800v. 
To get ACPI working I compiled the following kernel+patches and applied
a modified dsdt-table.

2.4.20 kernel +
patch-2.4.21-pre4.gz +
acpi-20030218-2.4.21-pre4.diff.gz +
linux-2.4.18-acpi-20020709-dsdt.patch +
custom dsdt

AC- and battery-status works just fine, as well as the Fn-combinations
for screen-brightness. 

But no matter how low the CPU-load, the clock frequency is always at
max.  Also, the fan(s?) are always on making noise (at relatively low
speed but still too high compared to under Win-XPP).  

I am not really familiar with how this is supposed to work. Can I force
the CPU to run on the low clock-frequency when, e.g., running on
batteries? Can I set load-thresholds when the speed-step functionality
should take action? Grateful for any help on this! 

Some data available below! 

$ cd /proc/acpi/processor/C000

$ cat throttling
state count:             8
active state:            T0
states:
   *T0:                  00%
    T1:                  12%
    T2:                  25%
    T3:                  37%
    T4:                  50%
    T5:                  62%
    T6:                  75%
    T7:                  87%

$ cat limit
active limit:            P0:T0
platform limit:          P0:T0
user limit:              P0:T0
thermal limit:           P0:T0

$ cat info
processor id:            0
acpi id:                 1
bus mastering control:   yes
power management:        yes
throttling control:      yes
performance management:  yes
limit interface:         yes


$ cat performance
state count:             2
active state:            P0
states:
   *P0:                  1900 MHz, 20000 mW, 500 uS
    P1:                  1200 MHz, 10000 mW, 500 uS


$ cat power
active state:            C2
default state:           C1
bus master activity:     04082fff
states:
    C1:                  promotion[C2] demotion[--] latency[000]
usage[00001020]
   *C2:                  promotion[C3] demotion[C1] latency[002]
usage[00348077]
    C3:                  promotion[--] demotion[C2] latency[085]
usage[00002681]





-------------------------------------------------------
This SF.net email is sponsored by: Etnus, makers of TotalView, The debugger 
for complex code. Debugging C/C++ programs can leave you feeling lost and 
disoriented. TotalView can help you find your way. Available on major UNIX 
and Linux platforms. Try it free. www.etnus.com

Re: Compaq evo n800v

[Daniele Boffi]

Anders,
I have a Compaq EVO N800v too.

> 2.4.20 kernel +
> patch-2.4.21-pre4.gz +
> acpi-20030218-2.4.21-pre4.diff.gz +
> linux-2.4.18-acpi-20020709-dsdt.patch +
> custom dsdt

I am using 2.4.20, no 21-pre patch.

> 
> AC- and battery-status works just fine, as well as the Fn-combinations
> for screen-brightness. 

Same to me.

> But no matter how low the CPU-load, the clock frequency is always at
> max.

Same, is that different under Windows?

> Also, the fan(s?) are always on making noise (at relatively low
> speed but still too high compared to under Win-XPP).  

I am not experiencing that. I can see 4 fans in /proc/acpi/fan and their status
depends on the (three) thermal zones.
I guess you did something wrong with your custom dsdt.

FYI:

ls /proc/acpi/fan

C1D9  C1DA  C1DB  C1DC

ls /proc/acpi/thermal_zone

TZ1  TZ2  TZ3



Daniele


-------------------------------------------------------
This SF.net email is sponsored by: Etnus, makers of TotalView, The debugger 
for complex code. Debugging C/C++ programs can leave you feeling lost and 
disoriented. TotalView can help you find your way. Available on major UNIX 
and Linux platforms. Try it free. www.etnus.com

Re: Compaq evo n800v

[Stephan Krings]

Hello,

> 2.4.20 kernel +
> patch-2.4.21-pre4.gz +
> acpi-20030218-2.4.21-pre4.diff.gz +
> linux-2.4.18-acpi-20020709-dsdt.patch +
> custom dsdt

I'm also using 2.4.20 but with a slightly older ACPI-patch.

> AC- and battery-status works just fine, as well as the Fn-combinations
> for screen-brightness.

Same for me, although the Fn-keys for changing the display brightness
no longer work, when DRI is enabled in XFree86. But I guess this has
nothing to do with ACPI but is a problem with the XFree86 driver for
Radeon cards.

> But no matter how low the CPU-load, the clock frequency is always at
> max.

I have the same problem. I already wrote about this problem some weeks
ago but got no answer up to now. The funny thing is that toggling the
frequency does indeed work (You can recognize that if you're running a
very CPU intensive task. It'll run slightly slower), but ACPI does not
detect the change.

I didn't really care much about this problem because the notebook just
seems to do the right thing: It'll automatically switch to a lower CPU
frequency when offline and reenable it again once power is online
again. Seems the BIOS is managing this.

> Also, the fan(s?) are always on making noise (at relatively low
> speed but still too high compared to under Win-XPP).

I also see this behaviour. Although the fan is reported to be off (all
fan states in /proc/acpi/fan/*/state say so), it runs at low speed.

I'm using a little hack in one of my startup-scripts to circumvent
this problem. The following shell function will check if the fan is
reported to be off. If it is, it will enforce this for all 4 states
switching the fan on and turning it off immediatly. Works fine for me.

# Make sure the fan is really off, if the ACPI system reports so.
function check_fan () {
    local fan_on="no"
    local i

    for i in /proc/acpi/fan/*/state; do
	grep -q 'status: *on$' $i
	if [ $? = 0 ]; then
	    fan_on="yes"
	    break
	fi
    done

    if [ $fan_on == "no" ]; then
	for i in /proc/acpi/fan/*/state; do
	    echo 0 > $i
	    echo 3 > $i
	done
    fi
}

> I am not really familiar with how this is supposed to work. Can I force
> the CPU to run on the low clock-frequency when, e.g., running on
> batteries?

It does so already, as I noted above. Try it out: Use a CPU intensive
tool which reports its progress and speed (I used transcode for that)
and you'll notice, that it'll run slower once you pull the plug.

You can also notice this by looking at the CPU temperature. It'll
decrease heavily when going offline from power.

It would still be nice to to able to control this manually, I think.

> Can I set load-thresholds when the speed-step functionality
> should take action? Grateful for any help on this!

Should be possible by writing a small monitoring daemon. But since
manually toggling the CPU frequency does not work right now, that
won't help you much right now.

Regards,

Stephan


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: Compaq evo n800v

[Karol Kozimor]

Thus wrote Stephan Krings:
> Same for me, although the Fn-keys for changing the display brightness
> no longer work, when DRI is enabled in XFree86. But I guess this has
> nothing to do with ACPI but is a problem with the XFree86 driver for
> Radeon cards.

I wouldn't be so sure -- I have an Asus L380C with a Radeon M7 and I've
had no problems changing brightness so far, be it with, or without DRI.
Best regards,

-- 
Karol 'sziwan' Kozimor
sziwan-DETuoxkZsSqrDJvtcaxF/A@public.gmane.org


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: Compaq evo n800v

[Stephan Krings]

Hello,

> I wouldn't be so sure -- I have an Asus L380C with a Radeon M7 and I've
> had no problems changing brightness so far, be it with, or without DRI.

On the Evo N800c I'm using right now, I got the impression that
changing the display brightness does completely bypass the ACPI
system. The associated keys (Fn+F7 or Fn+F8) do not show up on the
console (at least showkey shows nothing for them) and patching some of
the DSDT methods with debug statements did not show anything.

So I guessed changing the display brightness would write directly into
some of the registers of the graphics card. But of course I may be
simply wrong here :-).

Regards and thanks for the input,

Stephan


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: Compaq evo n800v

[Karol Kozimor]

Thus wrote Stephan Krings:
> On the Evo N800c I'm using right now, I got the impression that
> changing the display brightness does completely bypass the ACPI
> system. The associated keys (Fn+F7 or Fn+F8) do not show up on the
> console (at least showkey shows nothing for them) and patching some of
> the DSDT methods with debug statements did not show anything.

I don't know how about the Evo, but on Asus those keys generate ACPI
events, yet there are methods in DSDT for handling those events and they
work irregardless to external conditions. Search your DSDT for such methods.
Best regards,

-- 
Karol 'sziwan' Kozimor
sziwan-DETuoxkZsSqrDJvtcaxF/A@public.gmane.org


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: Compaq evo n800v

[Karol Kozimor]

Thus wrote Stephan Krings:
> > I don't know how about the Evo, but on Asus those keys generate ACPI
> > events, yet there are methods in DSDT for handling those events and they
> > work irregardless to external conditions.
> Just out of curiousity: How are these methods called or do these
> methods comply with the VGA extensions, which are described in the
> ACPI spec?

No, they do not. These are internal \_SB.PCI0.PX40.ECD0._Q0E and
\_SB.PCI0.PX40.ECD0._Q0F to increase and decrease brightness (these are 
mapped to the hotkeys in the DSDT), and GPLV and SPLV to get and set it 
directly, probably designed for ospm (which I use in the driver).

> > Search your DSDT for such methods.
> Already did so: And none of the methods get hit when toggling the
> display brightness. Which is pity, because that would have made it
> possible to write a module like the one for Toshiba notebooks to alter
> display parameters programmatically.

Well, these might not be obvious to spot. Cross-referencing and indexing
help a lot in that. Check for IndexFields with interesting names, perhaps
you'll find something anyway.
Best regards,

-- 
Karol 'sziwan' Kozimor
sziwan-DETuoxkZsSqrDJvtcaxF/A@public.gmane.org


-------------------------------------------------------
This SF.net email is sponsored by:Crypto Challenge is now open! 
Get cracking and register here for some mind boggling fun and 
the chance of winning an Apple iPod:
http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0031en