powerbook doubles as a frying pan

powerbook doubles as a frying pan

[Brad Midgley]

hi,

the only time i ever noticed the fan on my powerbook turn on was when i could smell something
burning... it was while i was running a long-running java program which was cpu-bound. the program
was probably going for about a half hour.

i shut the machine down as soon as i realized where the smell was coming from. after that it
wouldn't boot. the tech says the ethernet chip had a "blister" on it. replacing the logic board
cost almost $600.

is everyone *totally* sure that activating the fan is entirely handled in hardware? is there no
register for controlling the threshold?

/proc/cpuinfo does not report the temperature (i'm not sure the hardware can report it).

for now, i put it to sleep whenever it gets hot. i suppose i should put sleep statements in
anything that is cpu bound to save the machine. also i pull out the pc card if it's not in use to
help with ventilation.

(what a day i'm having... i just got it back and spent the day trying to find out why java would
not run. after examing working- and non-working-strace's i noticed that the date was set to 1904.
the java vm will not initialize with a date like that. maybe it's because sun realizes that no
java machine was known to be working in the year 1904.)

--
Brad
brad@turbolinux.com      http://www.turbolinux.com/~brad/


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Iain Sandoe]

> /proc/cpuinfo does not report the temperature (i'm not sure the hardware
> can report it).

yes, it can.
Someone posted a patch some time last year to get the info into the /proc
entry.  If that no longer applies it shouldn't be too difficult to update
it.

ciao,
Iain.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Benjamin Herrenschmidt]

>
>is everyone *totally* sure that activating the fan is entirely handled in
>hardware? is there no
>register for controlling the threshold?

I've seen nothing in Apple's own Darwin OS related to controlling the
powerbook fan neither. However, they do have code that use the CPU's
temperature sensors to control the ICTC (instruction cache throttling).

Troy have some code that allow to read the temp from /proc. However, it
looks like a lot of CPUs are so badly calibrated that the information
returned is almost useless... Maybe that's not the case in CPUs used in
portables, that's the case in some of the G4s used in dual G4s.

>/proc/cpuinfo does not report the temperature (i'm not sure the hardware
>can report it).
>
>for now, i put it to sleep whenever it gets hot. i suppose i should put
>sleep statements in
>anything that is cpu bound to save the machine. also i pull out the pc
>card if it's not in use to
>help with ventilation.
>
>(what a day i'm having... i just got it back and spent the day trying to
>find out why java would
>not run. after examing working- and non-working-strace's i noticed that
>the date was set to 1904.
>the java vm will not initialize with a date like that. maybe it's because
>sun realizes that no
>java machine was known to be working in the year 1904.)


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Michael Schmitz]

> >is everyone *totally* sure that activating the fan is entirely handled in
> >hardware? is there no
> >register for controlling the threshold?
>
> I've seen nothing in Apple's own Darwin OS related to controlling the
> powerbook fan neither. However, they do have code that use the CPU's
> temperature sensors to control the ICTC (instruction cache throttling).

Which would be not-so-useful to control the temperature of the ethernet
chip, right? The PowerPC CPU isn't the primary source of heat in a laptop.
At least not to the extend the Intel CPU would be.
CPU temperature monitoring wouldn't be too useful either - the CPU
core always runs a lot hotter than the rest of the motherboard.

But I'm mostly guessing here - has anybody stuck a couple of temperature
probes on a PB motherboard and really measured the operating temperature?
What would the best 'hot spots' be for such an experiment?

	Michael


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Iain Sandoe]

> probes on a PB motherboard and really measured the operating temperature?
> What would the best 'hot spots' be for such an experiment?

from the work I did in my last company - it is usually the bridge chips that
are the hottest and therefore the weakest link.

I guess that Brad's experience could be bad luck (component do sometimes
fail randomly after all).

Iain.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Chas Williams]

In message <Pine.LNX.4.30.0101291602350.2636-100000@opal.biophys.uni-duesseldor
f.de>,Michael Schmitz writes:
>What would the best 'hot spots' be for such an experiment?

since heat is a result of power consumed by the device, i would guess that
devices that consume the most power would be the hottest.  the cpu is
one of those.  the other would be the disk drive.  and indeed my powerbook
gets fairly hot around the disk drive area.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Troy Benjegerdes]

On Mon, Jan 29, 2001 at 12:01:26PM +0100, Benjamin Herrenschmidt wrote:
>
> >
> >is everyone *totally* sure that activating the fan is entirely handled in
> >hardware? is there no
> >register for controlling the threshold?
>
> I've seen nothing in Apple's own Darwin OS related to controlling the
> powerbook fan neither. However, they do have code that use the CPU's
> temperature sensors to control the ICTC (instruction cache throttling).
>
> Troy have some code that allow to read the temp from /proc. However, it
> looks like a lot of CPUs are so badly calibrated that the information
> returned is almost useless... Maybe that's not the case in CPUs used in
> portables, that's the case in some of the G4s used in dual G4s.

The IBM manufactured 750's are much better in calibration than the G4's.
Brad, recent 2.4 kernels should have my 'CONFIG_TAU' config option to turn on
the CPU temp monitoring. This code is interrupt driven, so someone could
easily ad code to tweak the ICTC when the temp exceeds certain thresholds.

--------------------------------------------------------------------------
Troy Benjegerdes                'da hozer'                hozer@drgw.net


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Timothy A. Seufert]

At 12:01 PM +0100 1/29/01, Benjamin Herrenschmidt wrote:

>Troy have some code that allow to read the temp from /proc. However, it
>looks like a lot of CPUs are so badly calibrated that the information
>returned is almost useless... Maybe that's not the case in CPUs used in
>portables, that's the case in some of the G4s used in dual G4s.

G3s generally have a useful temperature sensor, but you have to know
the correction formula specific to the die revision you're using.
The raw data is not good for much other than deltas (i.e. if the raw
value changes by 12 C, you know that it really did change by about 12
C, even if you don't know what the real starting and ending temps
were).

However, from what I've heard the G4 sensor is essentially useless.
On my own dual processor 500, a MacOS temperature readout utility
consistently tells me that one processor is 24 degC (or more) cooler
than the other.  Both CPUs are the same die revision of course.
There is just no way for this to be correct; the CPUs are thermally
coupled by a nice thick chunk of aluminum (1" x 0.5" cross section at
least), due to the design of the heatsink.  Both are getting good
contact (I checked).

   Tim Seufert

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Dan Bethe]

> However, from what I've heard the G4 sensor is essentially useless.
> On my own dual processor 500, a MacOS temperature readout utility
> consistently tells me that one processor is 24 degC (or more) cooler
> than the other.  Both CPUs are the same die revision of course.

	Maybe one is being used more than the other.  :)  That would often be
the case.
	I'm not sure if you guys discussed this, but I have two points on the
temperature issue.
	One is that the Wallstreet II series had a revision that had a
motherboard bus somewhere around 83 MHz and was way hotter than the
rest, which were often at 66 MHz.  I'm pretty sure the cpu speed was
292 MHz on that faster motherboard.  They had to lower the motherboard
bus for subsequent models because it was too hot.
	The second point is that I heard that about 1 week before Macworld
Expo, somebody checked in Macintosh power management code to Darwin's
CVS.  :)

=====
"Don't expect your own messiah; this neverworld which you desire is
only in your mind." -- http://www.dreamtheater.net/songb4.htm#IV5


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Dan Bethe]

	And have you guys tried asking about this power management stuff on
the MacOSX-dev mailing lists at Omnigroup or on the Darwin mailing
lists at Apple?

http://omnigroup.com/community/developer/mailinglists/

http://www.opensource.apple.com//projects/mail.html

=====
"Don't expect your own messiah; this neverworld which you desire is
only in your mind." -- http://www.dreamtheater.net/songb4.htm#IV5


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Timothy A. Seufert]

At 7:03 PM -0800 2/2/01, Dan Bethe wrote:
>>  However, from what I've heard the G4 sensor is essentially useless.
>>  On my own dual processor 500, a MacOS temperature readout utility
>>  consistently tells me that one processor is 24 degC (or more) cooler
>>  than the other.  Both CPUs are the same die revision of course.
>
>	Maybe one is being used more than the other.  :)  That would often be
>the case.

The construction of the heatsink used in the dual processor G4s
guarantees that the two processors should always be about the same
temperature.  They're thermally coupled by a very thick chunk of
aluminum.

Besides, the difference between the readings remains the same whether
the machine is doing nothing or crunching RC5 full blast (the MacOS
RC5 client does use both CPUs).

>	One is that the Wallstreet II series had a revision that had a
>motherboard bus somewhere around 83 MHz and was way hotter than the
>rest, which were often at 66 MHz.  I'm pretty sure the cpu speed was
>292 MHz on that faster motherboard.  They had to lower the motherboard
>bus for subsequent models because it was too hot.

That was the original Wall Street series, not Wall Street II (which
was actually codenamed "PDQ" by the way).  The 233 MHz WS had a 66
MHz bus but the 250 and 292 MHz WS models had 83 MHz busses.

But the bus wasn't the reason they got so hot.  That was caused by
the use of first generation 0.29 micron Motorola G3 processors (all
that was available), and inadequate cooling.  Later versions of the
G3 were process shrunk and use a lower core voltage, resulting in
less than half the power consumption at the same clock.  And later
versions of the PowerBook G3 got better cooling technology.

   Tim Seufert

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: powerbook doubles as a frying pan

[Benjamin Herrenschmidt]

>	The second point is that I heard that about 1 week before Macworld
>Expo, somebody checked in Macintosh power management code to Darwin's
>CVS.  :)

There have been bits of power management in Darwin CVS since public beta.
PB could put some (but not all) powerbooks to sleep, had temperature/ICTC
control, etc...

However, AFAIK, the current Darwin CVS lacks all the support for recent
machines. Looks like almost all of the "platform expert" code that
contains driver for the Apple custom ASICs have been removed from the
kernel tree (it's beeing moved to separate drivers) and didn't yet show
up in the IO tree.

Ben.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/