Home

Bah!

Fri, 03/13/2009 - 21:55 — mikealborn

Yeah, so today I decided I'd spend some time troubleshooting the hibernate functionality on my HP tx1000 laptop, to find out why it doesn't consistently work. (it suspends and resumes fine, except that it will occasionally reboot instead of giving me back my desktop.)

So, I start looking into it, and re-discover TuxOnIce, formerly called suspend2. I used to use this on my Toshiba Satellite 2800, and I think on my Dell Inspiron 6000 as well, and from what I recall, it worked awesomely. A little bit more investigating led me to the understanding that suspend2 had not, in fact, been integrated into the kernel since I last played with it, which means I'm using something different on my tx1000. No wonder it doesn't work. So I figured I'd give it a try. TuxOnIce requires a kernel patch, but thankfully everything's all nicely packaged up by some friendly Debian developer.

I installed the kernel patch package, the kernel source package, and the "kernel-package" package, which automates the process of building a Debian package of a custom kernel. I patched and configured the kernel, and started building. To make things go faster, I forced the CPU on the machine to maximum speed, 2GHz. Shortly afterward, I discovered that this HP laptop's cooling system leaves a little to be desired. The temperature of both cores quickly reached 90°C, and the hardware choked the CPU clock back to 1.6GHz. I tried to reposition the laptop to give the screaming fan some clearer air to draw across the heatsink, but no sooner had I touched the machine than I heard five beeps from the kernel. I caught the briefest glimpse of some kind of kernel message in the terminal window before the power went out.

So, I rebooted, forced the CPU clock down to the lowest speed (800MHz) and started again. The cores are running between 60-70°C now, but the machine is very warm, and the heat has bled through the wooden table on which the laptop sits. I'm using the jet-exhaust from the fan to keep my coffee warm. But so far, the laptop has not overheated and crashed. The HDD, however, is reporting a temperature of 51°C, so I guess the race is on, to see whether the kenel compile finishes before the machine melts.

I'm just glad this isn't my laptop...

Update: to combat the heating problem, I installed cpufreqd and set up some profiles. Basically, the CPU uses the "conservative" governor unless the temperature breaks 80°C, at which point it switches to the "powersave" governor and bottoms the CPU frequency. I've attached my cpufreqd.conf to my "Debian on a tx1000" page in case anyone's interested.

Comments

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Fri, 03/13/2009 - 21:57 — mikealborn

Of course

Of course, once I finish the kernel compile, and test everything out, I'll update my page on the subject.

Fri, 03/13/2009 - 22:37 — pachocanadian

heh

wow, too bad the cpu clock doesn't scale back on its own when the temp rises. i'm refering to the reboot after the hardware forced it back down to 1.6 GHz above.
 
hibernate is wonderful when it works. that's the extent of my commentary.
 
on an unrelated note, i think i bullshitted excessively here at work in terms of potential power savings in turning systems off on the weekend/evenings. long story short, power management doesn't work for us because almost every system in this office runs some sort of database engine and none of them play well with PM. anyways, i used the maximal case of assuming the power supplies were actually consuming their max capacity which I understand now to be ridiculously excessive (i.e. a 450W PSU does not ever consume 450W... unless it's on fire).
 
for a fun project, estimate the heat savings gained by extending the nightly reduced hours by one and by running on reduced heat on the weekend as a percentage. ;)

Sat, 03/14/2009 - 17:41 — mikealborn

Heating with computers

We use PowerOff throughout the district to shut systems down nightly, to reduce power consumption.

An interesting anectode comes from one of our smaller schools. The principal is fanatic about keeping the lab's computers shut down when they're not in use. I was amused when a teacher came in to turn up the thermostat on the gas-fired radiant ceiling heat in the lab. Heating with computers that are powered by hydro is greener, and probably cheaper. And those thirty computers will make that room nice and toasty.

Tue, 03/17/2009 - 15:19 — pachocanadian

interesting

yeah, i'm stuck on using solutions like poweroff again because of the database end of thing. unfortunately a lot of DB software doesn't handle the system initiated shutdown well, particularly if the user is in an input screen. sucks because an automated solution is easier to sell to users.
 
heh, i love the concept of a computer lab heating a room. you're paying for the energy anyways. :)