Fanless Aspire One

Is under volting not working with the stock set up? The specs for the atom are like 1 volt at 800mhz and 1.2 volts at 1.6. However I noticed that when changing cpu governors that the power draw from the wall cord never changes so I think it is not working.

 

Well, I am not using that because I lost other kernel and I don't have time to build it - I made a few tests on kernel build, were my first time

About governors, I can say that I got the first value "12:24" by using a script for phc that did put stress on CPU while was going to lower that values than the original. I clearly saw(a few times) the system going crazy. The value at it were crazy was 22 or 21 - the script suggest going 2 values higher for a bit of stability.

After the first value, changing on other values didn't have any effect! So, the others were a guess mine. May this means that the same value are used for every frequency? so you didn't see changes on power draw from the wall cord? -- thats why I would like to test others programs.. who knows?

As I wasn't to much time with that kernel and under voltage working, I can't say If it was working or not, If AA1 were more cold.

About different frequency scalings, maybe they can't mean less power used but at least mean less hot -- I remember to being use the actual daemon and build kernel, the fan were always turning on and off, while now "doing Internet" and a bit o C hacks, fan almost never run. Well, I did changed to my convenience the FANAUTO=70 to value 75 instead.

 

I've used SpeedFan with other computers have gotten good results, but it's unable to read neither voltage nor fan speed for A1. PC Wizard 2008 shows max voltage at 1.226v and min at 1.013v. I'll try clockgen and see if I can get a matching motherboard ID.

 

One interesting thing, I did read online that Atom not better in speed than other older processors from AMD, for example. The only work that Intel did made was getting a processor that have equivalent speed, can do equivalent work, as actual low grade processors - the real advantage over others is the price!!! - they tested the Atom, it real use less energy but the chipset uses more energy :-( -- so another low grade from AMD uses more energy but their chipset uses less energy, in the end, the only good advance is the price... which is mainly important for Intel and vendors... well, also for us, because we have now more cheap PCs - information more accessible, IMO.
Well, just to say that I will not be surprised If we found that Atom(and all the system) don't have all bits to control every things that others expensive processors have - maybe like this voltage control.

 

Hello!

Here you can find my Fan Control software for Windows XP and Vista:

http://www.aspireoneuser.com/forum/viewtopic.php?f=7&t=1888

Please do only answer in the given thread!

CU Ralf

 

good news everyone..

my kernel module is now complete and working (for me).. it creates entries in /proc where you can obtain the cpu temperature, the fan rpm (not completely implemented) and set the fan speed..

i will upload it today, so that anyone interested may test it..

also i'm working on an fan daemon and an gnome applet..

greetz

 

GREAT

If you can help, provide some instructions of on how to build the kernel... maybe helping the actual page on the wiki - I will try it

But I must say that Rachel approach is better for a new user... build the kernel and others modules is a bit more difficult, no?

Hmmm... we must figure out more things about EC controller... so then it can be used on future. I am waiting for undervoltage, I think is the next big thing that can be done.

 

Well.. its done..

i release an alpha version.. so don't excpect to much

Please test my kernel module.. you can obtain it here:

http://code.google.com/p/aaoacpi/downloads/list

under ubuntu you first need to install "build-essential"

Code:

sudo apt-get install build-essential

then exctract and build the downloaded archive somewhere

Code:

tar xzvf aao_acpi-0.0.1.tgz
cd aao_acpi
make

install

Code:

sudo make install

now you can load it

Code:

sudo modprobe aao_acpi

To read the cpu temperature enter

Code:

cat /proc/acpi/aao/cpu_temperature

Code:

To disable the fan:
echo -1 > /proc/acpi/aao/fan_speed

To set auto fan mode:
echo 0 > /proc/acpi/aao/fan_speed

To set low
echo 1 > /proc/acpi/aao/fan_speed

To set high
echo 2 > /proc/acpi/aao/fan_speed

To remove the module:

Code:

sudo rmmod aao_acpi
cd aao_acpi
sudo make uninstall

please report any errors here..

thank you everyone!

 

I tried it and it seems to work. Settings 1 and 2 have some strange behaviour. Both will turn the fan off for 5 seconds, and then set it to max speed.

Also the temperature is about 20 °C off from the temperature reported by coretemp. I wouldn't trust 0x58, nobody knows what it really reports or how it should be interpreted.

And since everybody seems to have their own preference i still prefer the perl one liner for the fan.

Code:

sudo perl -e 'use POSIX; $i=IOPORTS; sysopen($i,"/dev/port",O_RDWR); sub i{sysseek $i,$_[0],0; sysread $i,$res,1; return unpack("C",$res)} sub o{sysseek $i,$_[1],0; syswrite $i,pack("C",$_[0]),1} sub w{$k=0; while((i($_[0])&0x02)&&($k<100)){sleep(.01); $k++} return -($k==100)} sub e{o(0x81,0x66); if(!w(0x66)){o($_[0],0x62)} if(!w(0x66)){o($_[1],0x62)}} e(0x55,0x1F); close($i);'

 

Do you really believe on that 20ºC?? - make a test with your finger, find something near 20ºC and compare with that from AA1

 

can you send me a dump of your ec registers?

sudo perl acer_ec.pl regs

im quite sure that 0x58 is the cpu temperature.. in the acpi table it is referred as CTMP.. other acpi tables from aspires use this in TZ0._TMP.. and it is the temperature in celcius..

don't know excactly how coretemp works.. but i think it uses the method intel provides access to access the core..

(from the coretemp documentation)

and the value TjMax is dependent of the cpu model.. it may be 85, 100, 105, or something other.. so.. if coretemp is just fixed by adding the atom is some "if", i dont think it reports the right value.. but i may be wrong..

 

It's not 20° C, but 20° C off (in this case higher). The value coretemp reports seems more reasonable for a chip with a TDP of only 2.5W.

TjMax is 90° C for the N270, i adjusted it in the source.

 

Well, I have aao_acpi working now. How don't you both share that information on wiki? -- even If is different, maybe later we can figure other informations....

 

I did a test a test trying to figure out which is the real temperature, and found sth. interesting in the process.

First is set the fan to 0x20 and let it idle for about half an hour. Everything was cold to the touch and the air coming out was cold as well. coretemp reports 23° C while 0x58 thinks it's 44° C.

Then i set the fan to 0x1F and launched super pi and glxgears. After some time, about an hour, the fan suddenly turned on by itself at maximum speed, and then turned itself off about 20 seconds later. I figured out the threshold, at 70°C coretemp (86°C 0x58) the fan starts spinning, and at 60°C coretemp (77°C 0x58) it turns off. It must be hardcoded in the BIOS.

I think both the idle temperature and the threshold indicate that coretemp shows the real temperature. Also it appears no emergency script is necessary, the AA1 will protect itself.

 

i think we cannot definitely say which value is the right one..

coretemp measures the temperature of the die.. CTMP (0x58) seems to measure the case temperature..

i think we could test this.. put your aao in suspend.. and let it cool down.. switch it on and read the temperatures..

CTMP says 28°C for me.. while my room temperature is around 27°C..

will compile coretemp support now.. then test again..

edit:

found in the n270 datasheet

 

I know about the inaccuracy. It still doesn't explain why 0x58 is about 20°C off. It is probably reading a probe on the motherboard, but why should it have a higher temperature than the processor? Might be close to the 945GSE, it has a TjMax of 105°C and 6W TDP, so it's probably hotter.

Certainly 60°C and 70°C are values somebody would actually hardcode as a threshold, unlike 77°C and 86°C.

 

sounds plausible.. i wonder why the autflowing air feels "hot", while coretemp says there are only 30-35°C on the die.. is the chipset / gpu also cooled by the fan?

perhaps one windows user might try this.. dont know if it works..
http://www.intel.com/design/motherbd/software/idu/

 

I think it is a safe conclusion that the temperature is not caused by the cpu.

I have checked the temp by sliding a temperature sensor up the fan shaft, this measures around 10 degrees lower then the internal diode.. (when the fan is running).

 

Yes they are next to each other, with the Atom being in the middle of the motherboard (at the height of the tabulator key) and the 945GSE to the left of it (towards the side of the VGA connector).

 

tested agian.. (suspend, wait, poweron, sensors / CTMP)

CTMP reports 28°C
coretemp reports 24°C

room temperature (near the aao): 28°C

strange world..

EDIT: Tj_max is 90°C..