Fanless Aspire One

Are you running Ubuntu or some other distro than the default one?

 

i'm running ubuntu..

can you test it yourself?

i used the instructions found at http://macles.blogspot.com/2008/08/reading-internal-thermal-diode-of-intel.html, modified for ubuntu, to enable coretemp..

perhaps this what the n270 datasheet describes.. saturation below 50°C..

i will try it again.. need some time to cool down..

 

I'm running the default install, that might explain why my 0x58 is never below 40°C (with coretemp 23° C), if it's really related to the 945GSE. Ubuntu probably has some additional power saving magic going on for the chipset because of a recent kernel and a recent intel xorg driver. The default install is still on 2.6.23 and intel 2.1.0, which is quite old.

 

reading other patches, that f.e. are suggested on the kernel list, the above mentioned patch repots the temperature 5 degrees to low.
It is relative to the tjmax, and while the max is 90 following the datasheet, other patches include 95 degrees as upper limit.

Withouth documentation we never know.

 

What's interesting is that the patch mentions 95°C according to the datasheet. Maybe the maintaine of the module has a general datasheet for the Atom. The 90°C are in the datasheet only for the N270.

 

quite interesting.. around 17°C difference between chipset (?) temperature and die temperature.. thats a funny gradient..

summarize..

coretemp reports the die temp, relative to the tj_max.. which is 90°C in the datasheet
CTMP (0x58) reports something like the temperature of chipset + cpu mixed..

which one controls the fan? i noticed that if the CTMP is below 33°C, the fan is off by default.. then it is put to a low speed mode till 49°C.. from there its at middle speed and CTMP stays at around 50°C.. now i measure the same with coretemp..

a question related to rpm.. someone knows the max rpm for the fan?


--- second test ---
CTMP: 33°C
coretemp: 15°C
room: 28°C

is there somewhere a peltier element?

 

15°C is definitely too low, that is probably misreporting because of the inaccuracy. It's funny that you mention 49°C with 0x58, if you look in the changelog of the BIOS.

1. Update new Fan table, change 49'C up to middle speed.

So 0x58 is actually being used, regardless of where it gets the temperature from. It would make sense to have the fan adjust to the average temperature of probably chipset and cpu, instead of relying on the die temperature of the cpu. I think it's a probe close to or under the thermal module.

 

information about the DTS thing..
http://intel.wingateweb.com/US08/published/sessions/TMTS001/SF08_TMTS001_100r.pdf

as far as i understand the value for Tj_max is not necessarily the one mentioned in the datasheet but depends also of some MSR..

 

I think this is not the case for Atom processors, because if you look at the patch that implemented reading that register it only switches between 85°C (mobile) and 100°C (desktop) based on that register. And the table it references is only for a specific processor.

 

Why don't be practice and forget that value? -- let us define a limit as we are doing now with initial hack - it is working perfect -- since we don't have information from Acer, we will never be sure of that value. The hack is working now, for me thats good.

I would put now my energies on undervoltage.

 

ok, sounds good.. i will rename the cpu_temperature entry from my module to temperature.. then we could use both values.. the coretemp and the CTMP..

macles, you mentioned that the high and low fan speed settings didn't work for you..

can you please try to set 0x55 to 0x1e (low) and 0x1d? perhaps the stock linux has some magic that controls the fan by itself?

 

I agree but i think we are close to solving the riddle. Also the hardware limit at 70°C coretemp is good enough for me.

And speaking of undervolting, i actually tried that PCH thing yout posted, i couldn't get it to work properly. I think it's not useful anyway because it added about 400 wakeup ticks per seconds to the kernel, i have no idea what this thing is doing. It's kinda pointless to undervolt the cpu to save energy, while at the same time preventing it to go to sleep.

 

how many wakeups do you get w/o this? powertop says 80-150 for me (not touching and not connected anything). without the uvcvideo module it lowers to 50-60.. without fan i have an discharge rate from about 8W.. 3 hours of battery life..

 

I tried 0x1D and 0x1E and it's the same behaviour, fan is turned off and then set to max speed after about 5 seconds. Did you see my previous post on the 49°C?

The stock kernel is lousy on the wakeup ticks, about 380. Power is 8.3W with WLAN off and everything else at default including the fan, +1W with WLAN.

 

sorry, didn't see it.. but it makes sense what you say.. especially if the gpu and chipset is also cooled by the fan. i think it is the right way to use this value..

can you send me a dump of "perl acer_ec.pl regs" and dmesg?

thank you in advance.

hm.. with defaults + disabled wlan i get around 9.4W.. seems that linpus is heavy optimized..

 

I guess the riddle is solved then. 0x58 reads a probe somewhere near the chipset and cpu, which is used to adjust the fan speed. coretemp is the die temperature of the cpu, not very accurate below 50C. coretemp 70°C turns the fan on at max speed if it's been previously turned off, and off again when the cpu has been cooled down to 60°C.

 

[/quote:32lvyube]

It is just tweaking. I get with wlan off 7.3-7.7 W under ubuntu. 8.4-8.8 with wlan on.. depending what I do.

 

Hey you all, and thanks to especially rachel and the others, who are involved with the fan-hack in Ubuntu. It works like a charm, and now the netbook has turned into a perfect little mate to carry with. About the noise of the fan, I consider my AA1's fan silent, compared to Acer Extensa 5220, which is next to the AA1 atm. I think I'm a lucky owner of a "silent" AA1 then, although it makes noise, it isn't at annoying level. I have bought the AA1 from Finland, it has Samsung SSD and I've also added 1GB of Samsung DDR2 memory in it, and like qball, I also have the power usages of ~8.5-9W with WLAN on, and ~7.8-8.5W with WLAN (and my mini bt-dongle) off. The powertop gives me 3h battery life straight after changed to the battery mode, but the Ubuntu's powermanager gives only 2h5min. I haven't tested the battery life with a clock or anything, but I can assure that it does run more than 2h30mins with non-heavy use.

BTW, I changed the fan auto- temp from 70 to 67, which seemed to be a nice compromise with a temp as close to 65 as possible, since the temp seems to flow around 62-65 degrees in normal use, so setting it at 65 made the fan spin once in a minute. The 70 seemed to me a little too high since the bottom of the netbook became significant hot on heavy use, so changing the value to 67 eased this situation a bit. I think that the chipset needs to warm like 6-10 degrees to heat up the temp-sensor, which is used in the script with 3 degrees, or something like that, at least the bottom feels cooler now.

Keep up the good work guys.

 

Is there anyway this can be done through windows too?

 

Yes: viewtopic.php?f=7&t=1888 :lol: :mrgreen: