[lm-sensors] Andigilog aSC7621 on Intel 975XBX2 - mostly working
Sebastian Flothow
lists at flothow.de
Sun Aug 12 00:29:56 CEST 2007
Am 11. Aug 2007 um 22:57 Uhr schrieb George Joseph ((dev)):
> Oops...
> Turn off "Hardware monitoring chip debugging messages". That's what's
> causing it to not compile and I'll take a look at it.
Yeah, I just noticed this too, but you were faster to post it :)
With debugging turned off, it compiles nicely, even against
2.6.16.52-xen.
After enabling PECI (why is it disabled by default?), and also enabling
temp 6 through 8, I get these readings:
| temp1_input:34750
| temp2_input:39000
| temp3_input:38500
| temp4_input:-49000
| temp5_input:-49000
| temp6_input:-92969
| temp7_input:-31953
| temp8_input:-29969
So temp4 essentially duplicates temp5, while 6 through 8 seem to be
random garbage for lack of real data. temp1-temp4 is always about 85
degrees or slightly less, which makes sense.
I tried setting peci_domain to 1 in the hopes of getting separate
readings for the CPU cores, but this behaves rather weird: temp4 reads
-128 degrees, while temp5 retains the correct value, and all fans go to
full speed.
BTW, there's a small mistake in the docs: it mentions files
temp[5-8]_enable
however they're actually called peci_temp[5-8]_enable.
As to the fan readings:
| fan1_input:361
| fan2_input:357
| fan3_input:302
| fan4_input:300
1 and 2 are fine, but 3 and 4 are about half the correct value.
(This is just an estimate - the BIOS shows them at about 800 rpm, but
the BIOS enables pulse stretching upon entering the monitoring screen,
resulting in an audible speed increase, and the very first readings
still show them at something in the 600 range.)
In fact, when reading the files repeatedly, they are sometimes, rarely,
shown at about 600.
I then decided to twiddle pwm3_freq, which defaults to 10: When set to
15 or 23, the readings seem pretty random. Set to 30, I get readings
slightly below 600 most of the time, but sometimes it's too high (about
900 to 1000). 38 is OK too, but the too-high values occur more often.
Anything higher causes wrong readings almost all of the time, sometimes
way beyond the fans' max speed. The high-frequency settings are
unsuitable for 3-wire fans and result in a 0 reading.
So for now im running with pwm3_freq = 30, which seems to give correct
readings most of the time, but this is just a kludge; it would be great
to have support for pulse stretching in the drivers.
Voltage readings are fine.
More information about the lm-sensors
mailing list