[lm-sensors] Andigilog aSC7621 on Intel 975XBX2 - mostly working

[George Joseph (dev)]

> -----Original Message-----
> From: lm-sensors-bounces at lm-sensors.org 
> [mailto:lm-sensors-bounces at lm-sensors.org] On Behalf Of 
> Sebastian Flothow
> Sent: Sunday, August 12, 2007 11:56 AM
> To: lm-sensors
> Subject: Re: [lm-sensors] Andigilog aSC7621 on Intel 975XBX2 
> - mostly working
> 
> Am 12. Aug 2007 um 00:43 Uhr schrieb Sebastian Flothow:
> > Am 12. Aug 2007 um 00:19 Uhr schrieb George Joseph ((dev)):
> >> Try ...
> >> http://www.peakin.com/download/asc76xx-2.6.22.2.patch
> >
> > Does compile against 2.6.16.52-xen, with hwmon debugging turned on.
> > I'll boot it tomorrow.
> 
> Works, debug messages appear in dmesg, and PECI is enabled by 
> default. 
> Nice.
> 
> I took a closer look at the PWM stuff this time and found 
> it's behaving 
> a bit weird.
> 
> Setting pwm?_enable to 0 should cause the fan to run at full speed, 
> right? This is not really clear in 
> Documentation/hwmon/sysfs-interface, 
> though.
> When setting pwm[12]_enable to 0, I can't observe any effect at all; 
> setting pwm3_enable to 0 causes the fans 3 and 4 to stop.

Pwm?_enable = 0 means disable the fan.  The driver makes no distinction
between the 3 pwm settings whatsoever but remember, pwm 1 & 2 are 4-wire
interfaces and pwm 3 is a 3 wire interface.  Also, there are other
settings that can affect what happens to the fan when "disabled".  Take
a look at the datasheet
(http://www.andigilog.com/downloads/aSC7621_70A06010.pdf) and in the
mean time, I'll run another set of tests.  Also, on 1 of my 3 boards,
the pwm1 output is inverted.  I.E.  0=full, 255=off.  Don't ask me why.
I have to set pwm1_invert=1 to get it to behave correctly.

> 
> With pwm?_enable set to 1 (manual), it seems that 51 is the effective 
> minimum for pwm1 - I can set it to lower values, and read them back, 
> but the fan won't get any slower than at 51. For pwm2 and pwm3, this 
> limit is 76. Why is this?
> 
Unfortunately, this is a question for Intel and Andigilog.  I see the
same.
> 
> _______________________________________________
> lm-sensors mailing list
> lm-sensors at lm-sensors.org
> http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
>