Re: [lm-sensors] No such subfeature known when sensors -s

[Guenter Roeck <linux@roeck-us.net>]

On Fri, Sep 14, 2012 at 05:14:45PM +0100, MrNice wrote:
> On 14/09/12 12:56, Jean Delvare wrote:
> >Sorry I forgot to reply to the list last time, adding it back...
> >
> >On Thu, 13 Sep 2012 15:07:30 +0100, MrNice wrote:
> >>On 13/09/12 10:15, Jean Delvare wrote:
> >>>On Thu, 13 Sep 2012 09:19:39 +0100, MrNice wrote:
> >>>>Following the advice
> >>>>http://lm-sensors.org/browser/lm-sensors/trunk/doc/fan-divisors
> >>>This doesn't apply to the NCT6776. See in the output above, there is no
> >>>"div =" on the fan lines.
> >>>
> >>>Maybe the document should be clarified.
> >I just did so [1], hopefully the updated document is clearer and users
> >will understand whether they have to read it or not.
> >
> >[1] http://www.lm-sensors.org/changeset/6071
> >
> >>Do you know if is there any way to collect the right speeds as UEFI
> >>values are good.
> >No idea. Check the motherboard documentation. If the fans are really
> >connected to the NCT6776, this may be a driver bug. BTW you said you're
> >using the w83627ehf driver from groeck, but there is a better driver
> >for the NCT6776 now:
> >   https://github.com/groeck/nct6775
> >Guenter, maybe this driver should be listed at
> >   http://roeck-us.net/linux/drivers/
> >so that users find it?
> >
> >If the new driver doesn't help, maybe the fans are connected to another
> >monitoring chip.
> >
> >>Does the mean all the AMD FX CPU have the same issue? What to do? Is a
> >>patch on the way?
> >There's no relation between fan speed monitoring and the CPU your
> >system is using.
> >
> >>Temperatures in the hardware area have a load of loose vocabulary.
> >>I try to be accurate:
> >>Could you confirm that CPUTIN is the temperature read by the sensor in
> >>the socket of the motherboard called Tjunction?
> >I don't think so. AFAIK Tjunction refers to a temperature limit inside
> >the CPU, and it related to the digital thermal sensor in the CPU.
> >CPUTIN OTOH is an analog thermal sensor. As it is a thermistor in your
> >case, it is likely under the CPU, in the middle of the CPU socket.
> >
> >>Could you confirm that K10temp, temp1 is what AMD call Tctl (sliding
> >>scale) accessible through D18F3xA4[CurTmp] register (from BIOS and
> >>Kernel Developer's Guide)?
> >Yes, this is correct.
> >
> >>Could you tell me what is SYSTIN (system temp input) and AUXTIN
> >>(auxiliary temp input)?
> >These are analog thermal sensor inputs. As already mentioned, AUXTIN is
> >apparently not used on your board, you can ignore it. SYSTIN could be
> >anywhere on the motherboard, we can't tell. Maybe the technical
> >documentation for your board has the information.
> >
> >>Are they in the list bellow?
> >I have no idea where this list comes from, so I can't really comment on
> >that, sorry. Other than:
> >
> >>TJunction is the temperature where the pins hit the board
> >>
> >>TJunction is still a diode on the board, under the processor
> >This is completely wrong. At least this is not what the rest of the
> >world, starting with Intel technical documentation, refers to as
> >Tjunction.
> >
> 
> The new document version is now clear to understand and moreover we
> know why we don't need the divisor in some chips.
> 
> I installed the new driver. For around 15 minutes I had fan2 value
> (about 1150 RPM).
> I did some changes in the sensors3.conf file, then sensors -s, was OK.
> After some try I have fan2 = 0 RPM. I did cold boot, reinstall
> driver, reinstall lm-sensors, still 0 RPM.
> I can't find why it's not working any more.
> Any idea?
> 

ASRock uses GPIO pins to multiplex multiple fan speed inputs to a single pin
at the NCT6776. It might be that you somehow triggered or changed a multiplexer
configuration.

Do you see fan speed(s) in the BIOS ? If you do, it might well be that you have
to do that once after powerup to connect a fan to the fan speed input of the
NCT6776, and then you might see it in the driver.

[ And, no, it is practically impossible to get the necessary information to set
the GPIO multiplexer from the board vendor. AFAIK they are using GPIO pins,
but which ones on which board is a mystery. ]

Guenter


_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors