[lm-sensors] w83627hf, fan speed always high, unchangeable?

[Sebastian Nowozin]

Hello,

I just installed an EPOX ES300 miniMe barebone with a EP-4PGF mainboard
using Debian/testing, kernel 2.6.14. The chip label on the board reads
"Winbond W83627HF-AW", and sensors-detect finds it as such with a
confidence of 8 at the ISA bus.

"sensors" displays fan1 (the case fan) with constant rpm around 4700,
fan2 and fan3 at 0 rpm. I tried everything with the w83781d driver,
which also loads fine and displays two fan with non-zero rpm (the CPU
fan with around 2600rpm). I am only interested in the case fan though,
as it is really loud. The 4700rpm is consistent with what is shown in
the BIOS.

Using the "pwmconfig" program, I get this output:

Found the following PWM controls:
9191-0290/pwm1
9191-0290/pwm2

Found the following fan sensors:
9191-0290/fan1_input current speed: 4687 RPM
9191-0290/fan2_input current speed: 0 ... skipping!
9191-0290/fan3_input current speed: 0 ... skipping!

Testing pwm control 9191-0290/pwm1 ...
9191-0290/fan1_input ... speed was 4687 now 4720
no correlation

Testing pwm control 9191-0290/pwm2 ...
9191-0290/fan1_input ... speed was 4687 now 4753
no correlation

With no slowing down of the case fan.

Now my question, limited by my understanding of the sensors package:
What is most likely:

1. The w83627hf chip on the board can only provide readings but no
control of the fanspeed.
2. The w83627hf chip may provide control of the fanspeed, but there is
some hardware connector/cable missing.
3. Something else is wrong, eg. /etc/sensors.conf (attached), ...

Thanks,
Sebastian
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: sensors.conf
URL: <http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20051102/22334203/attachment.ksh>

[Jean Delvare]

Hi Sebastian,

Post by Sebastian Nowozin
I just installed an EPOX ES300 miniMe barebone with a EP-4PGF mainboard
using Debian/testing, kernel 2.6.14. The chip label on the board reads
"Winbond W83627HF-AW", and sensors-detect finds it as such with a
confidence of 8 at the ISA bus.

Great :)

Post by Sebastian Nowozin
"sensors" displays fan1 (the case fan) with constant rpm around 4700,
fan2 and fan3 at 0 rpm. I tried everything with the w83781d driver,
which also loads fine and displays two fan with non-zero rpm (the CPU
fan with around 2600rpm). I am only interested in the case fan though,
as it is really loud. The 4700rpm is consistent with what is shown in
the BIOS.

The w83627hf driver should work better for you, don't use w83781d.

2600 RPM is the low limit of what the W83627HF chip can measure with a
fan clock divider of 2. I guess that you have fan2_div set to 2. That
would explain why you sometimes see the real CPU fan speed, and
sometimes 0 RPM.

Try setting fan2_div to 4 (add "set fan2_div 4" in the w83627hf-*
section of your configuration file, then run "sensors -s".) I expect
you to then have consistent CPU fan speed readings.

Post by Sebastian Nowozin
9191-0290/pwm1
9191-0290/pwm2
9191-0290/fan1_input current speed: 4687 RPM
9191-0290/fan2_input current speed: 0 ... skipping!
9191-0290/fan3_input current speed: 0 ... skipping!
Testing pwm control 9191-0290/pwm1 ...
9191-0290/fan1_input ... speed was 4687 now 4720
no correlation
Testing pwm control 9191-0290/pwm2 ...
9191-0290/fan1_input ... speed was 4687 now 4753
no correlation
With no slowing down of the case fan.
1. The w83627hf chip on the board can only provide readings but no
control of the fanspeed.

Yes, this is the most probable. It is always a bit frustrating to know
that the chip could do it, but the motherboard manufacturer did not wire
it properly to do so. Unfortunately, this is a rather common situation.

Note however that the run of pwmconfig above only shows that the speed of
the case can cannot be controlled. When you get valid readings for your
CPU fan, try again. It is possible, though unlikely, that the CPU fan
can be controlled.

Likewise, if your motherboard has a third fan header, you may try to use
it for the case fan instead of fan1, and run pwmconfig again. Who knows?

Post by Sebastian Nowozin
2. The w83627hf chip may provide control of the fanspeed, but there is
some hardware connector/cable missing.

No, no additional cable is required. Either the motherboard is properly
wired, or it's not.

Post by Sebastian Nowozin
3. Something else is wrong, eg. /etc/sensors.conf (attached), ...

Attaching a 70 kB file wasn't exactly wise ;) please only include the
relevant part next time.

The configuration look OK, but it's really only affecting monitoring,
not control, so it isn't related to your problem.

You may try to load the w83627hf driver with reset=1. Please report if it
helps. I don't expect it to actually help, but you may still want to
try.

Remember that there are always hardware solutions to reduce the fan
noise. You can use a potentiometer. You can use a thermoregulated fan.
Or you can simply use a fan with a lower nominal speed.

--
Jean Delvare

[Grant Coady]

Post by Jean Delvare
Hi Sebastian,

Post by Sebastian Nowozin
I just installed an EPOX ES300 miniMe barebone with a EP-4PGF mainboard
using Debian/testing, kernel 2.6.14. The chip label on the board reads
"Winbond W83627HF-AW", and sensors-detect finds it as such with a
confidence of 8 at the ISA bus.

...

Post by Jean Delvare
Yes, this is the most probable. It is always a bit frustrating to know
that the chip could do it, but the motherboard manufacturer did not wire
it properly to do so. Unfortunately, this is a rather common situation.

I have similar sensor chip in one box (w83697hf), no fan speed control,
turns out desktop AMD Sempron doesn't have CPU power idle therefore CPU
temp varies only a couple degrees on load --> no point having fan speed
control, thus mobo make didn't wire it. If CPU is running at ~50'C there's
not much fan speed reduction can be done, in my situation I put the PC
box under the desk to reduce the fan noise :o)

Grant.

[Sebastian Nowozin]

Hi Jean and the others who have helped me,

Post by Jean Delvare

Post by Sebastian Nowozin
1. The w83627hf chip on the board can only provide readings but no
control of the fanspeed.

Yes, this is the most probable. It is always a bit frustrating to know
that the chip could do it, but the motherboard manufacturer did not wire
it properly to do so. Unfortunately, this is a rather common situation.
Note however that the run of pwmconfig above only shows that the speed of
the case can cannot be controlled. When you get valid readings for your
CPU fan, try again. It is possible, though unlikely, that the CPU fan
can be controlled.

Today I opened the case and examined both the CPU and power supply
fan. The power supply fan has a simple two-wire connection, without
any speed reading line physically present. It must be a mistake the
BIOS displays a reading for it.

The loud fan was actually the CPU fan (I only discovered that _after_
I exchanged the power supply fan). Luckily I could reuse the
potiometer for the power supply fan for the CPU fan then. With the
potiometer on the lowest voltage setting it is much more quieter now
and testing with the "cpuburn" program for a long time still only had
a max peak temperature at 68? degrees. So its sort of a workaround
solution now, but I'm closer to the original goal.

Post by Jean Delvare
Likewise, if your motherboard has a third fan header, you may try to use
it for the case fan instead of fan1, and run pwmconfig again. Who knows?

It only has two fan connectors, for two CPU fans, both connected.
(The EPOX board is a bit strange in that way, it has a combined
Northbridge and CPU cooler with two fans, one is loud and the other is
hidden inside).

I really don't know where from the BIOS could come up with a chassis
fan speed. Maybe just a wrong text label.

Post by Jean Delvare

Post by Sebastian Nowozin
3. Something else is wrong, eg. /etc/sensors.conf (attached), ...

Attaching a 70 kB file wasn't exactly wise ;) please only include the
relevant part next time.

Ooops, next time I will remember :)

Post by Jean Delvare
You may try to load the w83627hf driver with reset=1. Please report if it
helps. I don't expect it to actually help, but you may still want to
try.

I tried it, no difference. I tried with init=0, but for 2.6.14 it
should be reset=1, as reset=0/init=0 is the default behaviour, right?
:)

Post by Jean Delvare
Remember that there are always hardware solutions to reduce the fan
noise. You can use a potentiometer. You can use a thermoregulated fan.
Or you can simply use a fan with a lower nominal speed.

Yes, but its a more expensive solution. In my case it turns out to be
the only solution, but if it would have been possible by changing a
configuration setting I would really dislike the thought of buying
extra hardware to fix it. :-)

Thanks for the great support you all are doing on lm-sensors, its a real help.

Regards,
Sebastian

[Jean Delvare]

Hi Sebastian,

Post by Sebastian Nowozin
The loud fan was actually the CPU fan (I only discovered that _after_
I exchanged the power supply fan). Luckily I could reuse the
potiometer for the power supply fan for the CPU fan then. With the
potiometer on the lowest voltage setting it is much more quieter now
and testing with the "cpuburn" program for a long time still only had
a max peak temperature at 68? degrees. So its sort of a workaround
solution now, but I'm closer to the original goal.

68 degrees C is quite high, even for a peak value. I wouldn't want any
of my CPUs to reach that temperature.

Post by Sebastian Nowozin
I tried it, no difference. I tried with init=0, but for 2.6.14 it
should be reset=1, as reset=0/init=0 is the default behaviour, right?
:)

The default is actually reset=0/init=1.

Post by Sebastian Nowozin
Thanks for the great support you all are doing on lm-sensors, its a real help.

Thanks for the nice words :') These are really valuable as these are
almost the only reward we ever get. Unhappy users usually make much
more noise than happy ones, and at times this becomes very frustrating,
although I guess it's unavoidable.

[Rudolf Marek]

Hello,

As the last resort you can try to change the base frequencies of PWM. All you need is to look into the dataseet and find register
0x5C bank 0. Then you can use isaset command to modify the register. If you need help with this just write back.

Regards
Rudolf