[lm-sensors] IT8721F fan control?

[lm-sensors] IT8721F fan control?

[Alexandros Diamantidis]

Hello!

I've just gotten an ASUS M4A87TD/USB3 motherboard, which has an IT8721F
chip. Although I've enabled a "silent" BIOS setting, the new system runs
its fans pretty fast and noisily, so I'm interested in controlling their
speed for quieter operation.

After a few web searches, I gather that IT8721F isn't completely
supported yet but it's being worked on - so thanks a lot for your work
on this!

Anyway, I just wanted to report that with with the new driver from
http://khali.linux-fr.org/devel/misc/it87/, sensors reports:

--------------------------------------------------------------------
k10temp-pci-00c3
Adapter: PCI adapter
temp1:       +20.0°C  (high = +70.0°C, crit = +90.0°C)  

radeon-pci-0500
Adapter: PCI adapter
temp1:       +50.0°C                                    

it8721-isa-0290
Adapter: ISA adapter
in0:         +2.81 V  (min =  +1.08 V, max =  +2.22 V)   ALARM
in1:         +2.76 V  (min =  +0.00 V, max =  +0.41 V)   ALARM
in2:         +1.31 V  (min =  +0.71 V, max =  +2.69 V)   
+3.3V:       +3.31 V  (min =  +5.66 V, max =  +5.35 V)   ALARM
in4:         +2.47 V  (min =  +0.74 V, max =  +0.88 V)   ALARM
in5:         +1.26 V  (min =  +1.33 V, max =  +2.76 V)   ALARM
in6:         +1.69 V  (min =  +2.48 V, max =  +0.35 V)   ALARM
3VSB:        +5.35 V  (min =  +5.71 V, max =  +1.30 V)   ALARM
Vbat:        +3.38 V
fan1:       1573 RPM  (min =   11 RPM)
fan2:       1819 RPM  (min =   11 RPM)
fan3:       1229 RPM  (min =   12 RPM)
temp1:       +28.0°C  (low  = +62.0°C, high = -118.0°C)  ALARM  sensor = thermistor
temp2:       +23.0°C  (low  = -53.0°C, high = -35.0°C)  sensor = thermistor
temp3:      -128.0°C  (low  =  -9.0°C, high =  -2.0°C)  sensor = disabled
--------------------------------------------------------------------

From dmesg:

it87: Found IT8721F chip at 0x290, revision 1
ACPI: resource it87 [io  0x0295-0x0296] conflicts with ACPI region ECRE [io  0x0290-0x02af pref window disabled]
ACPI: This conflict may cause random problems and system instability
ACPI: If an ACPI driver is available for this device, you should use it instead of the native driver

It seems that pwmconfig doesn't do anything... should it have worked or
is support for fan control still incomplete?

For reference, with asus_atk0110 sensors reports:

--------------------------------------------------------------------
k10temp-pci-00c3
Adapter: PCI adapter
temp1:       +21.0°C  (high = +70.0°C, crit = +90.0°C)

atk0110-acpi-0
Adapter: ACPI interface
Vcore Voltage:     +1.31 V  (min =  +0.80 V, max =  +1.60 V)
+3.3V Voltage:     +3.31 V  (min =  +2.97 V, max =  +3.63 V)
+5V Voltage:       +4.93 V  (min =  +4.50 V, max =  +5.50 V)
+12V Voltage:     +12.05 V  (min = +10.20 V, max = +13.80 V)
CPU Fan Speed:    675000 RPM  (min =  600 RPM)
Chassis Fan Speed:1844 RPM  (min =  600 RPM)
Power Fan Speed:  1240 RPM  (min =  600 RPM)
CPU Temperature:   +29.0°C  (high = +60.0°C, crit = +95.0°C)
MB Temperature:    +23.0°C  (high = +45.0°C, crit = +75.0°C)

radeon-pci-0500
Adapter: PCI adapter
temp1:       +53.0°C
--------------------------------------------------------------------

sensors-detect from lm-sensors 3.1.2 reports:

Probing for Super-I/O at 0x2e/0x2f
[...]
Trying family `ITE'...                                      Yes
Found unknown chip with ID 0x8721
    (logical device 4 has address 0x290, could be sensors)


Thanks again,
Alexandros

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

Re: [lm-sensors] IT8721F fan control?

[Luca Tettamanti]

Hi,
I can't comment on the native driver, but:

On Sat, Nov 6, 2010 at 1:22 AM, Alexandros Diamantidis <adia@hellug.gr> wrote:
> atk0110-acpi-0
> Adapter: ACPI interface
> Vcore Voltage:     +1.31 V  (min =  +0.80 V, max =  +1.60 V)
> +3.3V Voltage:     +3.31 V  (min =  +2.97 V, max =  +3.63 V)
> +5V Voltage:       +4.93 V  (min =  +4.50 V, max =  +5.50 V)
> +12V Voltage:     +12.05 V  (min = +10.20 V, max = +13.80 V)
> CPU Fan Speed:    675000 RPM  (min =  600 RPM)

There's something wrong with the reading of this fan. Can you send me
a copy of the DSDT table (/sys/firmware/acpi/tables/DSDT)?

Luca

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

Re: [lm-sensors] IT8721F fan control?

[Alexandros Diamantidis]

* Luca Tettamanti [2010-11-08 13:53]:
> > atk0110-acpi-0
> > CPU Fan Speed:    675000 RPM  (min =  600 RPM)
> 
> There's something wrong with the reading of this fan. Can you send me
> a copy of the DSDT table (/sys/firmware/acpi/tables/DSDT)?

Oops, I hadn't noticed that!

In fact, that fan speed is usually read correctly, but intermittently
jumps to some absurd value. I don't think it's a problem of the driver,
since I see this in the BIOS "system info" screen as well. Here's an
example of the usual output:

atk0110-acpi-0
Adapter: ACPI interface
Vcore Voltage:     +1.27 V  (min =  +0.80 V, max =  +1.60 V)
+3.3V Voltage:     +3.29 V  (min =  +2.97 V, max =  +3.63 V)
+5V Voltage:       +4.93 V  (min =  +4.50 V, max =  +5.50 V)
+12V Voltage:     +12.00 V  (min = +10.20 V, max = +13.80 V)
CPU Fan Speed:    1562 RPM  (min =  600 RPM)
Chassis Fan Speed:   0 RPM  (min =  600 RPM)
Power Fan Speed:  1216 RPM  (min =  600 RPM)
CPU Temperature:   +42.0°C  (high = +60.0°C, crit = +95.0°C)  
MB Temperature:    +29.0°C  (high = +45.0°C, crit = +75.0°C)  

For the moment I've disconnected the chassis fan because it was very
distracting.

If it's of any interest, I've put a copy of the DSDT table here:
http://users.uoa.gr/~adia/DSDT

Thanks!

Alexandros

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

Re: [lm-sensors] IT8721F fan control?

[Luca Tettamanti]

On Mon, Nov 8, 2010 at 6:13 PM, Alexandros Diamantidis <adia@hellug.gr> wrote:
> * Luca Tettamanti [2010-11-08 13:53]:
>> > atk0110-acpi-0
>> > CPU Fan Speed:    675000 RPM  (min =  600 RPM)
>>
>> There's something wrong with the reading of this fan. Can you send me
>> a copy of the DSDT table (/sys/firmware/acpi/tables/DSDT)?
>
> Oops, I hadn't noticed that!
>
> In fact, that fan speed is usually read correctly, but intermittently
> jumps to some absurd value. I don't think it's a problem of the driver,
> since I see this in the BIOS "system info" screen as well.

Ok, then. The DSDT looks fine too :)

> For the moment I've disconnected the chassis fan because it was very
> distracting.

It should be controllable with QFan, no?

Luca

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

Re: [lm-sensors] IT8721F fan control?

[Jean Delvare]

On Mon, 8 Nov 2010 19:13:54 +0200, Alexandros Diamantidis wrote:
> * Luca Tettamanti [2010-11-08 13:53]:
> > > atk0110-acpi-0
> > > CPU Fan Speed:    675000 RPM  (min =  600 RPM)
> > 
> > There's something wrong with the reading of this fan. Can you send me
> > a copy of the DSDT table (/sys/firmware/acpi/tables/DSDT)?
> 
> Oops, I hadn't noticed that!
> 
> In fact, that fan speed is usually read correctly, but intermittently
> jumps to some absurd value. I don't think it's a problem of the driver,
> since I see this in the BIOS "system info" screen as well.

This is a common problem when using external fan speed control
solutions (potentiometers, fanbus and the like.) The degraded power
signal makes it harder for the tachometer to sample the speed signal.

I presume that some self-controlled fans could also suffer from the
same problem, depending on the exact implementation.

-- 
Jean Delvare

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

Re: [lm-sensors] IT8721F fan control?

[Alexandros Diamantidis]

* Luca Tettamanti [2010-11-11 14:56]:
> > For the moment I've disconnected the chassis fan because it was very
> > distracting.
> 
> It should be controllable with QFan, no?

Do you mean the BIOS Q-Fan option? It only mentions a "CPU Q-Fan", which
is enabled and doesn't seem to do much for the front fan. The other two
fans indeed aren't so loud, so it probably works for them. Could that
mean that the front fan speed isn't controllable at all by the hardware?

Since even with Q-Fan enabled it's not as quiet as I'd like and
temperatures don't rise much even under full load, I'd be nice if fans
could go even slower if possible. My initial message was mostly a query
about whether the native driver will allow this when complete.

* Jean Delvare [2010-11-11 15:37]:
> > In fact, that fan speed is usually read correctly, but intermittently
> > jumps to some absurd value. I don't think it's a problem of the driver,
> 
> This is a common problem when using external fan speed control
> solutions (potentiometers, fanbus and the like.)

There's nothing of the sort in use here, but maybe you're right that the
CPU fan itself causes the misreading. No big problem anyway...

Thanks again!

Alexandros

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

Re: [lm-sensors] IT8721F fan control?

[Luca Tettamanti]

On Fri, Nov 12, 2010 at 12:23 AM, Alexandros Diamantidis <adia@hellug.gr> wrote:
> * Luca Tettamanti [2010-11-11 14:56]:
>> > For the moment I've disconnected the chassis fan because it was very
>> > distracting.
>>
>> It should be controllable with QFan, no?
>
> Do you mean the BIOS Q-Fan option? It only mentions a "CPU Q-Fan", which
> is enabled and doesn't seem to do much for the front fan. The other two
> fans indeed aren't so loud, so it probably works for them. Could that
> mean that the front fan speed isn't controllable at all by the hardware?

They are probably controllable, but the control is not implemented in
the firmware. My board (M4A79XTD) also has a Q-Fan options for the
chassis fan.

> Since even with Q-Fan enabled it's not as quiet as I'd like and
> temperatures don't rise much even under full load, I'd be nice if fans
> could go even slower if possible. My initial message was mostly a query
> about whether the native driver will allow this when complete.

The problem with the native driver is that it uses resources that are
claimed by ACPI, which might be unsafe.

Luca

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

Re: [lm-sensors] IT8721F fan control?

[Jean Delvare]

On Fri, 12 Nov 2010 01:23:05 +0200, Alexandros Diamantidis wrote:
> * Luca Tettamanti [2010-11-11 14:56]:
> > > For the moment I've disconnected the chassis fan because it was very
> > > distracting.
> > 
> > It should be controllable with QFan, no?
> 
> Do you mean the BIOS Q-Fan option? It only mentions a "CPU Q-Fan", which
> is enabled and doesn't seem to do much for the front fan. The other two
> fans indeed aren't so loud, so it probably works for them. Could that
> mean that the front fan speed isn't controllable at all by the hardware?
> 
> Since even with Q-Fan enabled it's not as quiet as I'd like and
> temperatures don't rise much even under full load, I'd be nice if fans
> could go even slower if possible. My initial message was mostly a query
> about whether the native driver will allow this when complete.

Looking at the it87 driver code, it is pretty clear that I messed up
the manual fan speed control support for the IT8721F. Sorry about that.
It requires much larger changes compared to other chips, than I first
thought.

Working on it, stay tuned...

-- 
Jean Delvare

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

Re: [lm-sensors] IT8721F fan control?

[Jean Delvare]

Hi Alexandros,

On Fri, 12 Nov 2010 16:09:50 +0100, Jean Delvare wrote:
> Looking at the it87 driver code, it is pretty clear that I messed up
> the manual fan speed control support for the IT8721F. Sorry about that.
> It requires much larger changes compared to other chips, than I first
> thought.
> 
> Working on it, stay tuned...

Can you please test the standalone driver at:
  http://khali.linux-fr.org/devel/misc/it87/

It has fixed manual fan speed control for the IT8721F. Hopefully it
works now...

-- 
Jean Delvare
http://khali.linux-fr.org/wishlist.html

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

Re: [lm-sensors] IT8721F fan control?

[Alexandros Diamantidis]

* Luca Tettamanti [2010-11-12 14:13]:
> The problem with the native driver is that it uses resources that are
> claimed by ACPI, which might be unsafe.

Hmm... How unsafe would it be? A crash once a year I wouldn't
mind much, frequent random memory corruption would be worse ;-)

* Jean Delvare [2010-11-12 19:03]:
> Can you please test the standalone driver at:
>   http://khali.linux-fr.org/devel/misc/it87/
> 
> It has fixed manual fan speed control for the IT8721F. Hopefully it
> works now...

Just tried the new version with no obvious change. pwmconfig still
doesn't seem to do anything. Also tried echo 1 > pwm[123]_enable;
echo 0 > pwm[123], which should have stopped the fans, right?

Thanks again!

Alexandros

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

Re: [lm-sensors] IT8721F fan control?

[Jean Delvare]

Hi Alexandros,

On Fri, 12 Nov 2010 22:01:25 +0200, Alexandros Diamantidis wrote:
> * Luca Tettamanti [2010-11-12 14:13]:
> > The problem with the native driver is that it uses resources that are
> > claimed by ACPI, which might be unsafe.
> 
> Hmm... How unsafe would it be? A crash once a year I wouldn't
> mind much, frequent random memory corruption would be worse ;-)

Data corruption when reading the sensor values or writing the limits.
This may simply mean that "sensors" will return wrong values once in a
while, but could also result in the BIOS taking actions to fight what
it thinks is a critical thermal condition. This would mean CPU
throttling or instant shutdown.

> * Jean Delvare [2010-11-12 19:03]:
> > Can you please test the standalone driver at:
> >   http://khali.linux-fr.org/devel/misc/it87/
> > 
> > It has fixed manual fan speed control for the IT8721F. Hopefully it
> > works now...
> 
> Just tried the new version with no obvious change. pwmconfig still
> doesn't seem to do anything. Also tried echo 1 > pwm[123]_enable;
> echo 0 > pwm[123], which should have stopped the fans, right?

Yes it should have... Hmm.

Could you please:

* Try again after disabling Q-Fan in the BIOS. Maybe I missed an
  initialization step for the manual mode.
* Both with and without Q-Fan, check whether the values you write to
  pwm[123] and pwm[123]_enable stick. Read them once 3 seconds after
  writing them, and then again 6 and 30 seconds after writing them. Test
  values 0, 1 and 2 for pwm[123]_enable, and any two random values for
  pwm[123] after writing 1 to pwm[123]_enable.
* With Q-Fan mode set, not trying to enable manual control mode, check
  the values of pwm[123]_enable. One of them at least should be 2. See
  if the corresponding reported pwmX value changes over time depending
  on CPU load.

Thanks,
-- 
Jean Delvare

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

Re: [lm-sensors] IT8721F fan control?

[Alexandros Diamantidis]

* Jean Delvare [2010-11-15 12:00]:
> > echo 0 > pwm[123], which should have stopped the fans, right?
> 
> Could you please:
> 
> * Try again after disabling Q-Fan in the BIOS. Maybe I missed an
>   initialization step for the manual mode.
[...]

Hello Jean!

First of all, I'd like a clarification, because I may be wasting your
time: I just did a few web searches about fan control, and some articles
I read indicated that to control fan speeds, you need fans with four
wires (one of which carries the PWM signal). Is this true? If so, that's
what's going on, since all my fans have three wires. The CPU fan header
on the motherboard does have four pins, but the fan only has a
three-wire connector. The other two fan headers have three pins.

So, I did the tests you indicated. First of all, a reminder that the
BIOS only has a CPU Q-Fan setting.

With Q-Fan disabled, after boot, pwm[123]_enable = 1, pwm[123] = 255.
I tried writing 0, 1 and 2 to pwm?_enable, and a couple of random values to
pwm? in each case. Reading from them, I get back the last values written, with
no change over time. Fan speeds stay almost constant regardless of CPU
temperature.

With Q-Fan enabled and set to "silent" mode, after boot, pwm1_enable = 2,
pwm1 = 54, pwm[23]_enable = 1, pwm[23] = 255. After running the system at full
load for a few minutes, CPU temperature as reported by it8721 rises from ~30°C
to ~42°C, but pwm1 always stays 54 (although CPU fan speed did rise a
bit, from ~1550 RPM to ~1610).

In this case, too, values written to pwm[123] stick and don't change over time,
for pwm[123]_enable set to 0, 1 or 2.

Thanks again!

Alexandros


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

Re: [lm-sensors] IT8721F fan control?

[Jean Delvare]

Hi Alexandros,

On Sat, 20 Nov 2010 18:14:18 +0200, Alexandros Diamantidis wrote:
> * Jean Delvare [2010-11-15 12:00]:
> > > echo 0 > pwm[123], which should have stopped the fans, right?
> > 
> > Could you please:
> > 
> > * Try again after disabling Q-Fan in the BIOS. Maybe I missed an
> >   initialization step for the manual mode.
> [...]
> 
> Hello Jean!
> 
> First of all, I'd like a clarification, because I may be wasting your
> time: I just did a few web searches about fan control, and some articles
> I read indicated that to control fan speeds, you need fans with four
> wires (one of which carries the PWM signal). Is this true? If so, that's
> what's going on, since all my fans have three wires. The CPU fan header
> on the motherboard does have four pins, but the fan only has a
> three-wire connector. The other two fan headers have three pins.

Things aren't that simple. It is perfectly possible to control a 3-wire
fan. People have been doing that for the last decade successfully. In
fact it even works with 2-wire fans.

What is true is that 4-wire fans are better at monitoring and
controlling the fan at the same time. With 3-wire fans, it was somewhat
difficult to monitor the fan when it's running at low speed (low duty
cycle or low DC). Some chips have advanced circuitry but this helps
only to some point. With 4-wire fans, the power and the control are on
different wires, which means that the monitoring signal (which goes on
a separate wire too) always has the same strength.

But the most important thing is that 3-wire and 4-wire fans aren't
controlled the same way, so it is important that the monitoring/control
chip and the motherboard agree on the number of pins. Some motherboards
have jumpers to decide whether fans have 3-wire or 4-wire are in use. I
guess that some boards could have BIOS options instead. If you have a
4-wire fan header and no option to claim you're using a 3-wire fan on
it, I guess the board assumes you're using a 4-wire fan, and a 3-wire
fan can't be controlled. It might be spelled out explicitly in the user
manual.

> So, I did the tests you indicated. First of all, a reminder that the
> BIOS only has a CPU Q-Fan setting.
> 
> With Q-Fan disabled, after boot, pwm[123]_enable = 1, pwm[123] = 255.
> I tried writing 0, 1 and 2 to pwm?_enable, and a couple of random values to
> pwm? in each case. Reading from them, I get back the last values written, with
> no change over time. Fan speeds stay almost constant regardless of CPU
> temperature.

At least this suggests that my patch isn't completely broken. But given
the above, I also start doubting that you can actually test it with
your current hardware.

> With Q-Fan enabled and set to "silent" mode, after boot, pwm1_enable = 2,
> pwm1 = 54, pwm[23]_enable = 1, pwm[23] = 255. After running the system at full

This means that PWM1 is used to control the CPU fan on your board.

> load for a few minutes, CPU temperature as reported by it8721 rises from ~30°C
> to ~42°C, but pwm1 always stays 54 (although CPU fan speed did rise a
> bit, from ~1550 RPM to ~1610).

The fact that the pwm1 value sticks to 54 suggests that the device did
not change the fan output despite the temperature increase. We'd have
to check the exact automatic control settings. You can provide a
register dump of your chip if you want me to take a look:

# rmmod it87
# isadump 0x295 0x296

You may want to try a different Q-fan profile, maybe it will react
faster to temperature changes. This would change the value of pwm1, but
probably not the actual fan speed as long as you don't have a 4-wire fan.

> In this case, too, values written to pwm[123] stick and don't change over time,
> for pwm[123]_enable set to 0, 1 or 2.

OK, thanks for the tests.

-- 
Jean Delvare

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