[lm-sensors] Testing LM-Sensor Support of SCH5127 in Acer easyStore H340

[Jeff Rickman]

Hi Jean,

> Hi Jeff,
>
> On Mon, 29 Nov 2010 00:20:44 -0600 (CST), Jeff Rickman wrote:
>> Providing some feedback now that SCH5127 support is "mainline" in
>> LM_Sensors; no "force_id" option required.
>> (...)
>> [root at anas-01 ~]# service lm_sensors status
>> coretemp-isa-0000
>> Adapter: ISA adapter
>> Core 0:      +21.0°C  (crit = +90.0°C)
>>
>> sch5127-isa-0870
>> Adapter: ISA adapter
>> V+1.5:       +1.42 V  (min =  +1.35 V, max =  +1.65 V)
>> 5VTR:        +4.78 V  (min =  +4.50 V, max =  +5.48 V)
>> VBAT:        +3.32 V  (min =  +0.00 V, max =  +4.38 V)
>> V+5:         +5.09 V  (min =  +0.00 V, max =  +6.72 V)
>> Vccp:        +1.48 V  (min =  +1.35 V, max =  +1.49 V)
>> VCC:         +3.33 V  (min =  +0.00 V, max =  +4.38 V)
>> VTR:         +3.24 V  (min =  +0.00 V, max =  +4.38 V)
>> Side Fan:   2250 RPM  (min = 600 RPM)
>> MCH Fan:    5137 RPM  (min = 4500 RPM)
>> CPU Temp:    +47.6°C  (low  = +20.0°C, high = +60.0°C)
>> SYS Temp:    +35.1°C  (low  = +20.0°C, high = +60.0°C)
>>
>> Here is my "/etc/sensors.d/local.conf" file:
>>
>> [root at anas-01 ~]# cat /etc/sensors.d/local.conf
>> # libsensors configuration file
>> # -----------------------------
>>
>> chip "sch5127-isa-0870"
>>
>>     label in0 "V+1.5"
>>     label in1 "5VTR"
>>     label in2 "VBAT"
>>     label in3 "V+5"
>>     label in4 "Vccp"
>>     label in5 "VCC"
>>     label in6 "VTR"
>>
>> # All inputs are listed here in order as displayed in BIOS.
>
> This is often the case, but not always. So this should only be used as
> a last resort decision factor.

I did attempt to match by voltage values between "sensors" and BIOS.

>> # Values seen by "sensors" closely match values seen in BIOS.
>>
>>     label fan1 "Side Fan"
>>     label fan2 "MCH Fan"
>>     ignore fan3
>>
>>     label temp1 "CPU Temp"
>>     ignore temp2
>> #    label temp2 "SIO Temp"
>
> This is an internal sensor, it should always be present and correct, so
> why would you ignore it?
>

The value is high...very high. Does a flucuating value between 120 and 128
Celsius make sense?

>>     label temp3 "SYS Temp"
>>
>>     compute in0 (@ * 0.8), (@ / 0.8)

Removing this compute line shows a fairly stable in0 value of 1.78

> This is extremely unlikely. While it is frequent to scale down voltage
> inputs so that the results fits in the ADC range, there is no point in
> scaling a voltage up. Especially not by a factor so close to 1: the
> error incurred by the operation would far outweigh the resolution
> improvement at the ADC level. Not to mention that scaling up needs an
> amplifier so it's not a cheap operation - no PC vendor would do this.
>
> So, I seriously doubt that in0 is +1.5V.
>
> In fact, the SCH5127 already has internal scaling resistors on most
> voltage inputs, so if the voltage lines are wired correctly, only
> voltages over +3.3V should need external scaling. These voltages (+5V
> and 5VTR in your case) should be wired to in3 and in4 per chip design,
> to limit the number of inputs which require scaling. VBAT, VCC and VTR
> should be on in2, in5 and in6, which matches your findings. Which means
> that +1.5V and Vccp should be on in0 and in1 (or vice-versa.)

I will dig out this chassis and install a video card in it so I can access
the BIOS values. I can get the system to boot very fast (<=1m boot to
login prompt) so checking "sensors" values against BIOS values is
reasonable.

> Now it is of course possible that Acer did things differently, either
> for pin proximity reasons, or just because they are bad ;)

They can be bad so anything is possible ;)

> Assuming that your CPU does frequency and voltage scaling based on
> load, you should try to put some load on the CPU and check which
> voltage input raises. This would be Vcore (Vccp) and should require no
> scaling. If you can figure that one out, it might help sort out the
> rest.

I will need to find some type of CPU stress program. Even moving 250,000
files (about 70+GB) between hard drives inside the chassis using Rsync
only placed <5% load on the CPU.

>>     compute in1 (@ * 4), (@ / 4)
>>     compute in3 (@ * 4.5), (@ / 4.5)
>>
>>     set in0_min 1.5 * 0.90
>>     set in0_max 1.5 * 1.10
>>     set in1_min 5.0 * 0.90
>>     set in1_max 5.0 * 1.10
>>     set in4_min 1.50 * 0.90
>>     set in4_max 1.50 * 1.10
>>
>>     set fan1_min 600 <- per fan vendor
>
> You're missing a # before your comment.

Comment added during "cut & paste" into email....

>>     set fan2_min 4500
>>     set temp1_min 20
>>     set temp1_max 60
>>     set temp2_min 20
>>     set temp3_min 20
>>     set temp3_max 60
>
> --
> Jean Delvare
>