Welcome, Guest. Please login or register.
Did you miss your activation email?

Author Topic: [EN] k10temp problem with latest kernel  (Read 5985 times)

Offline jyp

  • User
  • Posts: 118
[EN] k10temp problem with latest kernel
« on: 2021/11/09, 20:42:48 »
Tdie does not show with latest kernel. I  reconfigured lm-sensors but no result.
Not a big thing but surprising and a bit annoying.

Code: [Select]
:~$ uname -a
Linux gamma 5.15.1-1-siduction-amd64 #1 SMP PREEMPT siduction 5.15-1 (2021-11-06) x86_64 GNU/Linux

:~$ sensors
...
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +45.2°C
Tccd1:        +31.0°C
...

Code: [Select]
~$ uname -a
Linux gamma 5.14.15-2-siduction-amd64 #1 SMP PREEMPT siduction 5.14-15.1 (2021-10-28) x86_64 GNU/Linux

:~$ sensors
...
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +41.1°C
Tdie:         +41.1°C
Tccd1:        +30.8°C
...

Thanks for your attention
jyp

Offline samoht

  • User
  • Posts: 478
Re: k10temp problem with latest kernel
« Reply #1 on: 2021/11/09, 23:50:44 »
Confirmed here:
No sensors at all found on current kernel.

On the other hand with kernel 5.14.16-2-siduction-amd64:

Code: [Select]
$ sensors
amdgpu-pci-0400
Adapter: PCI adapter
vddgfx:      718.00 mV
vddnb:       999.00 mV
edge:         +31.0°C
power1:      1000.00 uW

nvme-pci-0300
Adapter: PCI adapter
Composite:    +36.9°C  (low  =  -0.1°C, high = +79.8°C)
                       (crit = +83.8°C)
Sensor 1:     +36.9°C  (low  = -273.1°C, high = +65261.8°C)

k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +33.5°C
Tdie:         +33.5°C
« Last Edit: 2021/11/09, 23:54:06 by samoht »

Offline fams

  • User
  • Posts: 37
Re: k10temp problem with latest kernel
« Reply #2 on: 2021/11/10, 07:42:22 »
Same here (Linux Ryzen 5.15.1-1-siduction-amd64 #1 SMP PREEMPT siduction 5.15-1 (2021-11-06) x86_64 GNU/Linux)
Code: [Select]
amdgpu-pci-0a00
Adapter: PCI adapter
vddgfx:      943.00 mV
fan1:          10 RPM  (min =    0 RPM, max = 4600 RPM)
edge:         +45.0°C  (crit = +94.0°C, hyst = -273.1°C)
power1:       17.03 W  (cap =  48.00 W)

k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +36.0°C 
Tccd1:        +32.5°C 

nvme-pci-0100
Adapter: PCI adapter
Composite:    +36.9°C  (low  = -273.1°C, high = +84.8°C)
                       (crit = +84.8°C)
Sensor 1:     +36.9°C  (low  = -273.1°C, high = +65261.8°C)
Sensor 2:     +36.9°C  (low  = -273.1°C, high = +65261.8°C)

gigabyte_wmi-virtual-0
Adapter: Virtual device
temp1:        +31.0°C 
temp2:        +30.0°C 
temp3:        +36.0°C 
temp4:        +34.0°C 
temp5:        +33.0°C 
temp6:        +41.0°C 

acpitz-acpi-0
Adapter: ACPI interface
temp1:        +16.8°C  (crit = +20.8°C)



I remember that I read somewhere that Tdie and Tclt are linked by a fixed factor (by AMD) and gave the hottest point.
This would  be an explanation for that ommission.
Tccd1 should be the temperatur of one CoreComplexDie, so one value for each core. Question is why is only one value (for core 1) is shown.
But maybe a more competent person comments here...
Some googling may help, too   ;)
« Last Edit: 2021/11/10, 07:46:04 by fams »

Offline unklarer

  • User
  • Posts: 816
Re: k10temp problem with latest kernel
« Reply #3 on: 2021/11/10, 10:01:53 »
Have you found this statement yet?
Quote
That depends on particular model and Tctl=Tdie only if the CPU doesn't use an offset (Tctl_offset).
If you see in sensors a "CPU (Tctl/Tdie)" value, it means they are same. Otherwise there will be 2 values shown Tctl and Tdie.

I don't have a Ryzen and everything is fine with me.   ;)
Code: [Select]
sensors -u
coretemp-isa-0000
Adapter: ISA adapter
Core 0:
  temp2_input: 36.000
  temp2_max: 74.000
  temp2_crit: 100.000
  temp2_crit_alarm: 0.000
Core 1:
  temp3_input: 36.000
  temp3_max: 74.000
  temp3_crit: 100.000
  temp3_crit_alarm: 0.000
Core 2:
  temp4_input: 33.000
  temp4_max: 74.000
  temp4_crit: 100.000
  temp4_crit_alarm: 0.000
Core 3:
  temp5_input: 32.000
  temp5_max: 74.000
  temp5_crit: 100.000
  temp5_crit_alarm: 0.000

radeon-pci-0100
Adapter: PCI adapter
temp1:
  temp1_input: 50.000
  temp1_crit: 120.000
  temp1_crit_hyst: 90.000

atk0110-acpi-0
Adapter: ACPI interface
Vcore Voltage:
  in0_input: 1.064
  in0_min: 0.800
  in0_max: 1.600
 +3.3 Voltage:
  in1_input: 3.248
  in1_min: 2.970
  in1_max: 3.630
 +5 Voltage:
  in2_input: 5.040
  in2_min: 4.500
  in2_max: 5.500
 +12 Voltage:
  in3_input: 11.928
  in3_min: 10.200
  in3_max: 13.800
CPU FAN Speed:
  fan1_input: 691.000
  fan1_min: 600.000
  fan1_max: 7200.000
CHASSIS1 FAN Speed:
  fan2_input: 225.000
  fan2_min: 600.000
  fan2_max: 7200.000
CHASSIS2 FAN Speed:
  fan3_input: 795.000
  fan3_min: 600.000
  fan3_max: 7200.000
POWER FAN Speed:
  fan4_input: 1430.000
  fan4_min: 600.000
  fan4_max: 7200.000
CPU Temperature:
  temp1_input: 23.500
  temp1_max: 60.000
  temp1_crit: 95.000
MB Temperature:
  temp2_input: 37.000
  temp2_max: 45.000
  temp2_crit: 95.000

Offline Mister00X

  • User
  • Posts: 198
Re: k10temp problem with latest kernel
« Reply #4 on: 2021/11/10, 13:02:19 »
I've tested this on my ryzen laptop.

With kernel 5.14.16-2 the output of sensors is:
Code: [Select]
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +40.0°C 
Tdie:         +40.0°C 

BAT0-acpi-0
Adapter: ACPI interface
in0:          12.07 V 
curr1:            N/A 

nvme-pci-0100
Adapter: PCI adapter
Composite:    +42.9°C  (low  = -273.1°C, high = +81.8°C)
                       (crit = +84.8°C)
Sensor 1:     +42.9°C  (low  = -273.1°C, high = +65261.8°C)

amdgpu-pci-0400
Adapter: PCI adapter
vddgfx:           N/A 
vddnb:            N/A 
edge:         +40.0°C 

acpitz-acpi-0
Adapter: ACPI interface
temp1:        +39.0°C  (crit = +125.0°C)
temp2:         +0.0°C  (crit = +200.0°C)

With kernel 5.15.1-1 it's:
Code: [Select]
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +45.1°C 

BAT0-acpi-0
Adapter: ACPI interface
in0:          12.01 V 
curr1:            N/A 

nvme-pci-0100
Adapter: PCI adapter
Composite:    +50.9°C  (low  = -273.1°C, high = +81.8°C)
                       (crit = +84.8°C)
Sensor 1:     +50.9°C  (low  = -273.1°C, high = +65261.8°C)

amdgpu-pci-0400
Adapter: PCI adapter
vddgfx:           N/A 
vddnb:            N/A 
edge:         +45.0°C 

acpitz-acpi-0
Adapter: ACPI interface
temp1:        +45.0°C  (crit = +125.0°C)
temp2:         +0.0°C  (crit = +200.0°C)

So Tdie is missing on 5.15.1-1.
One side effect caused by this is that htop is no longer able to display the cpu temperature :-/

I will also test this on my ryzen desktop PC and will report back.
Arguing that you don't care about the right to privacy because you have nothing to hide is no different than saying you don't care about free speech because you have nothing to say. – Edward Snowden

Offline whistler_mb

  • User
  • Posts: 198
Re: k10temp problem with latest kernel
« Reply #5 on: 2021/11/10, 13:21:31 »
On my desktop PC with an AMD 4700G it behaves the same as on Mister00X's laptop.

Code: [Select]
~$ uname -r
5.14.16-1-siduction-amd64

~$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +49.8°C
Tdie:         +49.8°C

nvme-pci-0a00
Adapter: PCI adapter
Composite:    +38.9°C  (low  =  -0.1°C, high = +86.8°C)
                       (crit = +89.8°C)
Sensor 1:     +39.9°C  (low  = -273.1°C, high = +65261.8°C)

amdgpu-pci-0b00
Adapter: PCI adapter
vddgfx:      724.00 mV
vddnb:       724.00 mV
edge:         +41.0°C
power1:        0.00 W

Code: [Select]
~$ uname -r
5.15.1-1-siduction-amd64

~$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +47.2°C


nvme-pci-0a00
Adapter: PCI adapter
Composite:    +37.9°C  (low  =  -0.1°C, high = +86.8°C)
                       (crit = +89.8°C)
Sensor 1:     +37.9°C  (low  = -273.1°C, high = +65261.8°C)

amdgpu-pci-0b00
Adapter: PCI adapter
vddgfx:      731.00 mV
vddnb:       937.00 mV
edge:         +38.0°C
power1:      1000.00 uW

Offline fams

  • User
  • Posts: 37
Re: k10temp problem with latest kernel
« Reply #6 on: 2021/11/10, 14:11:33 »
Tdie and Tctl are linked by a fixed factor (in your case 0), so omission of one seems to be correct.
What is missing is Tccd1 (and others...) that seem to give the real (not peak) cpu core temperature(s).
On my laptop with Bullseye Kernel 5.10.0 and Ryzen 7 5700u I have only Tdie and Tctl with the same value, too.

Offline Mister00X

  • User
  • Posts: 198
Re: k10temp problem with latest kernel
« Reply #7 on: 2021/11/10, 17:14:46 »
So I have now also testet it on my ryzen desktop PC Tdie is missing on the 5.15.1 Kernel too and with it the temperatures in htop.

Edit: This appears to be intended though according to https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/hwmon/k10temp.c?id=02a2484cf8d17a2acf3b9b151147bafaa55ad38c

« Last Edit: 2021/11/10, 17:18:26 by Mister00X »
Arguing that you don't care about the right to privacy because you have nothing to hide is no different than saying you don't care about free speech because you have nothing to say. – Edward Snowden

Offline jyp

  • User
  • Posts: 118
Re: k10temp problem with latest kernel
« Reply #8 on: 2021/11/10, 18:48:30 »
To say the least, it seems that, on amd ryzen, temperatures monitoring is not very reliable.

For instance, issuing <sensors>, I get _Tccd1 = +29.0°C_ and conky shows _Tccd1 = +37.0°C_ using the same source (/usr/bin/sensors | grep Tccd1). Really strange.

Offline samoht

  • User
  • Posts: 478
Re: k10temp problem with latest kernel
« Reply #9 on: 2021/11/11, 00:13:50 »
Remaining question:
What is the reason for the different results with the two kernels?

Code: [Select]
CPU: Quad Core AMD Ryzen 3 PRO 4350GE with Radeon Graphics
Legacy kernel:

Code: [Select]
$ uname -a
Linux tuxxy2-sid 5.14.16-2-siduction-amd64 #1 SMP PREEMPT siduction 5.14-16.1 (2021-11-04) x86_64 GNU/Linux

$ sensors
amdgpu-pci-0400
Adapter: PCI adapter
vddgfx:      718.00 mV
vddnb:       724.00 mV
edge:         +29.0°C
power1:        0.00 W

nvme-pci-0300
Adapter: PCI adapter
Composite:    +36.9°C  (low  =  -0.1°C, high = +79.8°C)
                       (crit = +83.8°C)
Sensor 1:     +36.9°C  (low  = -273.1°C, high = +65261.8°C)

k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +32.4°C
Tdie:         +32.4°C

$ inxi -s
Sensors:   System Temperatures: cpu: 40.0 C mobo: N/A gpu: amdgpu temp: 31.0 C
           Fan Speeds (RPM): N/A

Current kernel:

Code: [Select]
$ uname -a
Linux tuxxy2-sid 5.15.1-3-siduction-amd64 #1 SMP PREEMPT siduction 5.15-1.2 (2021-11-10) x86_64 GNU/Linux

$ sensors
amdgpu-pci-0400
Adapter: PCI adapter
vddgfx:      718.00 mV
vddnb:       999.00 mV
edge:         +28.0°C
power1:      1000.00 uW

nvme-pci-0300
Adapter: PCI adapter
Composite:    +37.9°C  (low  =  -0.1°C, high = +79.8°C)
                       (crit = +83.8°C)
Sensor 1:     +37.9°C  (low  = -273.1°C, high = +65261.8°C)

k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +31.1°C

$ inxi -s
Sensors:   Message: No sensor data found. Is lm-sensors configured?
« Last Edit: 2021/11/11, 00:20:01 by samoht »

Offline unklarer

  • User
  • Posts: 816
Re: k10temp problem with latest kernel
« Reply #10 on: 2021/11/11, 09:39:27 »
Quote from: samoht
Remaining question:
What is the reason for the different results with the two kernels?

You may read the linked pages....    ;D

Offline samoht

  • User
  • Posts: 478
Re: k10temp problem with latest kernel
« Reply #11 on: 2021/11/12, 08:29:56 »
Quote
Remaining question:
What is the reason for the different results with the two kernels?

Quote
$ inxi -s
Sensors:   System Temperatures: cpu: 40.0 C mobo: N/A gpu: amdgpu temp: 31.0 C
           Fan Speeds (RPM): N/A

versus

Quote
$ inxi -s
Sensors:   Message: No sensor data found. Is lm-sensors configured?

Offline unklarer

  • User
  • Posts: 816
Re: k10temp problem with latest kernel
« Reply #12 on: 2021/11/12, 10:02:09 »
I had meant that it was because of the 'unification'
of the values Tctl and Tdie for certain models.

However, you mean something else.

1. Linux has NEVER been at the table for the production of new hardware at the corporations.

2. you have not configured sensors.

3. it might be worth looking at hwmon, which seems the better choice for laptop machines.

Offline samoht

  • User
  • Posts: 478
Re: k10temp problem with latest kernel
« Reply #13 on: 2021/11/13, 00:13:32 »
Thanks for trying to help.

Quote
... you have not configured sensors.

No, that message comes only with the current kernels

Quote
$ inxi -s
Sensors:   Message: No sensor data found. Is lm-sensors configured?

but not with the kernel 14 releases.

Offline unklarer

  • User
  • Posts: 816
Re: k10temp problem with latest kernel
« Reply #14 on: 2021/11/13, 10:37:03 »
Sorry, what's wrong with running sensors-detect again?  Or have you already done that?

You don't say anything about hwmon either. I would be interested in the output of this command:
Code: [Select]
$ for m in /sys/class/hwmon/* ; do echo -n "$m = " ; cat $m/name ; done
Everything with the "new" kernels, of course.