[SOLVED] lm_sensors (k10temp)

coralfang · 06-30-2014, 08:20 PM

Been meaning to ask this for a while, as i have noticed the same behaviour on all kinds of distributions and kernels.

Code:

$ uname -p
AMD FX(tm)-8320 Eight-Core Processor

Code:

$ sensors
atk0110-acpi-0
Adapter: ACPI interface
Vcore Voltage:      +1.09 V  (min =  +0.80 V, max =  +1.60 V)
+3.3V Voltage:      +3.38 V  (min =  +2.97 V, max =  +3.63 V)
+5V Voltage:        +5.04 V  (min =  +4.50 V, max =  +5.50 V)
+12V Voltage:      +12.18 V  (min = +10.20 V, max = +13.80 V)
CPU Fan Speed:      857 RPM  (min =  600 RPM, max = 7200 RPM)
Chassis Fan Speed:    0 RPM  (min =  600 RPM, max = 7200 RPM)
CPU Temperature:    +37.0°C  (high = +60.0°C, crit = +95.0°C)
MB Temperature:     +28.0°C  (high = +45.0°C, crit = +75.0°C)

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +21.4°C  (high = +70.0°C)
                       (crit = +83.5°C, hyst = +80.5°C)

fam15h_power-pci-00c4
Adapter: PCI adapter
power1:      102.87 W  (crit = 125.02 W)

What i want to know is, what exactly is CPU Temperature? Is it provided by the motherboard, and reading a sensor nearby the cpu? Should this be higher than the actual cpu temperature?

AFAIK, k10temp-* reads values from the sensor built into the CPU. This value always differs from the value provided by the acpi interface.

Typical values for the k10temp-* output on a cool day will be 20°C on this machine, and if i make the cpu busy with a `make -j9' for an extended period of time it maxes out at roughly 62°C. I recently installed a hyper 212 heatsink, and i'm wondering if these sound like correct (sane) temperature values, assuming that the k10temp-* sensor is displaying accurate values.

If i leave `watch -n1 sensors', i noticed that CPU Temperature doesn't update very often compared to the k10temp-* value.

Which is the REAL cpu temperature? :S

notsure · 06-30-2014, 08:36 PM

--I'm starting this post over again again--

I don't believe the K10temp reading is your CPU temp. 21C VERY low for that CPU. What is your ambient temperature?

Boot into your BIOS and check the CPU temp. I'll guess it reports closer to 37C.

What does "CPU Temperature" report when under load compared to the K10temp reading?

BTW what is your motherboard?

Code:

dmidecode | grep Product

Thanks!

273 · 07-01-2014, 09:29 AM

I don't think you can get a real CPU temperature on these modern AMD CPUs. As I understand it one of the temperatures is a false temperature which doesn't have any units (despite being labelled as C) and that temperature is the one that AMD recommends doesn't go above about 55. However, I have never been able to confirm for certain which temperature that is and all my temperatures can reach well above 55 under heavy load despite having a fairly decent and large cooler.
I've come to the conclusion that AMD don't care about their customers knowing whether their chips are too hot or not and that AMD chips tend to heat up very quickly anyway. As such my next desktop machine will have an Intel part assuming it's AMD64 architecture.

coralfang · 07-01-2014, 12:03 PM

Quote:

Originally Posted by notsure

--I'm starting this post over again again--

I don't believe the K10temp reading is your CPU temp. 21C VERY low for that CPU. What is your ambient temperature?

Boot into your BIOS and check the CPU temp. I'll guess it reports closer to 37C.

Ah, i checked the bios, and the CPU temp there seems to be the same value provided by the acpi output after watching it for a while.

idle CPU temperature in `sensors' = 37-38C
idle bios = 37-38C

So that seems to be the real cpu temperature i guess.

Quote:

Originally Posted by notsure

What does "CPU Temperature" report when under load compared to the K10temp reading?

It reaches exactly 60C by the time the full speeds up alot, then it drops down to 53C and slowly goes back and forth but doesn't exceed 60C, i assume that is good (sane) temperatures?

Quote:

Originally Posted by notsure

BTW what is your motherboard?

Code:

dmidecode | grep Product

Thanks!

The motherboard is an ASUS M5A78L-M/USB3.

Here's a screenshot of the sensors under full load if it's any use:

273 · 07-01-2014, 12:18 PM

Sorry, forgot to link to the source of the information I mentioned earlier:
http://www.overclock.net/t/1128821/a...tion-and-guide

notsure · 07-01-2014, 01:09 PM

After doing some research and testing I tend to agree with 273.

The idle temp is incorrect.

I (installed Windows on a spare SSD) installed the MSI BIOS app (allows adjusting the BIOS values in Windows) and Core Temp.

MSI utility / Core Temp

Idle: 37C / 20C
Prime95 load after ~5 minutes: 54C / 52C
Stopped Prime95, recorded temp after 1 minute: 39C / 23C
recorded temp after 3 minutes: 37C / 21C

And linux tests:

Idle:

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +21.4 C  (high = +70.0 C)

f71889a-isa-0480
fan1:        5791 RPM
fan2:        2269 RPM
fan3:        2086 RPM
temp1:        +32.0 C  (high = +85.0 C, hyst = +81.0 C)
                       (crit = +80.0 C, hyst = +76.0 C)  sensor = transistor
temp2:        +42.0 C  (high = +85.0 C, hyst = +79.0 C)
                       (crit = +109.0 C, hyst = +103.0 C)  sensor = thermistor
temp3:        +43.0 C  (high = +70.0 C, hyst = +68.0 C)
                       (crit = +85.0 C, hyst = +83.0 C)  sensor = transistor

Load using burnK7(8 processes) 1minute:

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +56.0 C  (high = +70.0 C)

f71889a-isa-0480
fan1:        6607 RPM
fan2:        2377 RPM
fan3:        2068 RPM
temp1:        +49.0 C  (high = +85.0 C, hyst = +81.0 C)
                       (crit = +80.0 C, hyst = +76.0 C)  sensor = transistor
temp2:        +62.0 C  (high = +85.0 C, hyst = +79.0 C)
                       (crit = +109.0 C, hyst = +103.0 C)  sensor = thermistor
temp3:        +47.0 C  (high = +70.0 C, hyst = +68.0 C)
                       (crit = +85.0 C, hyst = +83.0 C)  sensor = transistor

Load 3minutes:

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +58.1 C  (high = +70.0 C)

f71889a-isa-0480
fan1:        7177 RPM
fan2:        2415 RPM
fan3:        2057 RPM
temp1:        +52.0 C  (high = +85.0 C, hyst = +81.0 C)
                       (crit = +80.0 C, hyst = +76.0 C)  sensor = transistor
temp2:        +73.0 C  (high = +85.0 C, hyst = +79.0 C)
                       (crit = +109.0 C, hyst = +103.0 C)  sensor = thermistor
temp3:        +49.0 C  (high = +70.0 C, hyst = +68.0 C)
                       (crit = +85.0 C, hyst = +83.0 C)  sensor = transistor

Stopped burnK7, record after 1minute:

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +26.0 C  (high = +70.0 C)

f71889a-isa-0480
fan1:        6976 RPM
fan2:        2133 RPM
fan3:        2063 RPM
temp1:        +37.0 C  (high = +85.0 C, hyst = +81.0 C)
                       (crit = +80.0 C, hyst = +76.0 C)  sensor = transistor
temp2:        +53.0 C  (high = +85.0 C, hyst = +79.0 C)
                       (crit = +109.0 C, hyst = +103.0 C)  sensor = thermistor
temp3:        +50.0 C  (high = +70.0 C, hyst = +68.0 C)
                       (crit = +85.0 C, hyst = +83.0 C)  sensor = transistor

After 3minutes stopped:

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +23.5 C  (high = +70.0 C)

f71889a-isa-0480
fan1:        6726 RPM
fan2:        2127 RPM
fan3:        2074 RPM
temp1:        +34.0 C  (high = +85.0 C, hyst = +81.0 C)
                       (crit = +80.0 C, hyst = +76.0 C)  sensor = transistor
temp2:        +45.0 C  (high = +85.0 C, hyst = +79.0 C)
                       (crit = +109.0 C, hyst = +103.0 C)  sensor = thermistor
temp3:        +50.0 C  (high = +70.0 C, hyst = +68.0 C)
                       (crit = +85.0 C, hyst = +83.0 C)  sensor = transistor

And this is the full output of sensors (idle) FYI:

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +21.4 C  (high = +70.0 C)

fam15h_power-pci-00c4
Adapter: PCI adapter
power1:      124.77 W  (crit = 125.02 W)

f71889a-isa-0480
Adapter: ISA adapter
+3.3V:        +3.20 V  
in1:          +1.45 V  (max =  +2.04 V)
in2:          +1.09 V  
in3:          +0.86 V  
in4:          +0.62 V  
in5:          +1.34 V  
in6:          +1.61 V  
3VSB:         +3.25 V  
Vbat:         +3.20 V  
fan1:        5791 RPM
fan2:        2269 RPM
fan3:        2086 RPM
temp1:        +32.0 C  (high = +85.0 C, hyst = +81.0 C)
                       (crit = +80.0 C, hyst = +76.0 C)  sensor = transistor
temp2:        +42.0 C  (high = +85.0 C, hyst = +79.0 C)
                       (crit = +109.0 C, hyst = +103.0 C)  sensor = thermistor
temp3:        +43.0 C  (high = +70.0 C, hyst = +68.0 C)
                       (crit = +85.0 C, hyst = +83.0 C)  sensor = transistor

f71889a temp1 looks similar to the MSI utility reading.

k10temp shares the same idle temp as Core Temp, but the load temp is off by a good amount.

In conclusion, I'm notsure.

273 · 07-01-2014, 01:16 PM

In the post I linked to he suggests that the core temp doesn't read a proper temperature until it reaches 45C, which it displays correctly, but what he doesn't say is whether it reflects the true temperature after that.
After seeing my CPU hit what AMD lists as dangerous temperatures even with a beefy after-market cooler on I've stopped worrying and decided if it goes then it's junk anyway and I'll just have to spend my hard earned money on an Intel setup if it does.
That is probably a gross overreaction on my part but after much googleing I don't have any faith in AMD.

notsure · 07-01-2014, 06:15 PM

Ya, good link you posted. It seems we're stuck with inaccurate temperature readings that maybe just give us an idea of the temps under load.

I have taxed that AMD workstation heavily throughout the past 2 years.

I must say the performance was excellent for multitasking while compressing data for 5 days straight and running 2 Windows VMs and remote work with X11vnc on a large display (2560x1600+1920x1080) @100% CPU. I'd like to compare it to an i7 but I don't send all my money to china anymore.

273 · 07-01-2014, 06:27 PM

I have to say that I have no issues with the performance of my FX 8120 and it runs VMs well in VirtualBox. In day to day use it appears to be a decent CPU. I just don't like AMDs attitude of not bothering to let their customers know whether their CPU is running at a sensible temperature and the very low limits they give are worrying. I also don't like that theoretically my CPU is out of warranty due to a third party cooler when it gives readings 10"C" lower than the stock.
If AMD actually commented on this with some decent information I'd not recommend Intel but AMD really do seem not to care at all. oh, the AMD microcode site was also down for a few months which for a company that large doesn't bode well.