Linux k10temp Driver For AMD CPUs Updated To Better Handle Power/Temp Analysis

[boxie]

I have been lobbying AMD in person at events and via emails over this matter for a while now... Unfortunately no apparent progress yet.

Keep at em!

For all the good products they have right now we need to keep on them to make sure they do not rest

[agd5f]

I'm not too familiar with the CPU side documentation, but if it's anything like the GPU side, there are no ready made programming guides just sitting around waiting to be released. Most of the documentation is sprinkled here and there in hardware design documents which are not suitable to release as is. Someone would need to write up something suitable for public consumption. It's probably as much or more effort to write up the documentation as it would be to write the code in this case. It sounds like in some cases there may be platform specific changes that are handled by the OEM. They may need to provide that information. I agree that having this information is nice, but it's not really critical to the platform's operation. The power management microcontoller on the CPU handles the clocking, power, and thermals dynamically and things like ACPI thermal zones cover critical areas of the platform in things like laptops and servers. Keep in mind that windows in still like 90% of the consumer market. Having feature parity requires a similar level of engineering resources on each OS.

[Hibbelharry]

Wouldn't EPYC servers look more attractive with proper thermal and power reporting on Linux?
Come on, AMD!

Thats not really a problem on servers. Monitoring is usually done via dedicated monitoring and maintenance ICs called BMC (Baseband management Controllers) like HPE iLO,.. These are mostly own autonomous systems inside the system, operating their own OS, offering additional features like a virtual console and such stuff. If you want to get any data like temperatures, health, power consumption and alike you interface to those.

This is really mostly a desktop user problem.

[shmerl]

I have been lobbying AMD in person at events and via emails over this matter for a while now... Unfortunately no apparent progress yet.

Did AMD explain why they don't do it yet?

[Michael]

Did AMD explain why they don't do it yet?

Cost / risk / benefit analysis seems to be the summary.

[S.Pam]

Cost / risk / benefit analysis seems to be the summary.

Seems silly. You (Phoronix) must be driving a lot of Windows customers to because of your in - depth benchmarks and analysis. Poor remarks must surely reflect on their reputation as a whole?

[c2p_]

Ryzen 2600 idle:

$ sensors
it8792-isa-0a60
Adapter: ISA adapter
CPU Vcore:      +0.81 V  (min =  +0.35 V, max =  +1.45 V)
DDR VTT:        +0.60 V  (min =  +0.55 V, max =  +0.80 V)
Chipset Core:   +1.06 V  (min =  +0.99 V, max =  +1.10 V)
+3.3V:          +3.36 V  (min =  +0.00 V, max =  +5.56 V)
CPU Vdd18:      +1.78 V  (min =  +1.74 V, max =  +1.85 V)
DDR Vpp A/B:    +2.52 V  (min =  +2.38 V, max =  +2.63 V)
in6:            +2.78 V  (min =  +0.00 V, max =  +2.78 V)  ALARM
3VSB:           +3.33 V  (min =  +3.21 V, max =  +3.40 V)
Vbat:           +3.14 V  
SYS_FAN5_PUMP: 1080 RPM  (min =    0 RPM)
SYS_FAN6_PUMP:    0 RPM  (min =    0 RPM)
SYS_FAN4:      1092 RPM  (min =    0 RPM)
PCIEX8:         +32.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
EC_TEMP:        +25.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
System 2:       +31.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
intrusion0:    ALARM

k10temp-pci-00c3
Adapter: PCI adapter
Vcore:        +0.82 V  
Vsoc:         +0.93 V  
Tdie:         +32.0°C  
Tctl:         +32.0°C  
Icore:        +7.00 A  
Isoc:         +7.75 A  

it8686-isa-0a40
Adapter: ISA adapter
CPU Vcore:      +0.82 V  (min =  +0.35 V, max =  +1.45 V)
+3.3V:          +3.25 V  (min =  +5.05 V, max =  +5.05 V)
+12V:          +12.17 V  (min = +18.36 V, max = +18.36 V)
+5V:            +4.95 V  (min =  +7.65 V, max =  +7.65 V)
CPU Vcore SOC:  +0.94 V  (min =  +0.90 V, max =  +1.26 V)
CPU Vddp:       +0.91 V  (min =  +0.85 V, max =  +0.95 V)
DRAM A/B:       +1.22 V  (min =  +1.10 V, max =  +1.60 V)
3VSB:           +3.26 V  (min =  +0.00 V, max =  +6.12 V)
Vbat:           +3.07 V  
CPU_FAN:       1054 RPM  (min =    0 RPM)
SYS_FAN1:       815 RPM  (min =    0 RPM)
SYS_FAN2:         0 RPM  (min =    0 RPM)
SYS_FAN3:         0 RPM  (min =    0 RPM)
CPU_OPT:          0 RPM  (min =    0 RPM)
System 1:       +29.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
Chipset:        +38.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
CPU Socket:     +32.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = AMD AMDSI
PCIEX16:        +31.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
VRM MOS:        +39.0°C  (low  =  +0.0°C, high = -118.0°C)  sensor = thermistor
VSOC MOS:       +41.0°C  (low  =  +0.0°C, high = -118.0°C)  sensor = thermistor
intrusion0:    ALARM

Load (Blender benchmark):

$ sensors
it8792-isa-0a60
Adapter: ISA adapter
CPU Vcore:      +1.24 V  (min =  +0.35 V, max =  +1.45 V)
DDR VTT:        +0.60 V  (min =  +0.55 V, max =  +0.80 V)
Chipset Core:   +1.06 V  (min =  +0.99 V, max =  +1.10 V)
+3.3V:          +3.36 V  (min =  +0.00 V, max =  +5.56 V)
CPU Vdd18:      +1.78 V  (min =  +1.74 V, max =  +1.85 V)
DDR Vpp A/B:    +2.52 V  (min =  +2.38 V, max =  +2.63 V)
in6:            +2.78 V  (min =  +0.00 V, max =  +2.78 V)  ALARM
3VSB:           +3.33 V  (min =  +3.21 V, max =  +3.40 V)
Vbat:           +3.14 V  
SYS_FAN5_PUMP: 1290 RPM  (min =    0 RPM)
SYS_FAN6_PUMP:    0 RPM  (min =    0 RPM)
SYS_FAN4:      1305 RPM  (min =    0 RPM)
PCIEX8:         +32.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
EC_TEMP:        +25.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
System 2:       +31.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
intrusion0:    ALARM

k10temp-pci-00c3
Adapter: PCI adapter
Vcore:        +1.23 V  
Vsoc:         +0.93 V  
Tdie:         +59.5°C  
Tctl:         +59.5°C  
Icore:       +79.00 A  
Isoc:         +8.00 A  

it8686-isa-0a40
Adapter: ISA adapter
CPU Vcore:      +1.28 V  (min =  +0.35 V, max =  +1.45 V)
+3.3V:          +3.23 V  (min =  +5.05 V, max =  +5.05 V)
+12V:          +12.10 V  (min = +18.36 V, max = +18.36 V)
+5V:            +4.95 V  (min =  +7.65 V, max =  +7.65 V)
CPU Vcore SOC:  +0.96 V  (min =  +0.90 V, max =  +1.26 V)
CPU Vddp:       +0.91 V  (min =  +0.85 V, max =  +0.95 V)
DRAM A/B:       +1.22 V  (min =  +1.10 V, max =  +1.60 V)
3VSB:           +3.26 V  (min =  +0.00 V, max =  +6.12 V)
Vbat:           +3.07 V  
CPU_FAN:       1412 RPM  (min =    0 RPM)
SYS_FAN1:      1403 RPM  (min =    0 RPM)
SYS_FAN2:         0 RPM  (min =    0 RPM)
SYS_FAN3:         0 RPM  (min =    0 RPM)
CPU_OPT:          0 RPM  (min =    0 RPM)
System 1:       +29.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
Chipset:        +38.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
CPU Socket:     +59.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = AMD AMDSI
PCIEX16:        +31.0°C  (low  = +127.0°C, high = +127.0°C)  sensor = thermistor
VRM MOS:        +53.0°C  (low  =  +0.0°C, high = -118.0°C)  sensor = thermistor
VSOC MOS:       +44.0°C  (low  =  +0.0°C, high = -118.0°C)  sensor = thermistor
intrusion0:    ALARM

I don't know which Vcore and Vsoc is real - 3 different results.

[indepe]

I'd think AMD has an obligation to allow its Linux customers access to this information.

[Guest]

Wouldn't EPYC servers look more attractive with proper thermal and power reporting on Linux?

Come on, AMD!

Yeah, AMD's hardware is in good shape, but their software strategy has been crap for years. Maybe even decades. There's simply no excuse for their bad Linux support, when it's necessary to succeed in the server space. If not for their GPUs , then you'd atleast expect good CPU support, but it's worse than it used to be.

After having been an AMD fanboy for years, I just gave up on them. It's obvious they don't give a shit about Linux support, and seem to be hyper focused on the Windows and console market.

[groeck]

I'm not too familiar with the CPU side documentation, but if it's anything like the GPU side, there are no ready made programming guides just sitting around waiting to be released. ...

The information is available from AMD under NDA, which is why Windows tools like HwINFO support it. The added Linux support was largely possible due to what I am sure are unintentional leaks by AMD, mostly in AMD's Linux kernel graphics code. Unfortunately, they learned; the latest version of their graphics code include files no longer provides temperature sensor addresses for Zen2.