AMD Linux CPU Temperature Driver Sees Latest Patches For Zen 4 & Likely Mendocino

[phoronix]

Phoronix: AMD Linux CPU Temperature Driver Sees Latest Patches For Zen 4 & Likely Mendocino

One of my personal gripes with AMD's Zen CPU support on Linux has been the lack of timely support for CPU temperature monitoring with their "k10temp" driver. Even though usually just new IDs are often needed and sometimes needing to adjust offsets or other minor changes, it has traditionally been done post-launch and sometimes left up to patches from the open-source community. Thankfully that has been changing and with Zen 4 it looks like that support will be ready for launch-day with the mainline Linux kernel...

https://www.phoronix.com/scan.php?page=news_item&px=AMD-k10temp-Zen-4-v5

[birdie]

When I wake up tomorrow I'll post a screenshot of HWiNFO64 which proves that AMD continues to not give a flying f about Linux where it really matters, for my 5800X I get this:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
Tctl:         +31.9°C  
Tccd1:        +30.8°C

This is just BS, not "support". Running Linux 5.18.3 in case people wanna say something about my "old" kernel. This temperature patch for Zen 4 CPUs changes nothing. It's the absolute minimum just so the community doesn't thrash them too much.

[tildearrow]

When I wake up tomorrow I'll post a screenshot of HWiNFO64 which proves that AMD continues to not give a flying f about Linux where it really matters, for my 5800X I get this:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
Tctl: +31.9°C
Tccd1: +30.8°C

This is just BS, not "support". Running Linux 5.18.3 in case people wanna say something about my "old" kernel. This temperature patch for Zen 4 CPUs changes nothing. It's the absolute minimum just so the community doesn't thrash them too much.

Man, that's terrible... Just 2 sensors?

I wonder whether the sensor read addresses/IDs are UUID'ized, scrambled or randomized.
If not then I hope somebody fuzzes the sensor driver and peeks at random addresses hoping to reveal more of the sensors...

Also, did you check sysfs?

[Paradigm Shifter]

This is just BS, not "support". Running Linux 5.18.3 in case people wanna say something about my "old" kernel. This temperature patch for Zen 4 CPUs changes nothing. It's the absolute minimum just so the community doesn't thrash them too much.

Just how much detail do you want? Because per-core temperatures get a bit frustrating with Intel when on a 24-core dual-socket system the reported temperatures for cores go: Core 0, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 16, 17, 18, 19, 20, 21, 25, 26, 27, 28, 29. Wait, what? What happened to 7, 14, 15, 22, 23, and 24? On a different board (with the same CPUs) it again skips some numbers, but not the same ones. The skips seem board specific, though, so maybe it's a Supermicro quirk.

[Danny3]

When I wake up tomorrow I'll post a screenshot of HWiNFO64 which proves that AMD continues to not give a flying f about Linux where it really matters, for my 5800X I get this:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
Tctl: +31.9°C
Tccd1: +30.8°C

This is just BS, not "support". Running Linux 5.18.3 in case people wanna say something about my "old" kernel. This temperature patch for Zen 4 CPUs changes nothing. It's the absolute minimum just so the community doesn't thrash them too much.

Indeed, that's awful!

Clearly AMD doesn't give a fuck about Linux users.

I wonder if on the server side things are better.

[oiaohm]

When I wake up tomorrow I'll post a screenshot of HWiNFO64 which proves that AMD continues to not give a flying f about Linux where it really matters, for my 5800X I get this:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
Tctl: +31.9°C
Tccd1: +30.8°C

This is just BS, not "support". Running Linux 5.18.3 in case people wanna say something about my "old" kernel. This temperature patch for Zen 4 CPUs changes nothing. It's the absolute minimum just so the community doesn't thrash them too much.

Indeed, that's awful!

Clearly AMD doesn't give a fuck about Linux users.

I wonder if on the server side things are better.

The reality here is horrible. Birdie statement is not truthful if he was using only AMD drivers under Windows he would also only be able to see two sensors with that CPU. This is case that AMD is horrible to everyone equally. HWINFO64 is using a third party reversed engineered driver.

Also the more turns out only a limit number of sensors in lower end cpus was in fact calibrated at the factory amd documentation admits to this. So great you get more information using HWINFO64 under windows yet those temperature differences you see between cores could be nothing more than that sensors were not calibrated so now have garbage in garbage out.

Yes higher end server CPU do report more sensors and more sensors were in fact calibrated at the factory.

This is kind of a middle finger to everyone and birdie is being anti-linux to wake up its a generic all around middle finger to anyone who want to use cheaper AMD chips.

[numacross]

The reality here is horrible. Birdie statement is not truthful if he was using only AMD drivers under Windows he would also only be able to see two sensors with that CPU. This is case that AMD is horrible to everyone equally. HWINFO64 is using a third party reversed engineered driver.

Also the more turns out only a limit number of sensors in lower end cpus was in fact calibrated at the factory amd documentation admits to this. So great you get more information using HWINFO64 under windows yet those temperature differences you see between cores could be nothing more than that sensors were not calibrated so now have garbage in garbage out.

Yes higher end server CPU do report more sensors and more sensors were in fact calibrated at the factory.

This is kind of a middle finger to everyone and birdie is being anti-linux to wake up its a generic all around middle finger to anyone who want to use cheaper AMD chips.

AMD's Ryzen Master also shows more information under Windows and it is the official AMD tool.
You can get some of it under Linux by utilizing the out-of-tree module Zenpower. However it is still not even close to what HWiNFO can display.

[syrjala]

Well that was confusing, seeing as I remember Mendocino being the Celeron 300A/etc.

[tildearrow]

Just how much detail do you want? Because per-core temperatures get a bit frustrating with Intel when on a 24-core dual-socket system the reported temperatures for cores go: Core 0, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 16, 17, 18, 19, 20, 21, 25, 26, 27, 28, 29. Wait, what? What happened to 7, 14, 15, 22, 23, and 24? On a different board (with the same CPUs) it again skips some numbers, but not the same ones. The skips seem board specific, though, so maybe it's a Supermicro quirk.

Having the option of detailed information is always nice.

[kbios]

AMD's Ryzen Master also shows more information under Windows and it is the official AMD tool.
You can get some of it under Linux by utilizing the out-of-tree module Zenpower. However it is still not even close to what HWiNFO can display.

To get the full array of information (same as HWiNFO) you can use the ryzen_smu module