[Bug 1013563] New: dell 7470 - cpu gets throttled a lot
http://bugzilla.suse.com/show_bug.cgi?id=1013563 Bug ID: 1013563 Summary: dell 7470 - cpu gets throttled a lot Classification: openSUSE Product: openSUSE Tumbleweed Version: Current Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: Kernel Assignee: kernel-maintainers@forge.provo.novell.com Reporter: tchvatal@suse.com QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- With a normal usage, on table, browsing+ssh the machine quite often throtlles: [39814.739914] mei_wdt mei::05b79a6f-4628-4d7f-899d-a91514cb32ab:02: Could not register event ret=-22 [39814.741816] mei_wdt: probe of mei::05b79a6f-4628-4d7f-899d-a91514cb32ab:02 failed with error -22 [39867.021572] CPU3: Core temperature above threshold, cpu clock throttled (total events = 1355) [39867.021573] CPU1: Core temperature above threshold, cpu clock throttled (total events = 1355) [39867.021574] CPU0: Package temperature above threshold, cpu clock throttled (total events = 1355) [39867.021575] CPU2: Package temperature above threshold, cpu clock throttled (total events = 1355) [39867.021576] CPU1: Package temperature above threshold, cpu clock throttled (total events = 1355) [39867.021580] mce_notify_irq: 1 callbacks suppressed [39867.021581] mce: [Hardware Error]: Machine check events logged [39867.021583] CPU3: Package temperature above threshold, cpu clock throttled (total events = 1355) [39867.021585] mce: [Hardware Error]: Machine check events logged [39867.022565] CPU1: Core temperature/speed normal [39867.022566] CPU3: Core temperature/speed normal [39867.022567] CPU0: Package temperature/speed normal [39867.022567] CPU2: Package temperature/speed normal [39867.022568] CPU3: Package temperature/speed normal [39867.022569] CPU1: Package temperature/speed normal -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c1
Takashi Iwai
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c2
Tomáš Chvátal
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c3
--- Comment #3 from Tomáš Chvátal
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c4
--- Comment #4 from Takashi Iwai
Also updated bios to latest (had 1.6 now is 1.10) and tried kernel 4.9rc7.
Still present there.
Care to report to upstream, e.g. bugzilla.kernel.org? Feel free to put me (tiwai@suse.de) in Cc. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1013563
Tomáš Chvátal
http://bugzilla.suse.com/show_bug.cgi?id=1013563
Takashi Iwai
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c5
Tomáš Chvátal
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c6
Takashi Iwai
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c7
Tomáš Chvátal
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c8
Takashi Iwai
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c9
--- Comment #9 from Tomáš Chvátal
We can add supplements to thermald specifying the CPU IDs that require thermald, too, instead. Adding thermald maintainters to Cc.
BTW, does this happen also with SLE12-SP2 kernel?
Yes for the old kernel; I tried leap first... :) Technically if we have way how to identify CPU then I am fine for it being supplements. But we also need to change default presets for systemd to enable thermald by default if installed. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c10
--- Comment #10 from Takashi Iwai
(In reply to Takashi Iwai from comment #8)
We can add supplements to thermald specifying the CPU IDs that require thermald, too, instead. Adding thermald maintainters to Cc.
BTW, does this happen also with SLE12-SP2 kernel?
Yes for the old kernel; I tried leap first... :)
Oh, then it's a problem since I'm not sure whether thermald is included in SLE[SD]12-SP2. Thomas?
Technically if we have way how to identify CPU then I am fine for it being supplements. But we also need to change default presets for systemd to enable thermald by default if installed.
That's true. Let's leave the decision to thermald maintainers. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c11
Tomáš Chvátal
(In reply to Tomáš Chvátal from comment #9)
(In reply to Takashi Iwai from comment #8)
We can add supplements to thermald specifying the CPU IDs that require thermald, too, instead. Adding thermald maintainters to Cc.
BTW, does this happen also with SLE12-SP2 kernel?
Yes for the old kernel; I tried leap first... :)
Oh, then it's a problem since I'm not sure whether thermald is included in SLE[SD]12-SP2. Thomas?
It is not present on the SLES12 atm, only Leap and TW.
Technically if we have way how to identify CPU then I am fine for it being supplements. But we also need to change default presets for systemd to enable thermald by default if installed.
That's true. Let's leave the decision to thermald maintainers.
@thermlad-maints: so what is your take, would you guys add those supplements? -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1013563
Jiri Slaby
http://bugzilla.suse.com/show_bug.cgi?id=1013563
Jeffrey Cheung
http://bugzilla.suse.com/show_bug.cgi?id=1013563
http://bugzilla.suse.com/show_bug.cgi?id=1013563#c12
--- Comment #12 from Jeffrey Cheung
http://bugzilla.suse.com/show_bug.cgi?id=1013563
Stefan Dirsch
https://bugzilla.suse.com/show_bug.cgi?id=1013563
https://bugzilla.suse.com/show_bug.cgi?id=1013563#c15
Brice DEKANY
https://bugzilla.suse.com/show_bug.cgi?id=1013563
https://bugzilla.suse.com/show_bug.cgi?id=1013563#c16
--- Comment #16 from Takashi Iwai
https://bugzilla.suse.com/show_bug.cgi?id=1013563
https://bugzilla.suse.com/show_bug.cgi?id=1013563#c17
Fabian Vogt
https://bugzilla.suse.com/show_bug.cgi?id=1013563
https://bugzilla.suse.com/show_bug.cgi?id=1013563#c18
Takashi Iwai
I vaguely remember some report(s) about thermald actually causing the device to overheat and do a hard shutdown. That's not too unlikely, considering its main purpose is to avoid throttling the CPU harder than necessary when the thermal limit is approached/reached.
The thermal limit shouldn't be reached under normal circumstances (like "browsing+ssh" in the original comment), and so thermald shouldn't be absolutely necessary with a properly designed cooling solution. Unfortunately many laptops do not have one, so thermald has a noticable impact on performance on those. However, that comes at a cost: The device will run hotter.
Yes. It's good to give users a choice :)
Any opinions about this?
Supplements: (modalias(cpu:type%3Ax86*ven0000*) if kernel)
I guess this would work as a first shot, that's similar to ucode-intel.
While there are more specific modaliases which could be used here (like some intel specific ACPI platform devices), those may change in the future and then the supplements would silently not work anymore.
Agreed.
Ideally it only matches physical hardware, but I don't think we have any capability for that (yet).
And, thermald won't be enabled / activated by the package installation alone, right? -- You are receiving this mail because: You are on the CC list for the bug.
https://bugzilla.suse.com/show_bug.cgi?id=1013563
https://bugzilla.suse.com/show_bug.cgi?id=1013563#c19
--- Comment #19 from Thomas Renninger
https://bugzilla.suse.com/show_bug.cgi?id=1013563
https://bugzilla.suse.com/show_bug.cgi?id=1013563#c20
--- Comment #20 from Takashi Iwai
Sorry to highjack this one with another issue: I have this bug lingering around for quite some time.
ucode-amd is recommended on Intel CPU systems https://bugzilla.suse.com/show_bug.cgi?id=1158704
From Kernel:HEAD kernel-firmware: %package -n ucode-amd ... # new style (after 3.12 kernel somewhen) Supplements: modalias(cpu:type%%3Ax86*ven0002*) # old style (before 3.16 kernel) Supplements: modalias(x86cpu:vendor%%3A0002%%3Afamily%%3A*%%3Amodel%%3A*%%3Afeature%%3A*)
any ideas or seeing something obvious ;)
The Supplements is fine, but it's pulled just because of Recommends in the patterns-base-enhanced_base (as already suggested in bug 1158704). You may reassign the bug to Component patterns if it's really superfluous. -- You are receiving this mail because: You are on the CC list for the bug.
participants (2)
-
bugzilla_noreply@novell.com
-
bugzilla_noreply@suse.com