You are not logged in.
Hi,
I need someone who points me in the right direction:
I have a workstation laptop with a Ryzen 9 3950X, which has 16 physical cores with 2 Threads each, aka 32 virtual cores.
I noticed however that the cpufreq virtual files in sysfs at /sys/devices/system/cpu/cpu*/ are only created for virtual cores cpu0-cpu23. cpu24-cpu31 are missing the cpufreq file (and the driver and firmare_node simlink).
There is also no policy folder for these cores in /sys/devices/system/cpu/cpufreq/.
Am I missing a setting for high cpu count somewhere in the system?
I also tried Ubuntu but it has the same problem.
I also tried different kernel versions: 5.6, 5.8, and 5.10 all with the same problem.
Offline
what is the output of
$ journalctl -b -g smpboot
for reference here's the output from my threadripper 19020x
$ journalctl -b -g smpboot
-- Logs begin at Wed 2019-09-25 17:40:45 CEST, end at Wed 2020-11-18 14:23:44 CET. --
nov 18 11:04:11 silverbolt kernel: smpboot: Allowing 128 CPUs, 104 hotplug CPUs
nov 18 11:04:11 silverbolt kernel: smpboot: CPU0: AMD Ryzen Threadripper 1920X 12-Core Processor (family: 0x17, model: 0x1, stepping: 0x1)
nov 18 11:04:11 silverbolt kernel: smpboot: Max logical packages: 6
nov 18 11:04:11 silverbolt kernel: smpboot: Total of 24 processors activated (167729.30 BogoMIPS)
$
Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.
(A works at time B) && (time C > time B ) ≠ (A works at time C)
Offline
I'm currently still on my ubuntu test install, but the output here is
-- Logs begin at Wed 2020-11-18 12:49:53 CET, end at Wed 2020-11-18 14:39:24 CET. --
Nov 18 13:03:40 test-TUXEDO-Book-XA1510 kernel: smpboot: Allowing 32 CPUs, 0 hotplug CPUs
Nov 18 13:03:40 test-TUXEDO-Book-XA1510 kernel: smpboot: CPU0: AMD Ryzen 9 3950X 16-Core Processor (family: 0x17, model: 0x71, stepping: 0x0)
Nov 18 13:03:40 test-TUXEDO-Book-XA1510 kernel: smpboot: Max logical packages: 1
Nov 18 13:03:40 test-TUXEDO-Book-XA1510 kernel: smpboot: Total of 32 processors activated (223593.08 BogoMIPS)
I'm creating an arch live usb stick right now to go back to arch and will check if there is any difference.
Offline
Another thing I tried in the meantime: Compiling cpufreq as a module and unloading and reloading it. Sadly that didn't help.
edit: forgot to write: the journalctl output is the same on the arch live usb
Last edited by Matombo (2020-11-18 14:05:35)
Offline
Ok trying out the CPU on a different mainboard where it worked just fine, gave the deciding hint: It's a broken ACPI table: https://bugzilla.redhat.com/show_bug.cgi?id=1874538#c2
Last edited by Matombo (2020-11-19 12:07:01)
Offline