You are not logged in.
Should I activate 12-15 by "that echo 1>......." Command?
Yes, but I'd do that incrementally (one CPU every or every other hour or so)
Offline
See this:
(All are tested in usb live iso Arch )
1. I used these parmeters: processor.max_cstate=1 idle=nomwait pci=nomsi module_blacklist=thunderbolt
Couldn't even boot and got kernelpanic : Logs
2. Tried maxcpus=10 and enable cpu one by one [ I have 16 cpu 0-15]
I enabled 12-15 one by one (one minute gap) and no kernel panic it was fully stable and was idle for 30 minutes
BUT : the moment I enable cpu 11, I got kernel panic after 10 seconds
Here are the Logs : Logs
Some Good Information:
I ran : dmesg | grep -i taint IN ARCH ISO
Drive image 1
I ran : dmesg | grep -i error IN ARCH ISO
Drive image 2
I ran : dmesg | grep -i error in Ubuntu
Drive image 3
To check if My physical core/thread is faulty, I ran 2 test
1. taskset -c 11 sysbench cpu run
Output :
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)
Running the test with following options:
Number of threads: 1
Initializing random number generator from current time
Prime numbers limit: 10000
Initializing worker threads...
Threads started!
CPU speed:
events per second: 5359.37
General statistics:
total time: 10.0001s
total number of events: 53597
Latency (ms):
min: 0.18
avg: 0.19
max: 0.64
95th percentile: 0.19
sum: 9995.56
Threads fairness:
events (avg/stddev): 53597.0000/0.00
execution time (avg/stddev): 9.9956/0.002. full stress test on cpu 11, ran : taskset -c 11 stress-ng --cpu 1 --taskset 11 --timeout 60s--metrics-breif
Output:
stress-ng: info: [4605] setting to a 120363906789 years, 90 days, 15 hours, 8 mins, 48 secs run per stressor
stress-ng: info: [4605] dispatching hogs: 1 cpu
^Cstress-ng: info: [4605] skipped: 0
stress-ng: info: [4605] passed: 1: cpu (1)
stress-ng: info: [4605] failed: 0
stress-ng: info: [4605] metrics untrustworthy: 0
stress-ng: info: [4605] successful run completed in 18 mins, 56.42 secsI think something is creating problem with cpu11 could it be "nvidia-wc-backlight " ?
Offline
BUT : the moment I enable cpu 11, I got kernel panic after 10 seconds
There's your (most) troublesome core.
To check if My physical core/thread is faulty, I ran 2 test
The problem is w/ idling/waking the core, stressing it won't expose anything.
If you cannot control PBO (the curve optimizer) at all you could try https://archlinux.org/packages/extra/x86_64/corectrl/
Offline
What should I do with this "corectrl" like just install pkg and check? and after that what??
I think something is creating problem with cpu11 could it be "nvidia-wc-backlight " ?
Should I blacklist this, will this help??
Also this may help, ran this in ubuntu to check how it handles cpu11 (claude AI help)
radon@Horizon ~ % cat /sys/devices/system/cpu/cpu11/power/*
disabled
cat: /sys/devices/system/cpu/cpu11/power/autosuspend_delay_ms: Input/output error
auto
0
0
0
disabled
unsupported
0
0Last edited by radon (2025-11-26 16:44:32)
Offline
just install pkg and check? and after that what??
Yes and make sure cpu11 doesn't idle/clock down.
Btw, try
grep . /sys/devices/system/cpu/cpu11/power/*Offline
I couldn't install corectrl in usb live environment,
This is the output of the command:
With and without cpu11 is provided, I ran 2 times
Offline
I couldn't install corectrl in usb live environment,
No, why would you?
The plan is to get core 11 in the installed system working.
The commands was juts to illustrate how to get this information better™ (because it logs the filenames w/ the contents)
Offline
Okay so I will install it fully, which kernel should I use, normal one or lts, or maybe zen
Offline
The main kernel or lts, rather not zen (this is not! a "ryzen optimized" kernel, the "zen" part is coincidental)
Offline
So yeah I installed arch linux fully.
Installed corectrl, didn't do anything just installed it.
In my first boot I booted with nosmt and enabled all core manually. (Corectrl is still running)
Status : 10 min idle no kernel panics, Sucess
2nd boot: booted normally, started corectrl
Status : 15 min into boot no kernel panic, Sucess
So like what happened now is it fixed, should I always start corectrl on boot?
If it's fixed then what corectrl do. Like I didn't configured it just installed. (Power profile is automatic - default)
Edit : booted but didn't open corectrl, kernal panic in 5 minutes
2. When I upgrade -Syu , the system freezes after downloading and installing,
When I force shut down and ran command again, it says up to date.
3. When I download some nvidia drivers it freezes (previous experience)
4. When I change global theme plasma shell crashes T_T (previous issues but still there)
Last edited by radon (Yesterday 20:11:25)
Offline
I don't expect just starting corectl to do much . the plan is to ramp up the minimum clock/voltage for core11 and It would seem that you've to (auto)start it to apply the settings, https://gitlab.com/corectrl/corectrl/-/wikis/Setup
Are you sure that there're no PBO related settings in your UEFI?
Worst case solution, keep core11 disabled…
Offline
There is no PBO related settings in my UEFI.
I don't know what corectrl do, but yeah it does fixes the main issue.
Is there something else should I do?
Last edited by radon (Today 07:28:16)
Offline