You are not logged in.
if I add intel_idle.max_cstate=2 then the sleep consumed 8% per 8 hours,
if i use intel_idle.max_cstate=3 then the bag appears.
Offline
So basically the limited c-state prevents the CPU from actually shutting down during s2idle.
Going back to #14, what is device:5c ?
Offline
Laptop: Lenovo ThinkPad X9-15 Gen 1 – same problem; this bug can only be fixed by Lenovo or Intel
Offline
what is device:5c ?
$cat /sys/bus/acpi/devices/device:5c/description -Pp
ICL TBT DMA0 controller
$ls -la /sys/bus/acpi/devices/device:5c/physical_node -d
Permissions Size User Date Modified Name
lrwxrwxrwx - root 24 Jul 12:25 /sys/bus/acpi/devices/device:5c/physical_node -> ../../../../pci0000:00/0000:00:0d.2
$lspci -nnks 00:0d.2
00:0d.2 USB controller [0c03]: Intel Corporation Core Ultra 200V Series Processors Thunderbolt DMA0 [8086:a833] (rev 10)
Subsystem: Lenovo Device [17aa:233b]
Kernel driver in use: thunderbolt
Kernel modules: thunderboltLast edited by Semyon0101 (2026-07-24 09:29:08)
Offline
I have disabled some variable in bios and have found, that
1. Disabling that:
- PCLe tunneling
- Cpu power management
- Intel speedstep technology
- always on USB
- Wake on lan from dock
- i/o port acess (all. including camera, micro, audio, etc)
- Virtualization: enchanted biometric security
- Security chip: pluton processor control
Gave the result as disabling state3 and state2, battery leaked fast (about 1.5-2% per hour), and there was no lags.
2. Disabling the same variables without {Cpu power management, Intel speedstep technology}
Gave the result as an usual behaviour, battery leaked not fast (about 0.3-0.5% per hour) and there was lags.
Last edited by Semyon0101 (2026-07-24 09:41:11)
Offline
I would have asked about "always on USB" - can you also toggle thunderbolt?
Offline
In section Thunderbolt 4 there is one toggle: PCLe tunneling (as I write before I disabled it)
In section I/o port acess there is thunderbolt (same)
Offline
This is still not solved.
I Have reported on Lenovo ((https://forums.lenovo.com/t5/Fedora/sys … p/10050661) and on the Xe driver gitlab (https://gitlab.freedesktop.org/drm/xe/k … items/9712). Should it be reported anywhere else? It feels more like an upstream driver bug (Intel/Kernel) as it does not only affect lenovo's machines (e.g. https://forum.level1techs.com/t/lunar-l … /246487/3).
Last edited by avatar1024 (2026-11-01 08:52:11)
Offline
How does https://forum.level1techs.com/t/lunar-l … 3/246487/3 in any way shape or form address the CPU frequencies in the immediate aftermath of a s2idle?
Since by https://bbs.archlinux.org/viewtopic.php … 8#p2303968 even pushing the CPU idle states after the wakeup doesn't seem to help and for all we know this is lenovo-specific the fix will have to come from a lenovo firmware update.
The lenovo link claims the problem would be xe/graphics rather than the CPU.
Idk whether that's true but
lsmod | grep xeand does either of "xe.enable_dc=0" "xe.enable_sagv=false" "xe.enable_dmc_wl=0" have any impact on the situation?
https://wiki.archlinux.org/title/Kernel_parameters
Maybe
Offline
Yes, it is. I found some interesting patch description in fwupdmgr when update bios. It was said, that maybe it was solved problem of stack of cpu frequently. But after update I found that it needs laptop to lag 2 hours, not 5. So you can test too.
Temporary solution that I found is use rctwake to wake up laptop every hour, so lags don't appear and battery consumption isn't big as disable sleep as general (about 1% per 3 hours)
Offline