HfiLpmEnable set to 1 prevents from entering the low power mode on i7-1260P #40
Comments
|
with HFI monitor activated lpmd only reacts when a HFI netlink message is raised from the kernel. and what is your output of "rdmsr 0x17d1"? |
|
BTW, I guess you would see similar symptom with 0.0.3. |
|
One thing to mention is that the HFI monitor and Util monitor are exclusive and intel_lpmd uses one of them as the input for decision making. So once you enabled HFI monitor, util monitor is not activated at all. |
Executing this command gives me
This is my previous 0.0.3 config: #24 (comment)
I've set EDIT: |
|
I have the same issue on an i5-12500h laptop. Might be ADL related? |
|
I see. And you're right that lpmd still activates util monitor when HFI monitor is enabled in v0.0.3. commit da9ffd3 |
|
Close this issue. |
|
I don't think this issue should be closed.
-> Relying on HFI with intel_lpmd is not working (never enters low power mode), similar to what @stereomato observed. Not relying on HFI, but utilization, is working (enters low power mode). I'm happy assisting with logs or changing my perspective if this is not supposed to work on my system. |
|
Yeah same here, zhang says:
is there no way to make them be honored? seems like something is broken that could be fixed, or if this is really not fixable there should be a fallback + warning to the user otherwise it's just confusing |
I agree this is an issue, but IMO, it is very likely not an intel-lpmd issue.
I will reopen this, and try to see if there is anything we can do. At the same time, can you please try a different laptop and see if that works? |
Here "honored" means, we do A following HFI, but then we do B following utilization, the system state is not what the HFI hints told us so we are not strictly following HFI hints.
I can add a message telling the users that
Does this make the things clearer? |
Yeah, it seems fine. Maybe also a warning that HFI might not work well? or at all? |
Unfortunately, I'm unable to try another Laptop with a recent Intel chipset, as don't get access to such a system. |
|
re-open per @rmeissn suggestion. @rmeissn , what kernel version you're using? |
|
I'm currently on 6.9.6-200.fc40.x86_64 (default fedora kernel while updating regularly). I've run those tests as follows:
These are the results for AC This is the result for BAT: |
|
from the log, we don't receive ANY thermal interrupts when on battery, this means there is no HFI hint updates from firmware. And this explains why you cannot enter low power mode with HFI monitor. |
|
I've retested while on AC and started lpmd in debug mode (logs below). In the middle, I've unplugged the docking station, changing to battery mode. In both modes, I've caused some load and times of no load, where lpmd should "deactivate" most cores. |
|
Might this somehow related to selinux or secure boot? For selinux, I can't spot any problems in the audit.log file. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I've migrated from intel_lpmd 0.0.3 to 0.0.4 and migrated most of my configuration. Unfortunately, intel_lpmd won't enter low power mode when HfiLpmEnable is set to 1. When set to 0, it intel_lpmd enters low power modes correctly.
Log for HfiLpmEnable 1 :
-> after issuing
sudo intel_lpmd_control AUTO, nothing happens in the logs.Log for HfiLpmEnable 0 :
-> after issuing
sudo intel_lpmd_control AUTO, the 4th line is logged.XML configuration (HfiLpmEnable 1):
System:
Framework 13, Intel i7-1260P, Fedora 40.
The text was updated successfully, but these errors were encountered: