I have the same issue. charge limit feature is broken now. It's not correctly respected.

Are you able to run "ectool version". It is a Linux command, but there might be a Windows equivalent.
Maybe the FW tool can be made to work on Windows. It is a rust program that you might be able to compile on windows.
Alternatively, use a Linux Live USB stick.

framework_tool --versions

Post the output here.

I found a Windows build of ectool here and ran it locally. Output here:

RO version:    lilac-3.0.3-413f018
RW version:    lilac-3.0.3-413f018
Firmware copy: RO
Build info:    lilac-3.0.3-413f018 2025-03-06 05:45:28 marigold2@ip-172-26-3-226
Tool version:  0.0.1-isolate Nov 13 2024 none

FW have not published any of the source code for the "lilac" EC Firmware.

Would you be able to clarify.
Say you set the charge limit to 80% and the current charge is 80%.
You then set the charge limit to 60%. What are you expecting?
FYI, FW expect it to immediately discharge to 60%, even with the AC plugged in.
I did not like that, so part of the reason I wrote my own bug fixes to the EC Firmware for my FW16 AMD 7840HS.

Whether the laptop discharges or relies on AC power is not high-priority on my list of concerns; the issue here is that objectively incorrect behavior happens, which is charging to 100%. It defeats the whole point of a battery charging limit if the limit is only enforced during that session for the EC.

As for how it looks at the BIOS setting level:

• The BIOS does remember what the entered charging limit was; re-entering the BIOS shows the limit as entered.
• The BIOS does not re-inforce the setting even after you re-enter the BIOS settings. I need to enter a different limit in order for the EC to enforce a charging limit, and of course a power-down makes the limit ineffective again.

Whether the laptop discharges or relies on AC power is not high-priority on my list of concerns; the issue here is that objectively incorrect behavior happens, which is charging to 100%. It defeats the whole point of a battery charging limit if the limit is only enforced during that session for the EC.

As for how it looks at the BIOS setting level:

* The BIOS does remember what the entered charging limit was; re-entering the BIOS shows the limit as entered.

* The BIOS does _not_ re-inforce the setting even after you re-enter the BIOS settings. I need to enter a different limit in order for the EC to enforce a charging limit, and of course a power-down makes the limit ineffective again.

I've observed the same issue since I got the laptop, running Fedora 42. To slightly expand, the setting is correctly maintained if rebooting or if the period of time the laptop stays off is "short enough" but any longer shutdown will not enforce the limit.

The messaging from the BIOS to the EC is via HC (Host Commands).
At boot up the BIOS sends an HC to the EC telling it what the charge limit is.
The EC then just has some bugs in its code that cause it to accidentally not take it into account in particular edge cases.
If you wish you can download and install my version of the "ectool". Instructions here:
https://github.com/jcdutton/EmbeddedController/wiki/ectool

To stop it charging or discharging:
sudo ectool chargecontrol idle

To cause it to start discharging down to the BIOS set Battery Limit level.
sudo ectool chargecontrol discharge

To cause it to start charging, if it is below the BIOS set Battery Limit level:
sudo ectool chargecontrol discharge

They are not fixes for the bugs, but at least let you bring the charge back to where you want it.
Due to EC bugs, the above commands have limited success, but might help.

I have written a version of the EC firmware that fixes all the bugs, but it currently has only been tested on FW16 AMD 7840HS and FW13 AMD 7840U/7640U.
I don't have a version that would work on the FW13 AMD AI 300.

Presumably, this is the same issue as found on the 7040 series, BIOS v3.07:

We found a known issue that "If the battery extender mode is disabled, and charge limit is set below 100%, the charge limit will not be enforced causing the battery to charge to 100%."

#52 (comment)

It was fixed in BIOS v3.09. @quinchou77, could you confirm? Will the same fix be brought over to the AI 300 series?

Presumably, this is the same issue as found on the 7040 series, BIOS v3.07:

We found a known issue that "If the battery extender mode is disabled, and charge limit is set below 100%, the charge limit will not be enforced causing the battery to charge to 100%."

#52 (comment)

It was fixed in BIOS v3.09. @quinchou77, could you confirm? Will the same fix be brought over to the AI 300 series?

Yes, it will also be fixed in the next AI 300 Series BIOS. We are planning for each product update. I will update the schedule for AI 300 Series.

Are you able to run "ectool version". It is a Linux command, but there might be a Windows equivalent. Maybe the FW tool can be made to work on Windows. It is a rust program that you might be able to compile on windows. Alternatively, use a Linux Live USB stick.

framework_tool --versions

Post the output here.

FYI, prebuilt binaries of framework_tool are available for windows: https://github.com/FrameworkComputer/framework-system/releases