[TRACKING] Battery flipping between charging and discharging / Draws from battery even on AC

Framework Laptop 13 Linux
648

The way modern chips work an average 35W load will be made up of bigger spikes and wait periods between. The ec tells the soc it can pull 176W for short periods of time so it’ll take those opportunities when it sees them, doubt it ever actually manages to get that high and part of the spike will be covered by capacitors but the rest hits the battery.

Probably needs some experimentation but I didn’t see any obvious downsides to just going with 89W, if you want a little bit of dipping I’d probably do like 1.2-1.4x charger (judging by how little 89W caused a 65W charger to dip into battery, granted that was also with the modification to use all of the charger not just 90%) but I’d actually have to find a workload that benefits from high P3t first.

Going too low with p3t reduces max clocks and achievable sustained power but above the 89W (battery only) value I have not seen measurable impact on anything I looked at.

I actually didn’t care enough to keep running my modified ec but this seems pretty fixable if they wanted to. Imo the reasonable thing would be to do the 176W thing on the max performance power profile and just 90% charger (with a min of 89 or something) for balanced and powersave (could probably go even lower there and accept some reduced peak performance).

649

I am seeing the battery state repeatedly flip between fully-charged and discharging on a Framework 13 while external AC power remains continuously online.

Setup:

  • CachyOS

  • Niri + Noctalia

  • Noctalia 5.2.0-1.1

  • UPower 1.91.4

  • Kernel 7.2.8-1-cachyos

  • Framework Laptop 13 (AMD Ryzen AI 300 Series)

  • BIOS 03.05

I first noticed this because Noctalia started randomly playing its power-plug / power-unplug shell sounds while the laptop was sitting plugged in at 100%.

After troubleshooting the audio side, it appears that the sounds are only a symptom of the battery-state changes.

upower --monitor-detail shows BAT1 switching between fully-charged and discharging, for example:

History (rate):
1790714537  0.000  fully-charged
1790714507  0.015  discharging
1790714467  0.000  fully-charged
1790714451  0.015  discharging

The reported discharge rate is extremely small, around 0.015 W.

At the same time I monitored the actual external power state via sysfs:

/sys/class/power_supply/ACAD/online:1
/sys/class/power_supply/ucsi-source-psy-USBC000:004/online:1

I watched both continuously at 0.2-second intervals while the state change / sound occurred, and both remained 1. So the USB-C charger connection itself does not appear to be dropping.

Noctalia’s log also confirms that these state changes result in power events. While I had PipeWire deliberately stopped for troubleshooting, it logged:

[sound] failed to connect stream for sound "power-plug": Host is down
[sound] failed to connect stream for sound "power-unplug": Host is down

With Noctalia shell sounds enabled this produces apparently random plug/unplug sounds; disabling shell sounds hides the symptom.

So in summary:

  • ACAD remains online

  • the active USB-C source remains online

  • battery is at 100%

  • BAT1 nevertheless alternates between fully-charged and discharging

  • desktop software sees those transitions and may treat them as power plug/unplug events

Is this expected behaviour with the Framework 13 charging design, or is this a regression in the kernel, UPower, firmware, or ACPI reporting?

650

If you read the thread before posting (at least partially), you’d see that the stock 65W charger just isn’t enough to power the laptop under peak load. This is not a bug in anything really (except the proposed charger power by Framework maybe). If you are annoyed by this — get yourself a 100W charger, and the issue will disappear. Or just live with it.

651

Probably not entirely but it will massively reduce them.

652

I’m yet to see a single charge flip with a 100w charger, it’s been years. I have a 7840U though, not gen9 Ryzen, but I doubt those CPUs could pull over 100w either.

653

If you mean reported to the os then yes I also have not had any of those in a while, even with much lower wattage chargers but that doesn’t mean the laptop isn’t dipping into battery they just fixed the reporting at least somewhat.

Same

The 7840U can pull way over 100W just for very short periods, there has been a bunch of experimentation in this thread by me and a bunch of other people. Also with cable losses and the framework just using 90% of max current limitation on an 100W psu you’ll only need to pull mayve 85W to dip into battery.