LEFTEROS / PERSONAL NOTES
Omarchy on the Pixel 8 Pro: it powers off, sleeps, reads NFC, and the 540 mA problem
15 November 2026 / Lefteris Iliadis / SomniusX

My Pixel 8 Pro (husky, Tensor G3) runs Omarchy Mobile from slot B, with GrapheneOS on slot A as the way back. In the last post I had just heard sound. Nine days later, a lot has gone upstream, and the phone has learned to power off, sleep and read NFC. The current it draws with the screen off is still the big open problem, though.
Everything below runs on my phone, on a test kernel. For each point I say what I tested and how many times.
What went upstream
- Bluetooth (#199) and haptics with the torch (#200) went into the kernel on 7 November.
- The husky sound package, with the audio DSP start and the PipeWire files (#453), went in on 9 November. The two kernel changes for sound (#203, #204) are still open: Derek wants them to land together with the core AoC code.
- The NFC clock (#209) went in on 10 November, and Derek wrote: "NFC reading tags on the 8 Pro is a real milestone".
- The clocks, supplies and node for GNSS (#210, #211, #239) went in, all disabled. The GNSS chip hasn't had power yet.
- A Wi-Fi fix (#235), the Display settings page (#628), noticing a resume from sleep straight away (#642, Derek ran it on his 7 Pro too) and brightness with the Power key (#640).
It powers off without starting again
systemctl poweroff restarted the phone instead of turning it off. The cause is a restart timer in the power management chip (S2MPG14) that stayed armed. gitColinX found the lead, and the fix is two lines in the kernel. I tested it twice: once without the cable, where it was still off when I came back about 15 minutes later, and once with the cable, about 7. For now it's only in my own kernel; gitColinX has the same change, and we're waiting on Derek. The results are on #578.
The first sleep
The first sleep results from a Pixel 8 Pro, as Derek put it, were mine: a 58-second and a 10-minute s2idle sleep through systemctl suspend, woken by the Power key. On 15 November, with an alarm from the clock, the phone slept for 270 seconds, once without and once with the setting below. Without a small setting, the systemd watchdogs killed services on wake; with the setting it came out clean. The cable was in, so I don't have a figure for current in sleep yet. I posted the results on #765, a Pixel 9a issue with the same watchdog problem, and the change to a single sleep state is in #806, in review.
One kernel with everything
On 14 November I built a kernel with everything published, the sound changes and the power-off fix, and tested it from scratch. It booted in 27 seconds, sound came up by itself, the speaker and the microphone were clear, the torch and haptics worked, and Bluetooth found 12 devices in a scan (no pairing). With the NFC and GNSS settings it also booted clean, with the NFC controller bound and the GNSS supplies off. I haven't read an NFC card on this kernel yet, or powered GNSS on for the first time. Cards and tags were read over NFC on 7 November, on an older test kernel.
Settings and the screen
The Display page in Omarchy Mobile's settings changes brightness and the refresh rate (60 or 120 Hz). Night light isn't possible on this screen, because its driver doesn't support a colour matrix, and the page says so. Two more pages are in review: keyboard haptics (#629) and Change PIN (#630).


The Display page, merged, and the keyboard haptics page, in review.
The screen looked washed out. A gamma-correcting shader in the Hyprland config fixes it, and I keep it on my phone for good. My first explanation of the cause was wrong, and I corrected it in public on #537.
The 540 mA problem
With the screen off and the cable out, the phone draws about 540 mA, with noise of about ±15 mA. A phone at normal idle is at 30 to 60. I tried stopping the recorder, taking CPUs 1 to 8 offline, dropping the clocks to the floor, and turning Bluetooth and NFC off: none of it brings the current down. Wi-Fi power save gains about 30 mA, but that's still only local.


Left: none of these bring the current down. Right: Wi-Fi power save gains about 30 mA.
What's behind it, I don't know yet. One thing I do know: the kernel has no CPU idle states at all (no cpuidle driver), and no power-domain management. I haven't measured how much each draws. Derek asked explicitly: no more power-domain tests without a plan he has seen. #217 has the details.
Charging stops at 92%
The phone stops charging at 92%. The cause is that the bootloader sets the charge voltage to 4.35 V and Linux never raises it. The plan is a setting for the husky battery at 4.45 V, with a temperature table, then a limit for where it stops. It exists only as local changes; nothing has been written to the charger. It's on #761.
What doesn't work yet
- The modem, waiting on an answer from Derek before any step.
- GNSS, which hasn't had power yet, the sensors, the fingerprint reader and the cameras.
- Hardware video decoding.
- Idle power, about 540 mA, and charging that stops at 92%.
- Once, after a cold start, USB didn't work until it was set up again (#757).
Thanks
To Derek Ross (@derekross) for Omarchy Mobile and for the reviews that keep me careful, to gitColinX for the power-off lead, and to Brett-Michael Green (@greenbrettmichael) for the kernel work. The Omarchy Mobile page keeps the current state.