🧠 Investigating Nothing OS CPU Idle Behavior — Wake Locks & Background Services

JochieSantiago · 2025-11-04T00:09:49+00:00

Hey everyone, I’ve been testing my Nothing Phone (Nothing OS 3.5) with ChatGPT (GPT-5) to understand why the CPU doesn’t seem to enter deep idle — even when the screen is off and nothing’s running in the foreground. We went through a full diagnostic session using ADB to identify which services or apps might be holding wake locks or background jobs that keep the CPU partially active ⚙️ Background The issue started with unusually high idle CPU usage (20–29%) and noticeable overnight battery drain (around 6–8%). After checking dumpsys power, several partial wake locks were visible — mainly from Smart Center (com.nothing.smartcenter) and occasionally other apps using Firebase (like WhatsApp, Messaging, or TubeMate). Smart Center’s BLE notification service in particular (NothingNotificationService) seems to stay active even when the phone should be sleeping. 🧩 Steps & Methods 1\. Checking active listeners adb shell settings get secure enabled_notification_listeners Found Smart Center’s BLE notification listener constantly enabled. 2\. Restricting background activity We temporarily limited Smart Center’s permissions to see if it affects CPU sleep states: adb shell am set-standby-bucket com.nothing.smartcenter restricted adb shell cmd appops set com.nothing.smartcenter RUN_ANY_IN_BACKGROUND ignore adb shell cmd appops set com.nothing.smartcenter WAKE_LOCK ignore adb shell cmd appops set com.nothing.smartcenter POST_NOTIFICATION ignore adb shell cmd jobscheduler cancel com.nothing.smartcenter 3\. Forcing deep idle manually adb shell cmd deviceidle force-idle adb shell cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq CPU frequency started dropping from \~1.9GHz down to \~441MHz, showing that deep idle can be reached manually. 4\. Testing third-party apps Some third-party apps (like TubeMate) were also seen holding KeepAlive jobs: adb shell cmd appops set devian.tubemate.v3 RUN_IN_BACKGROUND ignore adb shell cmd appops set devian.tubemate.v3 WAKE_LOCK ignore adb shell am force-stop devian.tubemate.v3 📊 Current Observations With restrictions in place, CPU frequency does fall lower during idle, sometimes reaching 441MHz. dumpsys power shows significantly fewer active wake locks. Still under testing to confirm long-term battery impact and if Nothing OS automatically re-enables those background services over time. 🔄 Reverting the Changes To restore everything: adb shell cmd appops set com.nothing.smartcenter RUN_ANY_IN_BACKGROUND allow adb shell cmd appops set com.nothing.smartcenter WAKE_LOCK allow adb shell am set-standby-bucket com.nothing.smartcenter active There’s also a revert_all_changes.bat script prepared to reset all settings at once. 💬 Closing Notes This doesn’t appear to be an app bug — it’s likely a system behavior tied to Nothing’s background daemons, Smart Center, and Firebase sync routines. We’re still testing whether these restrictions improve idle drain over time. Big thanks to ChatGPT (GPT-5) for helping me trace and document each step clearly. Please, feel free to drop any suggestions or ideas. Keep in mind that my device was not directly bought with Nothing, so it does not have warranty. At this point, I'm just trying different stuff out to see if anything helps with the battery life problems and the apparent causes of it. (At least for me)

JochieSantiago

So i ran some tests today with ADB and CHATGPT guiding me through each step.

A link with the full report will be attached if you guys wanna take a look. @CarlPei @Rob ⚡️ can someone form the Software or related area take a look?

# Nothing Phone 3 — Deep Sleep Diagnostic Report ## Background:

Why this test was performed

The diagnostic session was initiated to determine why the Nothing Phone 3 appeared **not to be entering deep sleep (Doze)** mode. The user noticed possible idle battery drain and wanted to verify if background processes or wakelocks were preventing the system from fully suspending during screen-off periods.

The goal: identify if the device was staying awake, which services were responsible, and confirm proper power management behavior.

--## Phase 1 — ADB environment setup

**Objective:** connect the device and verify Doze state.

**Actions:**

1. Connected device via ADB.

2. Inspected Doze states using `adb shell dumpsys deviceidle`.

3. Found that the phone was in **OVERRIDE mode**, meaning Doze was forced for testing and not behaving naturally.

4. Restored normal operation with: ```bash adb shell cmd deviceidle unforce adb shell cmd deviceidle enable all adb shell dumpsys battery unplug ```

5. Reset battery stats.

**Result:** Doze returned to normal (`mState=INACTIVE`, `mLightState=INACTIVE`).--

## Phase 2 — Deep Sleep verification

**Actions:**

1. Device left idle for ~40 minutes, screen off and unplugged.

2. Checked device state: ```bash adb shell dumpsys deviceidle ``` Output confirmed `mState=IDLE` and `mLightState=IDLE`.

3. Verified no wakelocks held: ``` Wake Locks: size=1 PowerManagerService.WakeLocks: ref count=0 [] ```

**Result:**

Device successfully entered deep sleep with CPU fully suspended.

--## Phase 3 — Log extraction

**Commands executed:** ```bash adb bugreport bugreport_nothingOS.zip

adb shell dumpsys batterystats --charged > batterystats_charged.txt

adb shell dumpsys batterystats --history > batterystats_history.txt

adb shell dumpsys alarm > alarms.txt

adb shell dumpsys deviceidle > deviceidle.txt

adb shell dumpsys power > power.txt

adb shell cat /d/wakeup_sources > wakeup_sources.txt ```

**Files collected:** `power.txt`, `alarms.txt`, `batterystats_history.txt`, `batterystats_charged.txt`, `dumpstate_log.txt`, and the full `bugreport_nothingOS.zip`.

--## Phase 4 — Log analysis findings

### Battery & Doze- Deep sleep active **>80% of total time**.

- Idle drain extremely low (~1%/h or less).

- Proper Doze cycles observed: `device_idle=full` ↔ `device_idle=off`.

### Wakelocks-

None persistent or abnormal.

- System-only wakelocks observed: `PhoneWindowManager.mPowerKeyWakeLock`, `SyncLoopWakeLock` — both released instantly.

### Alarms-

Average **6 wakeups/hour**, all expected.- Key sources:- `DeviceIdleController.deep/light` (Doze maintenance)-

`GCM.HEARTBEAT_ALARM` (Google Play Services)-

`AOD.dismiss.fingerprint` (Always-On Display)

- No third-party apps abusing alarms.

### Radio / Network-

Brief modem wakelocks (`RILJ_ACK_WL`, `telephony-radio`) confirm LTE/IMS heartbeats.- No repeated reconnects or radio instability detected.

### Temperature- Maximum 34°C — perfectly normal.- No thermal throttling or overheating.

### Errors / Crashes- No ANR, FATAL, or crash loops detected.- `dumpstate_log.txt` confirms clean bugreport completion.-

-## Phase 5 — Final conclusion-

Device power management is **fully functional and optimized**.- Doze and deep sleep operate as designed.- No rogue apps or services preventing idle suspension.- Battery consumption in idle is **negligible**.- System and kernel wakelocks behave normally.

**Root cause:** the initial “OVERRIDE” mode kept Doze in a forced testing state, making it appear as though deep sleep was disabled. Once reset, the phone resumed normal sleep cycles.

--## Recommendations

1. Keep Always-On Display disabled when testing deep sleep efficiency.

2. Avoid widgets that refresh too often (clocks, weather, etc.).

3. Use AccuBattery to monitor real deep sleep percentage.

4. Retain this report as a **baseline reference** for future firmware updates.--**End of report — NothingOS Deep Sleep Diagnostic (Nov 5, 2025)**

Link for the full Logs and text results:

Nothing Phone 3 Bug & Tests Results

CMF1_hacker

Interesting.

My NP3 is entering deep sleep well, but my issue is the constant connection to the CMFW3P as I'm doing around 1h of cycling per day with the screen off, which uses BT and constant data transfer. Also, it requires a constant BT connection. The uptick in my watch usage for tracking has affected my maximal SoT, which makes sense. It's just much worse than a Garmin that have enough memory to store all of this data and only use BT infrequently.

I just cycled into worse and noticed a very large uptick in % battery/h used with the screen off.

I guess there are the compromises for data tracking.

JochieSantiago

CMF1_hacker and the cmf watch 3 pro is supposed to handle better Bluetooth connection. Did that happen after the recent updates?

Also, my phone is now entering deep sleep way better than before.

This is from last night.

Also, I'm running a test with adb and making a videocall on WhatsApp to see if it is actually draining the battery:

And it is. Abnormally as a matter of fact

GalibHossain_EuIYRZpoTs

I have been talking about the battery issue of my Nothing 3a since the week I bought the phone, because the idle drain seems too much for me—almost 8-10%. I was afraid that something was wrong with my new phone, because my 6-year-old phone almost had 0 to 2% battery drain overnight. I contacted Nothing Support several times, but their reply was: "We will look into the issue." After that, I installed 'Enforce Doze' with ADB; still nothing! Some people say that it is the Snapdragon 7s Gen 3 chipset that has a bad network modem. However, the NP3 also has the same issue. Even if I turn on airplane mode, my phone still loses over 5 to 6% overnight. Now it has become a habit for me to charge my phone twice a day. (You can find all the screenshots on my profile)

JochieSantiago

GalibHossain_EuIYRZpoTs I think that a future fix is coming, but for now, using adb commands to disable some stuff is the way to go.

And the modem chip is not bad. I ran a few tests and the problem came from the snapdragon ship

CMF1_hacker

JochieSantiago I've made it through another day and it's much worse than when I was getting 10-11h SoT per charge.

I think my phone has become your phone.

Check this out and the Splash action for my children and the System

What would you if you were me with ADB?

I removed the Playground widgets, sadly and let's see what happens and the launcher is always on.

Anonymous1004

CMF1_hacker What would you if you were me with ADB?

Uninstall all google apps that have tracker-free alternatives (Revanced and/or MicroG). Removing the telemetry makes a lot of difference. Unfortunately Gemini won't work properly without the google app

JochieSantiago That's crazy, I don't think I could drain that in a single day

Stats for removing my phone from charge at night to turning it off during school. It is in deep sleep but not for long enough. My alarm was at 6:30 but it exits sleep at 5.

10% per hour when just listening to music on vlc can't be good…

JochieSantiago

I think the link to the OneDrive folder where I saved all the tests I ran is available in the post.

But I would run a adb test to see what is draining the battery.

Chatgpt have me the commands, and when I copied and pasted a code in cmd, I did the same if it returned any results into chatgpt.

You have to download the SDKs platform files from Google and run the cmd from that folder.

I also deleted all third party widgets 🥲 and disabled A core and other stuff that someone posted here (I believe).

Also the tests will give you a result where the glyph matrix is also draining battery but it's minimal, so you can ignore it.

I don't know what code I used, but my WhatsApp drainage is now low, which is weird but also I got the Xiaomi 15T pro today and I've been using it for a while, so maybe that played a part

Anonymous993

I noticed on my baby monitor that my 3a proximity sensor is active ALL night, even aeroplane mode and locked.

Kind of odd - would have thought it would eventually turn off without movement etc? Wondered if that is a possible source of battery drain/unnecessary CPU usage?

JochieSantiago

Anonymous993 it possibly can be.

Try running a live adb test for 10 minutes with airplane mode on and screen off. And don't touch the phone.

If that's wakelock comes up in the test even without you moving the phone, then there's one cause.

CMF1_hacker

I guess I use my flashlight a lot as it's darker now.

Because of this:

JochieSantiago

CMF1_hacker damn, that sucks 😂 hope you could fix it

CMF1_hacker

JochieSantiago after deleting one or two apps that were battery hogs and likely don't work with A16, it looks about standard again, around 10h SoT.

Anonymous993

CMF1_hacker Maybe you use it in the background - but Chrome seems to have way too much background usage here.

I always toggle off background usage, and in Settings > Site Settings there is a "Background Sync" you can disable which I think also triggers background usage.

Slightly annoying all of the Google apps do not behave well as they are system apps and seem to do whatever the fancy at times

CMF1_hacker

Anonymous993 yeah, I always have it sync so the machines at the two house and two offices are always synced up with my phones. I assume this is Google checking the sync. I don't want to mess with it. I also have 2 TiB Google One that the machines access. I'm OK with Google having free reign is my data is present when/where I need it.

JochieSantiago

Haven't been using my NP3 that much since I got the Xiaomi. Yesterday I got 13 hours of SoT with it and still had 13% battery. (Battery saver turns on at 10%)

JochieSantiago

Anonymous1004 I used adb to disable some stuff. The commands are in the post and in the folder linked in it.

Also. This is from my 15T Pro. Number that my NP3 will never reach 😂

CMF1_hacker

JochieSantiago got it fixed now on my NP3

JochieSantiago

CMF1_hacker what settings did you change?

« Previous Page Next Page »