Task Manager: CPU column in Processes and Details stops updating after a few minutes (memory, I/O and GPU keep updating)

After Task Manager has been open for a few minutes, the CPU column on the Processes and Details tabs stops updating. The values stay frozen until Task Manager is closed and reopened. The Memory, Disk/I/O and GPU columns on the same tabs keep updating normally. F5 (refresh now) doesn’t restart the CPU column.

Steps to reproduce:

1. Open Task Manager and leave it on the Processes tab. It can stay visible or be minimized; both reproduce it.
2. Wait. It freezes after a random few minutes: about 3 minutes and about 10 minutes in my two timed runs.
3. The CPU column (and CPU in Details) stays frozen, while the other columns keep changing.

Expected: the CPU column keeps updating at the selected update speed. Actual: the CPU column freezes permanently for that Task Manager instance.

What I checked:

• Both modes: it happens both elevated (default) and non-elevated (started with __COMPAT_LAYER=RunAsInvoker).
• Process state during the freeze:
  • only one taskmgr.exe instance is running;
    • the window is responsive, not suspended, not in efficiency mode (no power throttling) and at normal priority.
    • System counters: Windows’ own CPU counters (\Processor Information\% Processor Time and % Processor Utility) keep updating normally the whole time.
    • Thread sampling (per-thread CPU time every 2 s):
      • one thread-pool worker did a small amount of work in almost every 2 s sample from launch.
        • at the exact moment of the freeze it stopped running and stayed idle in the thread pool (waiting on its work queue), so it isn’t blocked.
          • the other worker threads kept running at the same rate as before.
            • So the job that updates per-process CPU seems to stop being scheduled, rather than hanging.
            • Possible trigger: the system has heavy process churn. Automated headless browser tests start and exit about 2–3 processes per second, with roughly 490 processes and 9,800 threads total. A race when a process exits during the CPU measurement may be involved.

Workaround: close and reopen Task Manager; Resource Monitor keeps showing per-process CPU correctly.

System:

• OS: Windows 11 IoT Enterprise LTSC 24H2, build 26100.9974
• Taskmgr.exe: 10.0.26100.9818
• CPU: Intel Core Ultra 9 275HX (24 cores, hybrid P/E cores, no Hyper-Threading)
• RAM: 64 GB After Task Manager has been open for a few minutes, the CPU column on the Processes and Details tabs stops updating. The values stay frozen until Task Manager is closed and reopened. The Memory, Disk/I/O and GPU columns on the same tabs keep updating normally. F5 (refresh now) doesn’t restart the CPU column. Steps to reproduce:
  • Open Task Manager and leave it on the Processes tab. It can stay visible or be minimized; both reproduce it.
  • Wait. It freezes after a random few minutes: about 3 minutes and about 10 minutes in my two timed runs.
  • The CPU column (and CPU in Details) stays frozen, while the other columns keep changing.

Expected: the CPU column keeps updating at the selected update speed.
Actual: the CPU column freezes permanently for that Task Manager instance.

What I checked:

• Both modes: it happens both elevated (default) and non-elevated (started with __COMPAT_LAYER=RunAsInvoker).
• Process state during the freeze:
  • only one taskmgr.exe instance is running;
    • the window is responsive, not suspended, not in efficiency mode (no power throttling) and at normal priority.
    • System counters: Windows’ own CPU counters (\Processor Information\% Processor Time and % Processor Utility) keep updating normally the whole time.
    • Thread sampling (per-thread CPU time every 2 s):
      • one thread-pool worker did a small amount of work in almost every 2 s sample from launch.
        • at the exact moment of the freeze it stopped running and stayed idle in the thread pool (waiting on its work queue), so it isn’t blocked.
          • the other worker threads kept running at the same rate as before.
            • So the job that updates per-process CPU seems to stop being scheduled, rather than hanging.
            • Possible trigger: the system has heavy process churn. Automated headless browser tests start and exit about 2–3 processes per second, with roughly 490 processes and 9,800 threads total. A race when a process exits during the CPU measurement may be involved.

Workaround: close and reopen Task Manager; Resource Monitor keeps showing per-process CPU correctly.

System:

• OS: Windows 11 IoT Enterprise LTSC 24H2, build 26100.9974
• Taskmgr.exe: 10.0.26100.9818
• CPU: Intel Core Ultra 9 275HX (24 cores, hybrid P/E cores, no Hyper-Threading)
• RAM: 64 GB