Wi-Fi triggered, Task Manager temporarily mitigates (Win11 24H2, Driver 591.59)”}
Product: NVIDIA GeForce RTX 5060 (Inno3D)
OS: Windows 11 24H2 (fully updated)
Driver versions tested:
591.59 (current, affected)
576.80 / 576.88 (affected)
581.80 (blocked / unsupported)
Summary
After every cold boot, the GPU becomes permanently locked in P0 performance state while idle, resulting in maximum clocks and ~145 W power draw at desktop idle. The issue is triggered when Wi-Fi/networking is enabled. Task Manager temporarily forces correct downclocking only while it remains open.
This appears to be a driver-side WDDM / DXGI idle-detection regression specific to recent drivers and Windows 11 24H2.
Symptoms
GPU stuck in P0 at idle
Idle clocks:
Core ~2700–3100 MHz
Memory ~14801 MHz
Idle power draw: ~145 W
PCIe bus usage reported at 60–100%
No active processes shown in Task Manager
nvidia-smi clock controls ignored while bug is active
Expected idle behavior is ~200 MHz core, ~50 MHz memory, ~10–15 W.
Reproduction Steps
Perform a cold boot (shutdown → power on)
Log into Windows
Enable Wi-Fi (or allow it to initialize automatically)
Observe GPU idle clocks and power draw
Result: GPU remains locked in P0 indefinitely.
Temporary Mitigation Behavior
Opening Task Manager → Performance → GPU immediately forces correct downclocking
Closing Task Manager causes GPU to return to P0 instantly
Warm reboot no longer resolves the issue
Disabling Wi-Fi restores normal idle behavior
Re-enabling Wi-Fi causes GPU to return to P0
This indicates Task Manager’s continuous DXGI engine utilization polling temporarily overrides incorrect driver state.
What Does NOT Fix the Issue
DDU + clean driver install (multiple times)
Disabling Fast Startup
NVIDIA Control Panel power mode changes
Disabling hardware acceleration in apps
Power plans
MSI Afterburner
nvidia-smi -lgc / -rgc / -pl (clock controls ignored while bug is active)
Cold boot or warm reboot
Partial Workarounds
Keeping Task Manager open (impractical)
Forcing continuous DXGI polling via GPU engine performance counters
Power limiting via nvidia-smi -pl (reduces heat but does not downclock)
Strong Indicators
Hardware is functioning correctly
VBIOS is intact
Issue is network/Wi-Fi related
Behavior points to driver idle-detection failure after DXGI/network stack initialization
Likely interaction involving WDDM, DXGI, MPO, or HAGS
Request
Please investigate this as a driver regression affecting RTX 50-series on Windows 11 24H2, where enabling Wi-Fi causes the driver to incorrectly flag the GPU as active and permanently lock it in P0 until continuous DXGI polling occurs.
This issue significantly impacts idle power consumption, thermals, and noise.
I am happy to provide:
Full nvidia-smi -q logs
GPUView traces
Reproduction videos
Additional system diagnostics
Thank you.
Hello,
Welcome to the NVIDIA Developer forums. This community serves the developer ecosystem.
For consumer issues, please visit the GeForce forums.
Or contact the GPU manufacturer since this might very well be a Hardware issue.
As a note to others, the mentioned Bug number in the title does not refer to any NVIDIA bug number in this context.
Thanks.