Fix DRIVER_POWER_STATE_FAILURE Windows 11 (2026)

Download BlueScreenView from nirsoft.net — free, no installation, reads the minidump files Windows creates after every crash. Open it, click your most recent BSOD entry, look at the “Caused By Driver” column in the bottom pane. That’s the exact .sys file responsible. Thirty seconds, no guessing.

A customer’s Dell XPS 15 blue-screened every time she closed the lid. Worked fine while she used it, but the moment she put it to sleep and opened it back up — DRIVER_POWER_STATE_FAILURE, crash, restart. She’d already replaced her GPU driver and run every scan she could find. Nothing worked because she was guessing at the cause. We opened her minidump in BlueScreenView and the Caused By Driver column said nvlddmkm.sys — NVIDIA’s display driver. Specifically version 572.16 was failing the D3-to-D0 power transition when the GPU woke from sleep. Rolled her back to 566.36 using DDU in Safe Mode, crashes stopped completely. Twelve minutes.

Stop code 0x9F means a driver sent a power state request to a hardware device and the device never responded. Windows waits a few seconds, gives up, and crashes because it can’t leave hardware stuck between power states. This happens during sleep, wake, hibernate, or shutdown — any moment Windows tells a device to change its power consumption level. About 65% of the cases we diagnose remotely, it’s the GPU driver. Network adapters are second.

Power Settings

If you need the crashes to stop right now while you investigate, change the power plan. Open Control Panel (not the modern Settings app), Power Options, switch to High Performance — you might need to click “Show additional plans.” The Balanced plan constantly transitions devices between power states to save energy, and if any driver has a power state bug, Balanced pokes it every few minutes. High Performance keeps devices awake. This fixes maybe 25% of cases outright, though it’s more bandage than cure.

The single setting that causes more of these crashes than any individual driver bug: PCI Express Link State Power Management. In Power Options, click “Change plan settings” next to your active plan, then “Change advanced power settings,” expand PCI Express, set Link State Power Management to Off. ASPM puts PCIe devices — GPU, NVMe SSD, sometimes WiFi — into low-power states between operations. Certain NVMe controllers and GPUs don’t wake up cleanly from ASPM L1 states, especially on AMD boards with B550 and B650 chipsets. The device sleeps, Windows sends the wake command, the device doesn’t respond, kernel crashes with 0x9F.

Fast Startup is the other one. It saves a partial hibernation image at shutdown so the next boot loads faster, but it doesn’t fully reinitialize drivers on startup — it restores them from the image. If a driver’s power state was corrupted when the image was saved, Fast Startup recreates that corruption on every boot. Disable it through Control Panel, Power Options, “Choose what the power buttons do,” “Change settings that are currently unavailable,” uncheck “Turn on fast startup.”

Fixing the Specific Driver

Once BlueScreenView tells you which .sys file is crashing, the fix depends on the hardware.

For GPU drivers — nvlddmkm.sys on NVIDIA, atikmpag.sys on AMD — a standard Device Manager reinstall almost never works because it doesn’t clean the old driver’s power management registry entries. These same drivers also cause SYSTEM_SERVICE_EXCEPTION crashes under load, different stop code but same root cause. Download DDU (Display Driver Uninstaller), boot into Safe Mode (Settings, System, Recovery, Restart now, Troubleshoot, Startup Settings, press 4), run DDU, select “Clean and restart.” This strips every trace including registry keys, shader caches, and the power profiles that are almost certainly the crash source. After restart, install a fresh driver from nvidia.com. Before installing, search “nvidia [version number] BSOD” to check if the current release has known issues — the Studio Driver is sometimes the safer choice over Game Ready. On NVIDIA, the 566.x series is significantly more stable for sleep/wake than 570.x-572.x on most RTX cards.

AMD uses their own Cleanup Utility instead of DDU — downloadable from amd.com. Removes all Radeon components including Adrenalin and power profiles. Their RDNA 3 cards still occasionally hit power state issues with specific AGESA firmware versions, so check for a BIOS update at the same time.

For network adapters — e1d65x64.sys is Intel Ethernet, rt640x64.sys is Realtek — the fix is different. Device Manager, find your adapter under Network adapters, double-click it, Power Management tab, uncheck “Allow the computer to turn off this device to save power.” Advanced tab, disable Wake on Magic Packet, Wake on Pattern Match, and Energy Efficient Ethernet. The Intel I225-V 2.5Gb adapter on Z590/Z690/B660 boards and the Realtek RTL8125B on budget B550/B650 boards are the two we see most often — both have known firmware bugs in their power state transition handling.

If BlueScreenView points to ntoskrnl.exe instead of a specific driver, the crash is at a level deeper than any single driver. Run DISM /Online /Cleanup-Image /RestoreHealth then sfc /scannow to fix corrupted system components. Also check for a BIOS update — your BIOS contains ACPI power management tables that define how every device handles power transitions, and an outdated BIOS with buggy tables causes power state failures even if every driver is perfect. AMD Ryzen 7000 and Intel 12th-14th Gen have both had half a dozen AGESA and microcode patches in 2025-2026 specifically for power state handling. Check your motherboard manufacturer’s support page, not Windows Update.

One more thing that trips people up — you fix the driver, everything works for two weeks, then Windows Update quietly installs a newer version of the exact driver you replaced and the crashes come back. Prevent this by going to Settings, System, About, Advanced system settings, Hardware tab, Device Installation Settings, pick “No.” Our full driver update guide covers how to manage this safely. For deeper blue screen analysis including CRITICAL_PROCESS_DIED, DPC_WATCHDOG_VIOLATION, and KERNEL_DATA_INPAGE_ERROR, or if your machine freezes without a blue screen instead, we can pull minidump files and run full WinDbg analysis remotely in about twenty minutes.