GTX 1650Ti - Video dxgkrnl fatal error Windows 11

Question

GTX 1650Ti - Video dxgkrnl fatal error Windows 11

Youssef Srasra 0

Problem Description

While launching or playing games (e.g., Counter-Strike 2), the system crashes with a VIDEO_DXGKRNL_FATAL_ERROR (or the NVIDIA Discrete GPU suddenly drops to 0% utilization and disconnects from the system).

To recover temporarily, I must manually disable and re-enable the GTX 1650 Ti in Device Manager or restart the PC. However, relaunching any game causes the crash to recur within minutes, even with graphics settings turned down to the lowest possible options.

Troubleshooting Already Attempted (No Resolution)

Windows Updates: System updated to the latest available Windows 11 build.

System File Checks: Executed sfc /scannow and DISM /Online /Cleanup-Image /RestoreHealth (no unresolved corruption found).

GPU Driver Clean Reinstallation (DDU):

Used Display Driver Uninstaller (DDU) in Safe Mode to remove all NVIDIA and AMD drivers.

  Clean-installed the latest official NVIDIA drivers directly from NVIDIA.
  
     Performed another DDU clean sweep and installed the official manufacturer-recommended graphics drivers directly from Lenovo’s Support page for this specific model.
     
     __In-Game Adjustments:__ Lowered all graphics settings and resolution in-game (_CS2_).

Diagnostic Questions & Requests

Could this issue be tied to a recent Windows 11 DirectX / Graphics Kernel component update interacting poorly with hybrid graphics switching (AMD iGPU + NVIDIA dGPU)?

What are the exact recommended steps to fully isolate whether this is a software/driver state issue vs. a hardware power delivery/PCIe link failure?

Is a rollback to an earlier Windows 11 build advisable, or are there specific power management / BIOS settings on this Lenovo chassis known to resolve DirectX Kernel driver resets?

0 comments

2 answers

Your answer

Answer 1

Since both the current NVIDIA package and Lenovo’s OEM package fail, clean-install one or two older GTX 1650 Ti notebook-driver versions and retest the identical game workload before rolling back Windows.

1. Test an older NVIDIA driver branch

A closely matching GTX 1650 case—games trigger VIDEO_DXGKRNL_FATAL_ERROR, after which the GPU disconnects—was resolved by testing older NVIDIA drivers rather than the newest release.^1^

Download an older notebook driver compatible with the GTX 1650 Ti and your Windows version. Record its version.
Disconnect from the internet so Windows Update cannot insert another display driver during installation.
Perform one final DDU removal of NVIDIA only. This is not a repeat of the same test: its purpose is to prevent files from the current branch being mixed with the older branch.
Install the older driver using Custom installation → Perform a clean installation.
Disable automatic driver replacement in NVIDIA App/GeForce Experience.
Reconnect, restart, and reproduce the same CS2 workload without changing other settings.
If it still fails, repeat with one additional older branch.

Interpretation:

Older driver is stable: likely a driver-branch compatibility/regression issue. Keep that version temporarily and prevent NVIDIA software from replacing it.
The dGPU disappears identically with Lenovo OEM, current NVIDIA, and older NVIDIA drivers: firmware, GPU power delivery, motherboard/PCIe connectivity, or the GPU itself becomes substantially more likely.

The stop code itself proves only that DirectX’s graphics-kernel subsystem detected a violation; it does not identify Windows Update, the NVIDIA driver, or hardware as the root cause.^2^

2. Capture evidence from the next failure

Configure a kernel dump before testing:

Open Control Panel → System and Security → System → Advanced system settings.
Under Startup and Recovery, select Settings.
Set Write debugging information to Kernel memory dump.
Ensure the paging file remains enabled on the Windows boot volume, then restart.^3^
After the crash, preserve:
- %SystemRoot%\Memory.dmp
- %SystemRoot%\Minidump\*.dmp
- The exact crash time

Open the dump in WinDbg using File → Open Crash Dump, then run:

!analyze -v
.bugcheck
lmvm nvlddmkm

Microsoft recommends beginning dump analysis with !analyze; .bugcheck reports the code and parameters.^4^

Also check Event Viewer → Windows Logs → System around the same timestamp for:

WHEA-Logger
PCI Express/root-port errors
display-driver or device-start failures

WHEA can record PCIe bus, PCIe device, and other hardware error data.^5^ Such records strengthen the hardware/firmware path, but do not independently distinguish the GPU, motherboard, firmware, or power-delivery circuitry.

3. Run controlled isolation tests

Test one variable at a time, always using the same game scene or workload:

Test	Result and implication
Force the game to Power saving under Settings → System → Display → Graphics → game → Options	If stable on AMD but consistently fails on NVIDIA, the fault is confined to the dGPU path. Verify the game’s actual GPU engine in Task Manager; the preference alone does not prove which adapter rendered it.
Run on AC with the original Lenovo-rated adapter, without a dock or unnecessary peripherals	Failure only under dGPU load supports a power/firmware/hardware path, but does not identify the failed component.
Perform a clean boot and repeat	Stability indicates interference from a third-party service or startup application. Persistent failure excludes most background software conflicts.

For a clean boot, run msconfig, select Services → Hide all Microsoft services → Disable all, disable startup applications in Task Manager, and restart.^6^ When testing is complete, return to msconfig and choose General → Normal startup.

4. BIOS and Windows rollback

Install BIOS, AMD chipset, and power-management updates only from the support page for the exact Lenovo machine type/model. Do not use BIOS executables supplied for another Lenovo model. Keep AC power connected and have the BitLocker recovery key available before firmware changes.
Do not change undocumented PCIe or hybrid-power parameters. Use a Lenovo-provided Hybrid/Discrete Graphics option only if it exists for that exact chassis; testing discrete-only versus hybrid can isolate switching behavior, but availability is model-specific.
A Windows rollback is justified only if the first crash began immediately after a known feature or quality update and older-driver testing does not resolve it. For a quality update: Settings → Windows Update → Update history → Uninstall updates.^7^
After a feature upgrade, Settings → System → Recovery → Go back is generally available for only ten days and removes apps and drivers installed afterward.^8^

There is no broadly documented Microsoft/NVIDIA bulletin in the available material establishing a single recent Windows 11 hybrid-graphics update as the cause.^9^

If the dGPU disappears across OEM, current, and older drivers, including during a clean boot—and especially if matching WHEA/PCIe records appear—submit the dump and event logs to Lenovo and request motherboard/GPU power-rail and PCIe-link diagnostics. On this laptop architecture, physical GPU reseating is normally a service operation rather than an end-user test.

References

Answer 2

1

Open the old Control Panel.

Set View at the top of the old Control Panel to large icons.

Open the Power Options applet.

On your current Power Plan, click Change plan Settings.

Click 'Change Advanced Power Settings'.

Expand the PCI Express setting in the resulting list.

Disable all settings in the 'Link State Power Management' section.

Restart (not shut down) your PC to apply the change.

2

If that does not resolve the problem.

I see you have already used the DDU method in Safe Mode, the best option is to perform that process again and go to the Nvidia website to install a couple of slightly older versions of the drivers they provide to find a version that is stable on oyur system.