Hi guys,
Same issues here: Made a post: ROG Strix Scar 16 / RTX 5090 – nvlddmkm.sys Crashes with External Monitor · Issue #47 · Zephkek/Asus...
Hardware
Laptop: ASUS ROG Strix Scar 16 (2025)
GPU: NVIDIA RTX 5090 Laptop GPU
External Monitor: ASUS ROG XG27UCDMG (4K 240 Hz, HDMI 2.1)
- BIOS: v327 (latest)
Windows 11: Fresh install, fully updated (twice)
NVIDIA drivers: Latest from ASUS support (v576)
Using G-Helper now in standard mode, optimus in Nvidia and Dynamic Bios. Got more issues in Ultimate and dGPU only.
Problem Description
Frequent BSOD with nvlddmkm.sys (SYSTEM_THREAD_EXCEPTION_NOT_HANDLED, exception code c0000094)
part of the dump(s)
Arg1: ffffffffc0000094, The exception code that was not handled
Arg2: fffff8057766f4fa, The address that the exception occurred at
Arg3: ffffd30f9a91d3a8, Exception Record Address
Arg4: ffffd30f9a91cbb0, Context Record Address
DUMP_FILE_ATTRIBUTES: 0x21808 Kernel Generated Triage Dump FAULTING_THREAD: ffffbf88f9b23080EXCEPTION_RECORD: ffffd30f9a91d3a8 -- (.exr 0xffffd30f9a91d3a8)ExceptionAddress: fffff8057766f4fa (nvlddmkm+0x000000000073f4fa) ExceptionCode: c0000094 (Integer divide-by-zero) ExceptionFlags: 00000000 NumberParameters: 0
Before reboot and crash:
Session "dc3a3596-71e1-45a3-b2ea-39ad5322fe51" failed to start with the following error: 0xC0000022
This event happens every 60 seconds:
Crashes occur: (Black screen, then "pc ran into problem")
Using external monitor (HDMI 2.1) in Edge / Netflix or randomly (Funny thing i got it 3 times in a row when browsing to Netflix...)
After closing games and doing something (switch from dGPU to iGPU probable cause)
Sometimes immediately after display wake / GPU power state changes
Internal laptop display havent tested much
DSC, VRR / G-Sync, and monitor power state seem to influence crash frequency
When rebooting the external display sometimes just wont start or boot, need to duplicate the screen and try several times before it wants to go.
Analysis / Likely Cause
NVIDIA driver encounters GPU power-state race when external monitor is connected, consistent with Zephkek GitHub ROG AML deep dive:
GPU enters impossible state: firmware insists OFF, Windows needs ON → driver thread blocks → 30s watchdog expires → BSOD
Contributing factors: (in my eyes possible)
High-res / high-refresh scenarios require DSC on HDMI 2.1 / DP 1.4
HDMI 2.1 / DP 1.4 bandwidth limitations
VRR / G-Sync active during DSC operation
ASUS PowerSync firmware interaction
Windows / NVIDIA Settings (currently testing)
HAGS (Hardware Accelerated GPU Scheduling) OFF
G-Sync / VRR OFF
DSC ON (Off only gets me 4K at 120Hz ofcourse)
NVIDIA GPU: Optimus + Dynamic in Bios (dGPU + Ultimate gave me alot black screens)
PCI Express Link State: max performance / disabled power savings
CPU: set to 100% performance mode in power plan
Turned off hardware acceleration in Edge / browsers / apps
Custom fan curve applied for stable temperatures (Using G-Helper instead of Armory Crate)
BIOS / Hardware Settings
BIOS GPU mode: Dynamic
Latest BIOS v327 installed (this got my temps up quite a bit) Weird thing is this bios came out end of september, hoping this is not their "FIX" for the stutters and black screen BSOD problems MUX related
Monitor Settings
ASUS PowerSync OFF
Variable refresh rate and G-Sync disabled
DSC ON
Other
Fresh Windows 11 install
Temperature monitoring: "normal" (<85°C) topping/spiking to 95-101°C but its not load-related
Nvidia driver recommended by Asus (v576.65)
Critical recommended drivers from Asus
LatencyMon:
Observations from LatencyMon
ACPI.sys shows very high DPC routine execution times (~34,665 µs), which is extremely high.
Wdf01000.sys (Windows Kernel Mode Driver Framework) also has medium high ISR times (~787 µs).
Dxgkrnl.sys (Direct X graphics kernel) also has medium high ISR times (~910 µs).
High DPC / ISR execution times like this often cause:
- Buffer underruns in audio/video
- System stutters
- Potential race conditions with GPU power state, especially with NVIDIA DSC / HDMI 2.1
Combined with previous nvlddmkm.sys BSODs, these ACPI / WDF delays strongly suggest a power-state / driver handshake issue between the laptop, NVIDIA GPU, and external monitor likely caused by ACPI.sys / BIOS issues: Laptop’s ACPI implementation may be delaying GPU power-on or system wake events.
High DPC from system drivers: WDF or ACPI calls take too long, blocking GPU threads → could trigger the “impossible state” Zephkek described.
External monitor + DSC / HDMI 2.1: DSC + VRR / high refresh rates rely on fast DPC/ISR; any delay can trigger GPU driver timeout / BSOD.
Not sure if related but im also getting 20k "Display-External" Errors:
Hi Falcon2_ROG,
This isn't a random crash; it's a specific, repeatable driver bug caused by a race condition, and LatencyMon data points directly to the underlying cause.
The system crashes with a "PC ran into a problem" black screen, most often when using the external monitor, waking the display, or switching GPU power states (e.g., closing a game).
The minidump analysis shows a clear "divide by zero" error within the NVIDIA driver:
The driver attempted to divide eax (0xeff) by r12d, which was 0. This is a classic symptom of a race condition where the driver expected a value in r12 (e.g., from a power state or clock-speed query) but received 0 because another system process was stalled.
Here is the exact sequence of events that matches what i'm seeing:
- The Hang (Time: 0 seconds): The firmware/OS conflict happens. The nvlddmkm.sys driver thread, which is responsible for the display, enters the "impossible state" and freezes.
- The Black Screen (Time: 0.1 seconds): Because the driver thread is frozen, it stops sending a signal to the monitor. My screen immediately goes black.
- The Countdown (Time: 0.1s to 30s): The rest of Windows is still running. A kernel-level "watchdog" timer sees that this critical system thread is no longer responding. This watchdog's job is to prevent the entire OS from being permanently frozen. It starts a countdown, which is typically 30 seconds. This is the 30-60 second delay i am actively counting.
- The Bugcheck (Time: ~30-60 seconds): The timer expires. The watchdog says, "The thread is dead. The system is in an unrecoverable state." It then intentionally triggers the BSOD (the 0x7E bugcheck) to collect the memory dump and force a reboot.
The Root Cause: LatencyMon Analysis
LatencyMon reveals the "smoking gun": ACPI.sys (the Windows driver that interfaces with the BIOS for power management) is experiencing catastrophic DPC latency.
ACPI.sys (ACPI Driver for NT)
DPC Routine Execution Time: ~34,665 µs (This is astronomically high)
dxgkrnl.sys (DirectX Graphics Kernel)
ISR Routine Execution Time: ~910 µs
Wdf01000.sys (Kernel Mode Driver Framework)
ISR Routine Execution Time: ~787 µs
A 34,000µs (0.034 seconds) stall in ACPI.sys is more than enough to cause the race condition. The BIOS (v327) is failing to respond to a power-state request from the OS in time. This stalls the NVIDIA driver, which is waiting for a value. The driver's watchdog times out or the function proceeds with a 0 value, causing the "divide by zero" crash.
This aligns perfectly with research done by others, such as the "Asus-ROG-Aml-Deep-Dive" on GitHub, suggesting firmware-level issues with GPU power states.
Triggers & Other Symptoms
- Using the external 4K 240Hz monitor (which requires DSC).
- Switching from dGPU back to iGPU
- Browsing to Netflix gave me the black screen and BSOD 3 times in a row. The browser requests Hardware Accelaration to decode the protected DRM video stream. This forces the system to wake up the more powerfull Nvidia RTX 5090 GPU and switch the task to it. The laptops BIOS firmware is causing this handshake to fail catastrophically.
- Over 20,000 "Display-External" errors logged in Event Viewer.
- The external display often fails to initialize on boot/reboot, requiring multiple attempts to duplicate/extend the screen.
Troubleshooting Performed
2x Fresh Windows 11 Installs
- Latest BIOS v327 (Note: this BIOS seems to have increased idle/load temps)
- NVIDIA driver v576 (from ASUS)
- HAGS (Hardware Accelerated GPU Scheduling) OFF
- G-Sync / VRR OFF
- Browser Hardware Acceleration OFF
- PCI Express Link State Power Management OFF
- BIOS Mode: Dynamic (Ultimate/dGPU mode is even less stable)
This is not a faulty Windows install or a simple driver conflict. This is a deep firmware/BIOS bug in how ACPI.sys handles power-state transitions for the GPU, especially under high-bandwidth (DSC) loads.
Please escalate this to the firmware engineering team. We need a BIOS update that fixes the ACPI.sys DPC latency.
Thanks,
Kind regards,
Kevin
