RTX PRO 5000 Blackwell Laptop GPU: Kernel-mode access violations (0x3B/0x1E) under sustained CUDA compute

Summary

Nine BSODs (seven 0x3B SYSTEM_SERVICE_EXCEPTION, two 0x1E KMODE_EXCEPTION_NOT_HANDLED) over five days of sustained CUDA training on a Dell Pro Max 18 Plus laptop with RTX PRO 5000 Blackwell 24GB GDDR7. All crashes share exception code 0xc0000005 (STATUS_ACCESS_VIOLATION) at different faulting addresses in ntoskrnl address space. Crashes occur on both driver 595.79 (R595 U2) and 595.97 (R595 U3). The same PyTorch RL training workloads run indefinitely without issues on NVIDIA DGX Spark GB10 GPUs (sm_121).

System Configuration

  • Laptop: Dell Pro Max 18 Plus

  • GPU: NVIDIA RTX PRO 5000 Blackwell Generation Laptop GPU, 24GB GDDR7 (sm_120)

  • CPU: Intel Core Ultra (with Intel Arc integrated graphics and Intel AI Boost NPU)

  • OS: Windows 11 Pro

  • Drivers tested: R595 U2 (595.79), R595 U3 (595.97) — clean install via Custom > “Perform a clean installation”

  • CUDA Version: 13.2

  • Framework: PyTorch (RL training, sustained GPU compute at 75–100% utilization)

Reproduction Steps

  1. Run a PyTorch reinforcement learning training workload that sustains 75–100% GPU utilization over hours

  2. System will BSOD with 0x3B or 0x1E, exception code 0xc0000005

  3. Crash latency varies: minutes under active user work, hours when system is unattended

  4. Network state changes (WiFi switch, mobile hotspot activation) are a repeated immediate trigger

Crash Data

Nine crashes total across Mar 23–27, 2026:

# Date/Time Stop Code Driver Faulting Address Minidump Trigger
1 Mar 23, 07:06 AM 0x3B 595.79 032326-30281-01.dmp Active user work
2 Mar 23, 07:15 AM 0x3B 595.79 032326-31609-01.dmp Active user work
3 Mar 23, 11:31 PM 0x3B 595.79 0xfffff80262c5effd 032326-31390-01.dmp Unattended
4 Mar 24, 05:41 AM 0x3B 595.79 0xfffff8077d6c364c 032426-30828-01.dmp Unattended
5 Mar 26, 2:09 PM 0x3B 595.79 0xfffff805170b2c23 032626-25250-01.dmp Active user work
6 Mar 27, 10:34 AM 0x1E 595.79 032726-31203-01.dmp WiFi switch + screen lock
7 Mar 27, 2:29 PM 0x1E 595.79 032726-30718-01.dmp None (user away >1 hour)
8 Mar 27, ~4:30 PM 0x3B 595.97 0xfffff802e88c3657 032726-31500-01.dmp Active user work
9 Mar 27, ~5:03 PM 0x3B 595.97 0xfffff80559902c23 032726-30343-01.dmp Phone hotspot activation

All nine share exception code 0xc0000005 at different faulting addresses, indicating a race condition or memory corruption rather than a single deterministic fault.

Key Findings

1. nvidia-smi confirms Disp.A = On — dGPU handles display compositing alongside CUDA compute. dwm.exe and explorer.exe run on the RTX PRO 5000 simultaneously with CUDA training. The internal display is physically wired through the dGPU (TCC mode fails: “TCC can’t be enabled for device with active display”). NVIDIA Control Panel “Preferred graphics processor = Integrated graphics” does not affect Disp.A status. BIOS “Hybrid Graphics / Advanced Optimus” and “Discrete Direct Output Mode” options do not provide iGPU-only display routing.

2. Minidump analysis (crash #5) identified nvlddmkm.sys and npu_kmd.sys involvement.

  • nvlddmkm.sys loaded at crash time

  • npu_kmd.sys (Intel NPU driver, build 32.0.100.4512) logged bugcheck callbacks with TDR references and all-zero fence IDs on the Compute node

  • Faulting RIP in ntoskrnl address space — a kernel-mode driver passed a bad pointer

3. Intel NPU disablement was a partial mitigation. Disabling the Intel AI Boost NPU in Device Manager eliminated fast crashes under active use (a 5-hour training run with 3 hours of active work succeeded — previously crashed within minutes). However, crashes continued (crashes #6–9), proving the NVIDIA driver is unstable independently.

4. Thermals are not a factor. GPU temperatures across all crashes ranged from 56–78°C, well below the 83°C throttle threshold for RTX PRO 5000.

5. Identical workloads run stably on DGX Spark. The same PyTorch RL training code runs for days without issues on NVIDIA DGX Spark GB10 GPUs (sm_121, 128GB unified memory), confirming the workload is not inherently problematic.

Mitigations Attempted (None Resolved)

Mitigation Result
BIOS: Disabled Adaptive C-States for Discrete Graphics Crashes continued
Dell Optimizer: Ultra Performance thermal mode Crashes continued
Windows: Best Performance power mode, PCI Express Link State PM off Crashes continued
Intel NPU disabled in Device Manager Partial — eliminated fast crashes under active use, but crashes continued under sustained load
Display compositing routed to iGPU (NVIDIA Control Panel) Never took effect — Disp.A = On persisted
Screen/standby timeouts disabled (powercfg) Crashes continued
TCC mode (nvidia-smi -dm 1) Failed — “TCC can’t be enabled for device with active display”
Driver update: 595.79 → 595.97 clean install Two 0x3B crashes within hours of install

Minidumps

All nine minidumps are available in C:\Windows\Minidump\. I can provide them upon request or via file sharing.

Expected Behavior

Sustained CUDA compute workloads should not cause kernel-mode access violations, particularly on a professional workstation GPU designed for extended compute tasks.

Request

This appears to be a driver-level instability in nvlddmkm.sys specific to the RTX PRO 5000 Blackwell laptop GPU under concurrent WDDM display compositing + CUDA compute. The display cannot be decoupled from the dGPU on this hardware (TCC blocked, no BIOS MUX switch). A driver fix that stabilizes the WDDM kernel-mode path during sustained CUDA workloads would resolve this issue.

Bug Report Update — Apr 1, 2026

Summary of New Findings

Crash count is now ten (was nine). The tenth crash confirms the issue is not CUDA-specificnvlddmkm.sys is now unstable under routine WDDM display workloads alone.


New Crash Event

# Date Stop Code Driver Trigger Minidump
10 Apr 1 0x3B 595.97 Firefox startup — no CUDA training running Not yet retrieved

Exception code 0xc0000005 (access violation), consistent with all prior crashes.

This is a critical escalation: The crash was triggered by Firefox launching (WebRender / GPU compositor) with zero CUDA workload active. The instability threshold in nvlddmkm.sys has lowered from “sustained CUDA compute + WDDM” to “any GPU-accelerated WDDM consumer.”


Updated Crash Totals

  • Ten crashes total: eight 0x3B, two 0x1E

  • By driver: seven on 595.79, three on 595.97

  • Driver 595.97 did not resolve: three crashes including one with no CUDA involvement


Additional Finding from Prior Minidump Analysis

Analysis of crash #9 (032726-30343-01.dmp, Mar 27 ~5:03 PM) identified two ExpressVPN kernel-mode drivers loaded at crash time:

  • expressvpn-pkf.sys (packet filter)

  • expressvpn-ovpn-dco.sys (OpenVPN DCO kernel driver)

That crash was triggered at the exact moment of phone hotspot activation — a network state change that the VPN kernel filter driver intercepts. A race condition between the VPN filter and nvlddmkm.sys WDDM path is a plausible secondary trigger for the network-correlated crashes (#6 and #9). This is under investigation on my end but is noted for completeness.


Faulting Address Pattern

Two crashes share an identical low offset, suggesting a reproducible code path within the same driver module:

Crash Faulting RIP Low Offset
#5 — Mar 26, 2:09 PM 0xfffff805170b2c23 ...2c23
#9 — Mar 27, 5:03 PM 0xfffff80559902c23 ...2c23

Different load bases, same function offset — consistent with a single bug in nvlddmkm.sys.


New Workaround Applied

Browser GPU acceleration disabled to prevent crashes from routine browsing:

  • Firefox: layers.acceleration.forced-disabled = true, gfx.webrender.all = false, hardware video decoding disabled

  • Chrome: HardwareAccelerationModeEnabled = 0 via enterprise registry policy


Updated Assessment

The original filing described this as a CUDA + WDDM concurrency issue. The Apr 1 crash proves the scope is broader: nvlddmkm.sys is unstable under any WDDM GPU workload on RTX PRO 5000 (sm_120), including desktop compositing and browser rendering. The Dell Pro Max 18 Plus cannot decouple display from the dGPU (confirmed in original filing), so this instability affects all normal GPU use.

All GPU compute remains offloaded to DGX Sparks. Browser GPU acceleration is now disabled. The laptop is functioning as a CPU-only orchestration node with the dGPU handling only basic display compositing — and even that triggered a crash on Apr 1.

A driver fix for nvlddmkm.sys WDDM stability on RTX PRO 5000 Blackwell laptop GPUs remains the only viable resolution.


Minidump Availability

Nine minidumps available (listed in original report) if anyone is interested. I already sent this as an official bug report.