[Windows Portable][Blackwell] Official MiniMax H3 Ref workflow ends in nvlddmkm DPC timeout and 0x133 reboot on a fresh v0.31.0 install

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

The official MiniMax H3 Reference-to-Video workflow should complete on a supported RTX 50-series GPU, or report a recoverable CUDA/application error. It should not leave the NVIDIA kernel driver in a DPC long enough to trigger `DPC_WATCHDOG_VIOLATION (0x133)` and reboot Windows.

### Actual Behavior

After a completely fresh installation of ComfyUI 0.31.0, I ran the official MiniMax H3 Reference-to-Video workflow. During video generation, pc automatically restarted without a visible Python exception.

The minidump, `081126-17718-01.dmp`, records:

```text
DPC_WATCHDOG_VIOLATION (0x133)
Arg1: 0 (one DPC/ISR exceeded its time allotment)
Arg2: 0x500
Arg3: 0x500
DPC_TIMEOUT_TYPE: SINGLE_DPC_TIMEOUT_EXCEEDED
NVIDIA driver: nvlddmkm.sys 32.0.16.1088 / 610.88
Crash time: 2026-08-11 17:02:18 +08:00
System uptime: 1:05:53
```

The active stack contains 17 consecutive `nvlddmkm.sys` frames before returning to the Windows DPC path:

```text
nvlddmkm+0xe9023
nvlddmkm+0x69f535
nvlddmkm+0x1252f3
nvlddmkm+0x309b2e
nvlddmkm+0x124d54
nvlddmkm+0x10ebb0
nvlddmkm+0xa98ed9
nvlddmkm+0xa97702
nvlddmkm+0xa978cd
nvlddmkm+0x14eea4
nvlddmkm+0x3a43b6
nvlddmkm+0xacf571
nvlddmkm+0xacde4a
nvlddmkm+0x129318
nvlddmkm+0x12a87c
nvlddmkm+0x12b8cc
nvlddmkm+0xe93ba
```

### Steps to Reproduce

1. Install ComfyUI 0.31.0 on Windows 11.
2. launch.bat (I haven't install any custom nodes so `--disable-all-custom-nodes` is not added):

```powershell
.\python_embeded\python.exe -s ComfyUI\main.py `
--output-directory "C:\Users\\Pictures\ComfyUI" `
--user-directory "C:\Users\\Documents\ComfyUI" `
--disable-auto-launch `
--disable-api-nodes `
--disable-metadata `
--fast fp16_accumulation
```

3. Open the official workflow `video_minimax_h3_r2v` from `comfyui_workflow_templates_json 0.1.37`.
4. Use the official Comfy-Org model files:

- `minimax_h3_ref2va_pruned_int8_convrot.safetensors`
- `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors`
- `minimax_h3_video_vae_fp16.safetensors`
- `minimax_h3_audio_vae_fp32.safetensors`

5. Set the generation parameters to `0.8 MP`, `3:4`, `5 seconds`, `20 steps`, `simple` scheduler, and `res_multistep` sampler.
6. Queue the generation. PC may automatically reboot during generation without a Python exception being printed.

### Debug Logs

```powershell
C:\Users\\AppData\Local\Programs\ComfyUI>.\python_embeded\python.exe -s ComfyUI\main.py --output-directory "C:\Users\\Pictures\ComfyUI" --user-directory "C:\Users\\Documents\ComfyUI" --disable-auto-launch --disable-api-nodes --disable-metadata --fast fp16_accumulation
[INFO] setup plugin alembic.autogenerate.schemas
[INFO] setup plugin alembic.autogenerate.tables
[INFO] setup plugin alembic.autogenerate.types
[INFO] setup plugin alembic.autogenerate.constraints
[INFO] setup plugin alembic.autogenerate.defaults
[INFO] setup plugin alembic.autogenerate.comments
[INFO] Adding extra search path checkpoints C:\Users\\Files\Models\checkpoints
[INFO] Adding extra search path configs C:\Users\\Files\Models\configs
[INFO] Adding extra search path loras C:\Users\\Files\Models\loras
[INFO] Adding extra search path vae C:\Users\\Files\Models\vae
[INFO] Adding extra search path text_encoders C:\Users\\Files\Models\text_encoders
[INFO] Adding extra search path text_encoders C:\Users\\Files\Models\clip
[INFO] Adding extra search path diffusion_models C:\Users\\Files\Models\unet
[INFO] Adding extra search path diffusion_models C:\Users\\Files\Models\diffusion_models
[INFO] Adding extra search path clip_vision C:\Users\\Files\Models\clip_vision
[INFO] Adding extra search path style_models C:\Users\\Files\Models\style_models
[INFO] Adding extra search path embeddings C:\Users\\Files\Models\embeddings
[INFO] Adding extra search path diffusers C:\Users\\Files\Models\diffusers
[INFO] Adding extra search path vae_approx C:\Users\\Files\Models\vae_approx
[INFO] Adding extra search path controlnet C:\Users\\Files\Models\controlnet
[INFO] Adding extra search path controlnet C:\Users\\Files\Models\t2i_adapter
[INFO] Adding extra search path gligen C:\Users\\Files\Models\gligen
[INFO] Adding extra search path upscale_models C:\Users\\Files\Models\upscale_models
[INFO] Adding extra search path latent_upscale_models C:\Users\\Files\Models\latent_upscale_models
[INFO] Adding extra search path datasets C:\Users\\Files\Models\datasets
[INFO] Adding extra search path hypernetworks C:\Users\\Files\Models\hypernetworks
[INFO] Adding extra search path photomaker C:\Users\\Files\Models\photomaker
[INFO] Adding extra search path classifiers C:\Users\\Files\Models\classifiers
[INFO] Adding extra search path model_patches C:\Users\\Files\Models\model_patches
[INFO] Adding extra search path audio_encoders C:\Users\\Files\Models\audio_encoders
[INFO] Adding extra search path background_removal C:\Users\\Files\Models\background_removal
[INFO] Adding extra search path frame_interpolation C:\Users\\Files\Models\frame_interpolation
[INFO] Adding extra search path geometry_estimation C:\Users\\Files\Models\geometry_estimation
[INFO] Adding extra search path optical_flow C:\Users\\Files\Models\optical_flow
[INFO] Adding extra search path detection C:\Users\\Files\Models\detection
[INFO] Setting output directory to: C:\Users\\Pictures\ComfyUI
[INFO] Setting user directory to: C:\Users\\Documents\ComfyUI
[INFO] Found comfy_kitchen backend hip: {'available': False, 'disabled': False, 'unavailable_reason': 'PyTorch ROCm/HIP runtime not available', 'capabilities': []}
[INFO] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_embedding', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'dequantize_w4a8_int8_weight', 'gemv_awq_w4a16', 'int8_linear', 'na3d', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'quantize_w4a8_int8_weight', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'rotate_int8_convrot_weight', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8', 'w4a8_int8_linear']}
[INFO] Found comfy_kitchen backend triton: {'available': False, 'disabled': True, 'unavailable_reason': "ImportError: No module named 'triton'", 'capabilities': []}
[INFO] Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope1_', 'apply_rope_', 'apply_rope_split_half', 'apply_rope_split_half1', 'apply_rope_split_half1_', 'apply_rope_split_half_', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'dequantize_w4a8_int8_weight', 'gemv_awq_w4a16', 'int8_linear', 'na3d', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'quantize_w4a8_int8_weight', 'rms_adaln', 'rms_rope', 'rms_rope1', 'rms_rope1_', 'rms_rope_', 'rms_rope_split_half', 'rms_rope_split_half1', 'rms_rope_split_half1_', 'rms_rope_split_half_', 'rotate_int8_convrot_weight', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8', 'w4a8_int8_linear']}
[INFO] Checkpoint files will always be loaded safely.
[INFO] Total VRAM 12227 MB, total RAM 31935 MB
[INFO] pytorch version: 2.13.0+cu130
[INFO] Enabled fp16 accumulation.
[INFO] Set vram state to: NORMAL_VRAM
[INFO] Device: cuda:0 NVIDIA GeForce RTX 5070 Ti Laptop GPU : cudaMallocAsync
[INFO] Using async weight offloading with 2 streams
[INFO] Enabled pinned memory 12773.0
[INFO] Using pytorch attention
[INFO] aimdo: src-win/cuda-detour.c:38:INFO:aimdo_setup_hooks: installing 6 hooks
[INFO] aimdo: src/control.c:262:INFO:comfy-aimdo NVML pressure enabled
[INFO] aimdo: src-win/shmem-detect.c:80:INFO:comfy-aimdo WDDM adapter match: NVIDIA GeForce RTX 5070 Ti Laptop GPU runtime_luid=00000000:0001065d dxgi_luid=00000000:0001065d
[INFO] aimdo: src/control.c:277:INFO:comfy-aimdo inited for GPU: NVIDIA GeForce RTX 5070 Ti Laptop GPU (VRAM: 12226 MB)
[INFO] DynamicVRAM support detected and enabled
[INFO] Python version: 3.13.14 (tags/v3.13.14:fd17997, Jun 10 2026, 13:03:48) [MSC v.1944 64 bit (AMD64)]
[INFO] ComfyUI version: 0.31.0
[INFO] comfy-aimdo version: 0.4.13
[INFO] comfy-kitchen version: 0.2.30
[INFO] comfyui-frontend-package version: 1.48.7
[INFO] comfyui-workflow-templates version: 0.11.37
[INFO] comfyui-embedded-docs version: 0.5.9
[INFO] comfy-kitchen version: 0.2.30
[INFO] comfy-aimdo version: 0.4.13
[INFO] [Prompt Server] web root: C:\Users\\AppData\Local\Programs\ComfyUI\python_embeded\Lib\site-packages\comfyui_frontend_package\static
[INFO] Asset seeder disabled
[INFO] No OpenGL_accelerate module loaded: No module named 'OpenGL_accelerate'
[INFO]
Import times for custom nodes:
[INFO] 0.0 seconds: C:\Users\\AppData\Local\Programs\ComfyUI\ComfyUI\custom_nodes\websocket_image_save.py
[INFO]
[ERROR] Failed to initialize database. Please ensure you have installed the latest requirements. If the error persists, please report this as in future the database will be required: (sqlite3.OperationalError) unable to open database file
(Background on this error at: https://sqlalche.me/e/20/e3q8)
[INFO] Using RAM pressure cache.
[INFO] Starting server

[INFO] To see the GUI go to: http://127.0.0.1:8188
```

### Other

### System

- Laptop: MECHREVO CANGLONG 16 Ultra
- BIOS: `N.1.12MRO16`
- OS: Windows 11 Pro, build `26200`
- CPU: AMD Ryzen 9 9955HX3D, 16 cores / 32 threads
- RAM: 32 GB
- GPU: NVIDIA GeForce RTX 5070 Ti Laptop GPU, 12 GB VRAM, Blackwell
- GPU VBIOS: `98.05.31.00.70`
- Graphics mode: hybrid graphics
- NVIDIA driver: `610.88` / `32.0.16.1088`, clean-installed
- NVIDIA Overlay: disabled

### Workflow File

[MiniMax-H3-Ref-Official.json](https://github.com/user-attachments/files/30930806/MiniMax-H3-Ref-Official.json)

The database initialization error is present at startup, before the generation begins. The server nevertheless starts successfully; there is no evidence in this transcript that this message is related to the later NVIDIA kernel timeout.

The clean-install environment was verified from the installed files:

```text
ComfyUI: 0.31.0
Python: 3.13.14
PyTorch: 2.13.0+cu130
comfy-aimdo: 0.4.13
comfy-kitchen: 0.2.30
comfyui-frontend-package: 1.48.7
comfyui-workflow-templates: 0.11.37
comfyui_workflow_templates_json: 0.1.37
SageAttention: not installed
Triton: not installed
Third-party custom nodes: none
```

Windows recorded the following for the primary occurrence:

```text
2026-08-11 17:02:39 Microsoft-Windows-Kernel-Power Event 41
BugcheckCode=307 (0x133)
BugcheckParameter1=0x0
BugcheckParameter2=0x500
BugcheckParameter3=0x500
BugcheckParameter4=0xfffff80072dc53c8
LongPowerButtonPressDetected=false
WHEABootErrorCount=0

2026-08-11 17:02:52 Microsoft-Windows-WER-SystemErrorReporting Event 1001
Bugcheck: 0x00000133 (0x0, 0x500, 0x500, 0xfffff80072dc53c8)
Dump: C:\Windows\Minidump\081126-17718-01.dmp
```

Start by reproducing the official video_minimax_h3_r2v workflow with the listed ComfyUI 0.31.0 launch command and model files on the reported RTX 5070 Ti setup. Review the startup log, workflow JSON, and minidump details, while separating the database initialization warning from the later nvlddmkm timeout. Done means identifying a reproducible ComfyUI cause and a scoped fix, or documenting evidence that the failure is outside ComfyUI.

Written by the indexing model from the issue text.