segfault on AMD AM4 with 4 DIMMs populated

### 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

Graph completes and image is rendered

### Actual Behavior

I get a segmentation fault and crash in torch_cpu.dll when all 4 DIMMs are populated (4 x 16GB). Doesn't matter the model, nodes or graph. Using the built in example templates in Comfy UI. Consistent with Flux 2.d and WAN 2.2. System memory fills up (expected) then segfault. Tried 2 different motherboards, a Ryzen 2900 and a Ryzen 3950X. Tried different memory (Samsung, Crucial). Happens even in CPU mode. BIOS updated. No weird OC settings. Tried mem voltage tweaks. Used DDU and reinstalled NVidia Studio Driver (Using 591.74). Disabled Sysmemm Fallback. Logs below are my system python 3.12 but this happened using Comfy UI portable with the embedded python 3.13. Latest everything.

I tried posting this on Reddit to ask but a bot deleted my question in about a minute (IDK how to use Reddit anymore) so I'm sorry for adding another issue here: Is this a known issue? Are memory freakouts just going to be a thing with 4 DIMMs on this AMD setup? Is this a Windowsism that would go away on Linux (not ideal but wondering).

### Steps to Reproduce

Using the build in WAN 2.2 or Flux 2.d template

### Debug Logs

```powershell
Log Name: Application
Source: Application Error
Date: 2/16/2026 3:00:22 PM
Event ID: 1000
Task Category: Application Crashing Events
Level: Error
Keywords:
User: SHED\Kevin
Computer: SHED
Description:
Faulting application name: python.exe, version: 3.12.11150.1013, time stamp: 0x68a37ca7
Faulting module name: torch_cpu.dll, version: 0.0.0.0, time stamp: 0x6969929f
Exception code: 0xc0000005
Fault offset: 0x000000000664a0b9
Faulting process id: 0x4A94
Faulting application start time: 0x1DC9F7EAAB1E2A4
Faulting application path: C:\Users\Kevin\AppData\Roaming\uv\python\cpython-3.12.11-windows-x86_64-none\python.exe
Faulting module path: M:\ComfyUI\.venv\Lib\site-packages\torch\lib\torch_cpu.dll
Report Id: bf647d36-6fb7-46af-8ab2-354d3240ee46
Faulting package full name:
Faulting package-relative application ID:
Event Xml:

1000
0
2
100
0
0x8000000000000000

3785


Application
SHED



python.exe
3.12.11150.1013
68a37ca7
torch_cpu.dll
0.0.0.0
6969929f
c0000005
000000000664a0b9
0x4a94
0x1dc9f7eaab1e2a4
C:\Users\Kevin\AppData\Roaming\uv\python\cpython-3.12.11-windows-x86_64-none\python.exe
M:\ComfyUI\.venv\Lib\site-packages\torch\lib\torch_cpu.dll
bf647d36-6fb7-46af-8ab2-354d3240ee46

** ComfyUI startup time: 2026-02-16 14:58:53.877
[2026-02-16 14:58:53.877] ** Platform: Windows
[2026-02-16 14:58:53.877] ** Python version: 3.12.11 (main, Aug 18 2025, 19:17:54) [MSC v.1944 64 bit (AMD64)]
[2026-02-16 14:58:53.877] ** Python executable: M:\ComfyUI\.venv\Scripts\python.exe
[2026-02-16 14:58:53.877] ** ComfyUI Path: C:\Users\Kevin\AppData\Local\Programs\ComfyUI\resources\ComfyUI
[2026-02-16 14:58:53.877] ** ComfyUI Base Folder Path: C:\Users\Kevin\AppData\Local\Programs\ComfyUI\resources\ComfyUI
[2026-02-16 14:58:53.878] ** User directory: M:\ComfyUI\user
[2026-02-16 14:58:53.878] ** ComfyUI-Manager config path: M:\ComfyUI\user\__manager\config.ini
[2026-02-16 14:58:53.878] ** Log path: M:\ComfyUI\user\comfyui.log
[ComfyUI-Manager] Skipped fixing the 'comfyui-frontend-package' dependency because the ComfyUI is outdated.
[2026-02-16 14:58:54.801] [PRE] ComfyUI-Manager
[2026-02-16 14:59:00.304] Checkpoint files will always be loaded safely.
[2026-02-16 14:59:02.009] Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
[2026-02-16 14:59:02.009] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
[2026-02-16 14:59:02.009] Found comfy_kitchen backend triton: {'available': False, 'disabled': True, 'unavailable_reason': "ImportError: No module named 'triton'", 'capabilities': []}
[2026-02-16 14:59:02.185] Total VRAM 24576 MB, total RAM 65461 MB
[2026-02-16 14:59:02.186] pytorch version: 2.10.0+cu130
[2026-02-16 14:59:02.186] Set vram state to: NORMAL_VRAM
[2026-02-16 14:59:02.186] Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
[2026-02-16 14:59:02.206] Using async weight offloading with 2 streams
[2026-02-16 14:59:02.206] Enabled pinned memory 29457.0
[2026-02-16 14:59:02.214] working around nvidia conv3d memory bug.
[2026-02-16 14:59:03.782] Using pytorch attention
[2026-02-16 14:59:21.299] Python version: 3.12.11 (main, Aug 18 2025, 19:17:54) [MSC v.1944 64 bit (AMD64)]
[2026-02-16 14:59:21.299] ComfyUI version: 0.13.0
[2026-02-16 14:59:21.322] [Prompt Server] web root: C:\Users\Kevin\AppData\Local\Programs\ComfyUI\resources\ComfyUI\web_custom_versions\desktop_app
[2026-02-16 14:59:21.323] [START] ComfyUI-Manager
[2026-02-16 14:59:21.479] [ComfyUI-Manager] network_mode: public
[2026-02-16 14:59:21.483] [ComfyUI-Manager] The matrix sharing feature has been disabled because the `matrix-nio` dependency is not installed.
To use this feature, please run the following command:
M:\ComfyUI\.venv\Scripts\python.exe -m pip install matrix-nio

[2026-02-16 14:59:23.133] Blocked by policy: C:\Users\Kevin\AppData\Local\Programs\ComfyUI\resources\ComfyUI\custom_nodes\ComfyUI-Manager
[2026-02-16 14:59:23.133]
Import times for custom nodes:
[2026-02-16 14:59:23.134] 0.0 seconds: C:\Users\Kevin\AppData\Local\Programs\ComfyUI\resources\ComfyUI\custom_nodes\websocket_image_save.py
[2026-02-16 14:59:23.134]
[2026-02-16 14:59:23.193] Context impl SQLiteImpl.
[2026-02-16 14:59:23.193] Will assume non-transactional DDL.
[2026-02-16 14:59:23.224] Assets scan(roots=['models']) completed in 0.023s (created=0, skipped_existing=23, orphans_pruned=0, total_seen=23)
[2026-02-16 14:59:23.337] Starting server

[2026-02-16 14:59:23.337] To see the GUI go to: http://127.0.0.1:8000
[2026-02-16 14:59:24.778] comfyui-frontend-package not found in requirements.txt
[2026-02-16 14:59:35.216] got prompt
[2026-02-16 14:59:35.309] Using pytorch attention in VAE
[2026-02-16 14:59:35.310] Using pytorch attention in VAE
[2026-02-16 14:59:35.549] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[2026-02-16 15:00:00.654] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
[2026-02-16 15:00:00.719] Requested to load Flux2TEModel_
[2026-02-16 15:00:18.875] loaded partially; 21784.80 MB usable, 20180.59 MB loaded, 12900.00 MB offloaded, 1600.00 MB buffer reserved, lowvram patches: 0
```

### Other

_No response_