Even after reboot, comfyui API is taking WAY TOO LONG to load the model!!! Like 300-400 seconds!!!

### 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 model should load from my SSD in 10-30 seconds at most.
I run a P40 and a RTX 2070 card. Both are now taking a ridicuously long time to load any model!

### Actual Behavior

- Rebooted machine for clean work environment
- Website and ComfyUI API interfaces started. Waited for ComfyUI portable to finish loading to 'All startup tasks complete'
- Send command from website to generate an image
- Stuck waiting for 402 seconds to load SD 3.5 model and execute! Image generation takes 30 seconds at most!
- Stuck waiting for 279 seconds to load SDXL model and execute! Image generation takes 12 seconds!

### Steps to Reproduce

- Rebooted machine for clean work environment
- Website and ComfyUI API interfaces started. Waited for ComfyUI portable to finish loading to 'All startup tasks complete'
- Send command from website to generate an image
- Stuck waiting for 402 seconds to load SD 3.5 model and execute! Image generation takes 30 seconds at most!
- Stuck waiting for 279 seconds to load SDXL model and execute! Image generation takes 12 seconds!

### Debug Logs

```powershell
P40 GPU
[START] Security scan
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-09-28 23:22:03.999
** Platform: Windows
** Python version: 3.12.10 (tags/v3.12.10:0cc8128, Apr 8 2025, 12:21:36) [MSC v.1943 64 bit (AMD64)]
** Python executable: C:\nginx-1.27.4\html\ComfyUI_windows_portable\python_embeded\python.exe
** ComfyUI Path: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI
** ComfyUI Base Folder Path: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI
** User directory: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\user
** ComfyUI-Manager config path: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\user\default\ComfyUI-Manager\config.ini
** Log path: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\user\comfyui.log

Prestartup times for custom nodes:
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy
48.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager

Checkpoint files will always be loaded safely.
Total VRAM 24474 MB, total RAM 65464 MB
pytorch version: 2.7.0+cu128
Set vram state to: NORMAL_VRAM
Device: cuda:0 Tesla P40 : native
Using pytorch attention
Python version: 3.12.10 (tags/v3.12.10:0cc8128, Apr 8 2025, 12:21:36) [MSC v.1943 64 bit (AMD64)]
ComfyUI version: 0.3.60
ComfyUI frontend version: 1.26.13
[Prompt Server] web root: C:\nginx-1.27.4\html\ComfyUI_windows_portable\python_embeded\Lib\site-packages\comfyui_frontend_package\static
ComfyUI-FreeMemory nodes loaded: Image, Latent, Model, CLIP, and String versions available.
ComfyUI-GGUF: Partial torch compile only, consider updating pytorch
### Loading: ComfyUI-Impact-Pack (V8.22.3)
[Impact Pack] Wildcards loading done.
### Loading: ComfyUI-Manager (V3.37)
[ComfyUI-Manager] network_mode: public
### ComfyUI Revision: 640 [b8730510] *DETACHED | Released on '2025-09-23'

Initializing ControlAltAI Nodes

[rgthree-comfy] Loaded 48 fantastic nodes. 🎉

Import times for custom nodes:
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\websocket_image_save.py
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\janus-pro
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-FreeMemory
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\controlaltai-nodes
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-KJNodes
0.1 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy
0.6 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager
0.8 seconds: C:\nginx-1.27.4\html\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack

[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
Context impl SQLiteImpl.
Will assume non-transactional DDL.
No target revision found.
Starting server

To see the GUI go to: http://127.0.0.1:8190
FETCH ComfyRegistry Data: 5/98
FETCH ComfyRegistry Data: 10/98
FETCH ComfyRegistry Data: 15/98
FETCH ComfyRegistry Data: 20/98
FETCH ComfyRegistry Data: 25/98
FETCH ComfyRegistry Data: 30/98
FETCH ComfyRegistry Data: 35/98
FETCH ComfyRegistry Data: 40/98
FETCH ComfyRegistry Data: 45/98
FETCH ComfyRegistry Data: 50/98
FETCH ComfyRegistry Data: 55/98
FETCH ComfyRegistry Data: 60/98
FETCH ComfyRegistry Data: 65/98
FETCH ComfyRegistry Data: 70/98
FETCH ComfyRegistry Data: 75/98
FETCH ComfyRegistry Data: 80/98
FETCH ComfyRegistry Data: 85/98
FETCH ComfyRegistry Data: 90/98
FETCH ComfyRegistry Data: 95/98
FETCH ComfyRegistry Data [DONE]
[ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.float32
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load SD3ClipModel_
loaded completely 22987.925 6102.49609375 True
gguf qtypes: F16 (923)
model weight dtype torch.float16, manual cast: None
model_type FLOW
Requested to load SD3
got prompt
loaded completely 17105.97333984375 15366.797485351562 True
100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:20<00:00, 5.07s/it]
Requested to load AutoencodingEngine
loaded completely 402.0546875 319.7467155456543 True
Prompt executed in 407.59 seconds
loaded completely 22463.546978759765 15366.797485351562 True
100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:21<00:00, 5.43s/it]
Prompt executed in 25.88 seconds

RTX 2070 GPU

[START] Security scan
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-09-28 23:22:03.999
** Platform: Windows
** Python version: 3.12.7 (tags/v3.12.7:0b05ead, Oct 1 2024, 03:06:41) [MSC v.1941 64 bit (AMD64)]
** Python executable: C:\nginx-1.27.4\html\ComfyUI\python_embeded\python.exe
** ComfyUI Path: C:\nginx-1.27.4\html\ComfyUI\ComfyUI
** ComfyUI Base Folder Path: C:\nginx-1.27.4\html\ComfyUI\ComfyUI
** User directory: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\user
** ComfyUI-Manager config path: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\user\default\ComfyUI-Manager\config.ini
** Log path: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\user\comfyui.log

Prestartup times for custom nodes:
49.2 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\ComfyUI-Manager

Set cuda device to: 1
Checkpoint files will always be loaded safely.
Total VRAM 8192 MB, total RAM 65464 MB
pytorch version: 2.6.0+cu124
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 2070 : cudaMallocAsync
Using pytorch attention
Python version: 3.12.7 (tags/v3.12.7:0b05ead, Oct 1 2024, 03:06:41) [MSC v.1941 64 bit (AMD64)]
ComfyUI version: 0.3.60
ComfyUI frontend version: 1.26.13
[Prompt Server] web root: C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\comfyui_frontend_package\static
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_url" in Model has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_name" in KlingImageGenerationsRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_name" in KlingVirtualTryOnRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_params" in MoonvalleyPromptResponse has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_name" in KlingDualCharacterEffectInput has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_name" in KlingImage2VideoRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_name" in KlingSingleImageEffectInput has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_name" in KlingText2VideoRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_version" in TripoTextToModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_seed" in TripoTextToModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_version" in TripoImageToModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_seed" in TripoImageToModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_version" in TripoMultiviewToModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_seed" in TripoMultiviewToModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\pydantic\_internal\_fields.py:132: UserWarning: Field "model_seed" in TripoTextureModelRequest has conflict with protected namespace "model_".

You may be able to resolve this warning by setting `model_config['protected_namespaces'] = ()`.
warnings.warn(
ComfyUI-GGUF: Partial torch compile only, consider updating pytorch
C:\nginx-1.27.4\html\ComfyUI\python_embeded\Lib\site-packages\timm\models\layers\__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning)
### Loading: ComfyUI-Manager (V3.37)
[ComfyUI-Manager] network_mode: public
### ComfyUI Revision: 3951 [b8730510] *DETACHED | Released on '2025-09-23'
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
WAS Node Suite: OpenCV Python FFMPEG support is enabled
WAS Node Suite Warning: `ffmpeg_bin_path` is not set in `C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\was-node-suite-comfyui\was_suite_config.json` config file. Will attempt to use system ffmpeg binaries if available.
WAS Node Suite: Finished. Loaded 220 nodes successfully.

"Do one thing every day that scares you." - Eleanor Roosevelt

Import times for custom nodes:
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\SDXLAspectRatio.py
0.0 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\websocket_image_save.py
0.1 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\ComfyUI-GGUF
0.6 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\ComfyUI-Manager
3.1 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\ComfyUI-Inspyrenet-Rembg
3.5 seconds: C:\nginx-1.27.4\html\ComfyUI\ComfyUI\custom_nodes\was-node-suite-comfyui

FETCH ComfyRegistry Data: 5/98
Context impl SQLiteImpl.
Will assume non-transactional DDL.
No target revision found.
Starting server

To see the GUI go to: http://127.0.0.1:8188
FETCH ComfyRegistry Data: 10/98
FETCH ComfyRegistry Data: 15/98
FETCH ComfyRegistry Data: 20/98
FETCH ComfyRegistry Data: 25/98
FETCH ComfyRegistry Data: 30/98
FETCH ComfyRegistry Data: 35/98
FETCH ComfyRegistry Data: 40/98
FETCH ComfyRegistry Data: 45/98
FETCH ComfyRegistry Data: 50/98
FETCH ComfyRegistry Data: 55/98
FETCH ComfyRegistry Data: 60/98
FETCH ComfyRegistry Data: 65/98
FETCH ComfyRegistry Data: 70/98
FETCH ComfyRegistry Data: 75/98
FETCH ComfyRegistry Data: 80/98
FETCH ComfyRegistry Data: 85/98
FETCH ComfyRegistry Data: 90/98
FETCH ComfyRegistry Data: 95/98
FETCH ComfyRegistry Data [DONE]
[ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
got prompt
model weight dtype torch.float16, manual cast: None
model_type EPS
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.float32
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load SDXLClipModel
loaded completely 5714.8 1560.802734375 True
Requested to load SDXL
loaded completely 5389.596817779541 4897.0483474731445 True
100%|██████████████████████████████████████████████████████████████████████████████████| 12/12 [00:07<00:00, 1.62it/s]
Requested to load AutoencoderKL
0 models unloaded.
loaded partially 128.0 127.99989318847656 0
Prompt executed in 279.03 seconds
```

### Other

I don't seem to have these issues that I remember when running from the GUI. I think these issues are happening in the API.

Everything works fine EXCEPT this extreme slowdown. I run a website and it's making it unusable!
I have been having this issue back and forth for atleast 2 years!
When is it going to be finally fixed for good??? Running updates never seems to help.

You also need to update your insturctions on disabling custom nodes!! the interface no longer looks like that!

For the RTX 2070, I am only using the GGUF,Manager, and Inspyrenet Rembg which all appear to be working fine.
The GGUF and Rembg nodes are required for my site to run correctly.
I did not remove any custom_nodes but it seems to have sped up after changes to the P40.

For the P40, I was using some custom nodes that I have now removed and it seems to be fast again for the moment but it never seems to stay that way. Happening for years.

Here are the nodes I removed that seemed to speed it up. Maybe it was the 2-3 restarts though.

ComfyUI-FreeMemory
ComfyUI-Impact-Pack
ComfyUI-KJNodes
controlaltai-nodes
janus-pro
rgthree-comfy

In case one of these is causing these massive issues for both cards...
Maybe this will help you track down the problem. I still feel like there is a bug in the API though.