13700cpu 32g 1g 7900xtx

[yongjun liu]

@terry ## System Info
OS: linux
Python Version: 3.12.3 (main, Mar 23 2026, 19:04:32) [GCC 13.3.0]
Embedded Python: false
Pytorch Version: 2.5.1+rocm6.2
Arguments: main.py --listen 127.0.0.1 --port 8188
RAM Total: 31.08 GB
RAM Free: 27.25 GB
Templates Version: 0.9.73

Devices

• cuda:0 AMD Radeon RX 7900 XTX : native (cuda)
  VRAM Total: 23.98 GB
  VRAM Free: 23.88 GB
  Torch VRAM Total: 0 B
  Torch VRAM Free: 0 B

[terry]

@yongjun-liu 可以，开始实践吧！如果6.2的包有问题，你可以换6.3，更好一点，这是我测试过的稳定版，跑comfyUI非常稳。驱动我是Rocm7.2。

[yongjun liu]

@terry 是不是我这个模型用的不对？我就用的官方带的自带的那个，现在运行起来巨慢，一张图还没出过呢。

[terry]

@yongjun-liu 和模型无关，你贴下日志，肯定是降级到CPU了。

[yongjun liu]

setup plugin alembic.autogenerate.schemas
setup plugin alembic.autogenerate.tables
setup plugin alembic.autogenerate.types
setup plugin alembic.autogenerate.constraints
setup plugin alembic.autogenerate.defaults
setup plugin alembic.autogenerate.comments
Found comfy_kitchen backend cuda: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
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']}
Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
Checkpoint files will always be loaded safely.
Total VRAM 24560 MB, total RAM 31825 MB
pytorch version: 2.5.1+rocm6.2
Set: torch.backends.cudnn.enabled = False for better AMD performance.
AMD arch: gfx1100
ROCm version: (6, 2)
Set vram state to: NORMAL_VRAM
Device: cuda:0 AMD Radeon RX 7900 XTX : native
Using async weight offloading with 2 streams
Enabled pinned memory 28642.0
Using sub quadratic optimization for attention, if you have memory or speed issues try using: --use-split-cross-attention
Python version: 3.12.3 (main, Mar 23 2026, 19:04:32) [GCC 13.3.0]
ComfyUI version: 0.21.0
comfy-aimdo version: 0.3.0
comfy-kitchen version: 0.2.8
ComfyUI frontend version: 1.43.18
[Prompt Server] web root: /home/liuyong/ai/ComfyUI/venv/lib/python3.12/site-packages/comfyui_frontend_package/static
Asset seeder disabled

Import times for custom nodes:
0.0 seconds: /home/liuyong/ai/ComfyUI/custom_nodes/websocket_image_save.py

Context impl SQLiteImpl.
Will assume non-transactional DDL.
Starting server

To see the GUI go to: http://127.0.0.1:8188
got prompt
VAE load device: cuda:0, offload device: cpu, dtype: torch.float32
Found quantization metadata version 1
Detected mixed precision quantization
Using mixed precision operations
model weight dtype torch.bfloat16, manual cast: torch.bfloat16
model_type FLUX
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
no CLIP/text encoder weights in checkpoint, the text encoder model will not be loaded.
Requested to load VideoVAE
loaded completely; 23236.80 MB usable, 1384.94 MB loaded, full load: True
/home/liuyong/ai/ComfyUI/venv/lib/python3.12/site-packages/torch/nn/modules/conv.py:720: UserWarning: Attempting to use hipBLASLt on an unsupported architecture! Overriding blas backend to hipblas (Triggered internally at ../aten/src/ATen/Context.cpp:296.)
return F.conv3d(
Found quantization metadata version 1
Using MixedPrecisionOps for text encoder
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load LTXAVTEModel_
loaded completely; 20540.44 MB usable, 11201.91 MB loaded, full load: True
Requested to load LTXAV

[terry]

CLIP启用了CPU，你发日志给Gemini，它会教你怎么做的。就是你不要设置lowvram模式就好。

[yongjun liu]

@terry 他说是加载了 Nvidia 的程序。

[yongjun liu]

@terry 跑 7% 就卡死了。

[terry]

我都给你标注了，它Current，就是当前设备时CPU，你去骂Gemini煞笔。

[williamlouis]

@terry 学习到了。驾驭 G mini大法了。
这是核心教程啊。

[yongjun liu]

@terry 他回答，CPU 正常，没问题。说是步数问题，操，指挥了半天，给我脑子指挥晕了。我休息两天吧。

[yongjun liu]

我重装了Windows 11。comfyui 结果是这个,晕死我了.

[yongjun liu]

这是我新装系统的启动日志啊
Adding extra search path custom_nodes C:\Users\cnbjy\Documents\ComfyUI\custom_nodes
Adding extra search path download_model_base C:\Users\cnbjy\Documents\ComfyUI\models
Adding extra search path custom_nodes C:\Users\cnbjy\AppData\Local\Programs\ComfyUI\resources\ComfyUI\custom_nodes
Setting output directory to: C:\Users\cnbjy\Documents\ComfyUI\output
Setting input directory to: C:\Users\cnbjy\Documents\ComfyUI\input
Setting user directory to: C:\Users\cnbjy\Documents\ComfyUI\user
[START] Security scan
[DONE] Security scan
** ComfyUI startup time: 2026-05-14 20:40:00.306
** Platform: Windows
** Python version: 3.12.11 (main, Aug 18 2025, 19:17:54) [MSC v.1944 64 bit (AMD64)]
** Python executable: C:\Users\cnbjy\Documents\ComfyUI.venv\Scripts\python.exe
** ComfyUI Path: C:\Users\cnbjy\AppData\Local\Programs\ComfyUI\resources\ComfyUI
** ComfyUI Base Folder Path: C:\Users\cnbjy\AppData\Local\Programs\ComfyUI\resources\ComfyUI
** User directory: C:\Users\cnbjy\Documents\ComfyUI\user
** ComfyUI-Manager config path: C:\Users\cnbjy\Documents\ComfyUI\user__manager\config.ini
** Log path: C:\Users\cnbjy\Documents\ComfyUI\user\comfyui.log
[ComfyUI-Manager] Skipped fixing the 'comfyui-frontend-package' dependency because the ComfyUI is outdated.
[PRE] ComfyUI-Manager
Found comfy_kitchen backend cuda: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4']}
Found comfy_kitchen backend triton: {'available': False, 'disabled': True, 'unavailable_reason': "ImportError: No module named 'triton'", 'capabilities': []}
Checkpoint files will always be loaded safely.
Total VRAM 24560 MB, total RAM 32506 MB
pytorch version: 2.9.1+rocm7.2.1
Set: torch.backends.cudnn.enabled = False for better AMD performance.
AMD arch: gfx1100
ROCm version: (7, 2)
Set vram state to: NORMAL_VRAM
Device: cuda:0 AMD Radeon RX 7900 XTX : native
Using async weight offloading with 2 streams
Enabled pinned memory 13002.0
Using pytorch attention
Python version: 3.12.11 (main, Aug 18 2025, 19:17:54) [MSC v.1944 64 bit (AMD64)]
ComfyUI version: 0.21.1
comfy-aimdo version: 0.3.0
comfy-kitchen version: 0.2.8
[Prompt Server] web root: C:\Users\cnbjy\AppData\Local\Programs\ComfyUI\resources\ComfyUI\web_custom_versions\desktop_app
Asset seeder disabled
[START] ComfyUI-Manager
[ComfyUI-Manager] Using GitPython backend
[ComfyUI-Manager] network_mode: public
[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:
C:\Users\cnbjy\Documents\ComfyUI.venv\Scripts\python.exe -m pip install matrix-nio

Import times for custom nodes:
0.0 seconds: C:\Users\cnbjy\AppData\Local\Programs\ComfyUI\resources\ComfyUI\custom_nodes\websocket_image_save.py

Context impl SQLiteImpl.
Will assume non-transactional DDL.
Starting server

To see the GUI go to: http://127.0.0.1:8000

[terry]

不至于吧，我用xtx跑过啊，官方的包速度挺不错啊，你用的rocm7.2不至于慢。看你截图，你用的官方fp8模型，你想多了，你的卡跑不了这个模型，去换成社区的Q4量化的GGUF模型。

[yongjun liu]

@terry 刚才跑通了一段就用的是这个配置三秒钟720的视频用了13分钟.跑的鼠标都动不了。说的那个Q4的那个模型在哪啊您给个链接我去看看。多谢了啊哥们儿。

[terry]

@yongjun-liu 大哥都说了你模型太大，用CPU计算了，没用显卡，卡的和锤子一样。你问AI去哪里下，不行去B站下载刘悦的整合包，它包里就带模型。b站，刘悦的技术博客。还不会就直接把电脑格式化，你不适合玩AI。

[williamlouis]

@terry 顶。我也不喜欢解释。浪费时间。