About the compatibility of NVIDIA TESLA M60 关于英伟达特斯拉M60的兼容性

[htppp254]

My computer is using an Intel Core i3-9100F and NVIDIA TESLA M60. Its driver is 539.72, CUDA 12.2, Maxwell architecture. It's unable to generate images in Comfy UI. 我的电脑使用的是Intel core i3-9100f+NVIDIA TESLA M60 。它的驱动是539.72，CUDA12.2，Maxwell架构。它在Comfy UI中无法生成图片。

以下是日志
上传至Github的日志.txt

[vavo]

The attached traceback fails at the first CUDA allocation with CUDA error: operation not supported, so this is belov the workflow/model layer.

The Tesla M60 is Maxwell (sm_52). Recent PyTorch CUDA 12.8/13.x binaries do not include Maxwell kernels; CUDA 13 supports Turing (sm_75) and newer. That combination produces exactly this kind of failure even though Windows and nvidia-smi can see the GPU.

Check the environment ComfyUI is actually using:

 .venv\Scripts\python.exe -c "import torch; print(torch.__version__, torch.version.cuda); print(torch.cuda.get_device_capability()); print(torch.cuda.get_arch_list())"

If the device reports (5, 2) but get_arch_list() does not include sm_52, that PyTorch build cannot execute on the M60. The practical options are to use a PyTorch CUDA 12.6 build that still contains Maxwell support, with a compatible NVIDIA driverr.

Do not keep changing checkpoints; the crash occurs before a checkpoint can finish loading.