[Feature Request]: # RTX 5060 (sm_120) Not Supported - CUDA Compatibility Issue

[paladryenn11-alt]

Is there an existing issue for this?

• [x] I have searched the existing issues and checked the recent builds/commits

What would your feature do?

System Information

• GPU: NVIDIA GeForce RTX 5060
• CUDA Capability: sm_120 (Blackwell architecture)
• Driver Version: 581.57
• OS: Windows 11
• Fooocus Version: 2.5.0 (win64)

Problem Description

Fooocus fails to start on RTX 5060 with the following error:

RuntimeError: CUDA error: no kernel image is available for execution on the device

Full error message:

NVIDIA GeForce RTX 5060 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90 compute_37.

Proposed workflow

1. Go to ....
2. Press ....
3. ...

Additional information

What I Tried

1. ✅ Updated to latest NVIDIA driver (581.57)
2. ✅ Downloaded official Fooocus 2.5.0 from releases
3. ✅ Tried switching to CUDA 11 version (previous_old_xformers_env.7z)
4. ❌ Still getting the same compatibility error

Root Cause

The RTX 50-series GPUs (including RTX 5060) use the new Blackwell architecture with sm_120 compute capability, which is not yet supported by the current PyTorch version bundled with Fooocus.

Request

Could you please:

1. Update PyTorch to a version that supports sm_120 architecture
2. Add RTX 50-series GPU support to Fooocus
3. Update the troubleshooting documentation with information about RTX 50-series compatibility

Additional Context

The RTX 5060 was released in early 2025 and is becoming more common. This issue will likely affect many users upgrading to RTX 50-series cards.

Workaround Status

Currently, there is no working workaround for RTX 50-series users. The application cannot run on these GPUs.

---

Thank you for your excellent work on Fooocus! I'm looking forward to using it once RTX 50-series support is added.

[zevnda]

Fixed for me by using pytorch nightly build 2.10.0. Go to the directory where you unpacked Fooocus and run

pip install --upgrade --force-reinstall --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128