The error "No CUDA detected" means the software cannot find a usable CUDA environment. This typically occurs due to one of the following reasons:
- No NVIDIA GPU – CUDA acceleration requires an NVIDIA graphics card. If you have an AMD or Intel GPU (or no dedicated GPU), CUDA is unavailable. Use CPU mode instead (uncheck "CUDA acceleration" in the main interface).
- Outdated or missing NVIDIA driver – Even if you have an NVIDIA GPU, the driver must support CUDA 12.8 or later. Open Command Prompt and run
nvidia-smi. Check the "CUDA Version" shown at the top. It should be ≥ 12.8. If not, update your driver from NVIDIA's website. CUDA Toolkit and cuDNN not installed – In addition to the driver, you need:
- CUDA Toolkit 12.8+ (verify with
nvcc -V) - cuDNN 9.11+ (ensure the
bin and lib directories are added to the system PATH environment variable)
- Environment variables missing – After installing CUDA, the
PATH must include C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8\bin and C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8\lib. Restart the software after setting them.
Note: Windows 10 build 18363 is quite old – ensure you have the latest NVIDIA driver that supports CUDA 12.8 (driver version ≥ 531.61 for CUDA 12.8).
If you have an NVIDIA GPU but still see this error, please provide:
- Output of
nvidia-smi - Output of
nvcc -V - The exact content of
PATH environment variable (relevant to CUDA)
Please check the documentation: