The CUDA SDK is not installed correctly. One has to understand that there's a difference between:
- NVIDIA CUDA library
- NVIDIA RTC library
The NVIDIA CUDA library comes with the CUDA SDK, but also with the NVIDIA Driver.
The NVIDIA RTC library comes with the CUDA SDK alone. If the install is not completed correctly, hashcat can't use CUDA. Maybe you had unticked it during installation options menu.
Hashcat felt back to OpenCL because the CUDA SDK is not installed correctly, this is why you see this:
OpenCL API (OpenCL 3.0 CUDA 11.4.56)
Because the NV Driver also ships with an OpenCL API. The NV OpenCL API replaces the NVRTC because it ships with it's on JiT compiler. This JiT compiler creates the PTX code similar to NVRTC does. It's good enough for the CUDA library to take over after JiT compile phase, which is why in this case NV CUDA library is enough and hashcat can start.
However, using CUDA specific architecture the NVRTC can compile more efficient code than the OpenCL JiT, which is why you should stick to the CUDA NVRTC. So you need to fix your CUDA SDK installation.
- NVIDIA CUDA library
- NVIDIA RTC library
The NVIDIA CUDA library comes with the CUDA SDK, but also with the NVIDIA Driver.
The NVIDIA RTC library comes with the CUDA SDK alone. If the install is not completed correctly, hashcat can't use CUDA. Maybe you had unticked it during installation options menu.
Hashcat felt back to OpenCL because the CUDA SDK is not installed correctly, this is why you see this:
OpenCL API (OpenCL 3.0 CUDA 11.4.56)
Because the NV Driver also ships with an OpenCL API. The NV OpenCL API replaces the NVRTC because it ships with it's on JiT compiler. This JiT compiler creates the PTX code similar to NVRTC does. It's good enough for the CUDA library to take over after JiT compile phase, which is why in this case NV CUDA library is enough and hashcat can start.
However, using CUDA specific architecture the NVRTC can compile more efficient code than the OpenCL JiT, which is why you should stick to the CUDA NVRTC. So you need to fix your CUDA SDK installation.