Hardware acceleration support for Blender is enabled if a CUDA toolkit is detected when Blender is being built. The CUDA toolkit is no longer required after Blender has been built; it is only required during building. The CUDA toolkit SlackBuilds which have been tested and are known to work are: development/cudatoolkit_12 development/cudatoolkit_13. The main difference between the toolkits is the collection of compute capabilities that each one supports. This, in turn, determines the graphics cards able to be supported by the final Blender package. The relationship between compute capability and various graphics cards can be seen at: https://developer.nvidia.com/cuda/gpus and https://developer.nvidia.com/cuda/gpus/legacy With one of the CUDA toolkits installed, the range of supported architectures can be displayed by executing the command: nvcc --list-gpu-arch The range of architectures (hence compute capabilities) supported by Blender itself may not support all those that are supported by a particular version toolkit. For instance, when building blender-5.1.2 with cudatoolkit_13, terminal output will show lines like: CUDA binaries for sm_50 are no longer supported with CUDA 13.0+ with similar lines for sm_52, sm_60, sm_61 and sm_70. This means that compute capabilities (CC's) 5.0, 5.2, 6.0, 6.1 and 7.0 are not suppported. If your graphics card happens to fall into this category, build Blender with cudatoolkit_12 which still supports these older architectures. More information: https://developer.nvidia.com/blog/navigating-gpu-architecture-support-a-guide-for-nvidia-cuda-developers/ cudatoolkit_12: nvcc --list-gpu-arch compute_50 compute_52 compute_53 compute_60 compute_61 compute_62 compute_70 compute_72 compute_75 compute_80 compute_86 compute_87 compute_89 compute_90 compute_100 compute_101 compute_103 compute_120 compute_121 cudatoolkit_13: nvcc --list-gpu-arch compute_75 compute_80 compute_86 compute_87 compute_88 compute_89 compute_90 compute_100 compute_110 compute_103 compute_120 compute_121