The catch, which AMD are quick to gloss over, is that Caffes performance on Nvidia hardware largely comes from its use of Nvidia's proprietary cuDNN kernels. A

The catch, which AMD are quick to gloss over, is that Caffes performance on Nvidia hardware largely comes from its use of Nvidia's proprietary cuDNN kernels. AMDs HIP port is using the "fallback" open-source CUDA kernels, which are nowhere near as fast as the hand-optimized cuDNN code.