Why did you not expect this? GPU cards (where OpenCL i.e AMD is concerned) will always be faster and/or more efficient than CPUs in any given Proof-of-Work algorithm.
The CryptoNight algorithm was designed to be GPU resistant, but this means nothing once GPU miners are actually released.
Scrypt was meant to be GPU-resistant as well, and there are ASICs now.
Ryzen is actually proving to be competative on efficiency for XMR vs any GPU (possible exception for the GTX 750 ti, but that thing seems to have hit THE sweet spot on XMR).
Some of the more modern Intel CPUs are pretty efficient too.
pre-Ryzen AMD CPUs though are definitely NOT XMR efficient by current standards, nor are pre-Maxwell NVidia GPUs or pre-RX AMD GPUs (possible exception for the Fury/FuryX/Nano?).
OP, keep in mind that your particular I7 model is a few generations old now, and at LEAST one if not 2 process nodes behind current state-of-the-art - of COURSE it is less efficient than stuff running on a CURRENT process node.