Okay, it is difficult to tell from that screenshot, but I would explain it as follows:
The NIC only has one queue in which it receives (and probably sends packets). That means only one processor core deals with incoming and outgoing packets while the other one is idle.
This is common on hardware that is designed for a different market (in your case an ultra-low power processor for the mobile market which normally do not come with many PCIe lanes anyways) or cheap hardware. So the single-core performance is the bottleneck of this machine.
Our appliances from LWL load-balance traffic across all processor cores so that any additional core improves network throughput: