While they are pretty impressive little machines, I would take those “benchmarks” with an ocean of salt. Comparing local metal to virtual servers is really not reasonable. There are too many factors that can cause impact on virtual server performance to make a meaningful benchmark on one. The first of which is the virtualization itself which kills a lot of performance. Further, system resource access is crippled both by the FACT of virtualization, as well as contention for the resources with other users of the host machine.
To suggest that…
Well, I’ve had my hands in physical servers for decades, and I’ll tell you straight up that a real metal server from more than a DECADE ago, will still outperform these in terms of number crunching. Maybe not disk I/O, because back then everything was still spinny magnets, but certainly in processing.
I actually still have a few servers in that age range, in fact. All upgraded with SSD’s now. They work marvelously still, and carry continuous loads that you simply could NOT do with an rk3399.
Still though, the rk3399 is impressive. For what it is.