Hype trains are always a two-sided coin when it comes to fans on both sides of the aisle. Remember the Pentium 4 hype and more recently the Intel Atom Bay Trail and Cherry Trail hype? As it is, Ryzen desktop processors are between Broadwell and Kaby Lake which equates to the claimed 52% IPC (and single-thread) improvement over Bulldozer. This is nothing to be sneezed at no matter how you try cutting it.
Again, CPU power consumption is approximately a cubic function of the voltage you see (link). The lower slope means a huge savings in power draw for the lower frequencies. The higher slope means a huge uptick in power draw for the higher frequencies. Numerically, this means for half of the voltage, we will see about an eighth of the power draw without any additional transistor-level optimization. Better said, for a quarter drop in the voltage, we will approximately half the power draw without any additional transistor-level optimization. In addition, halving the number of cores will halve the power draw, where the processor in this chart is an eight-core configuration. Based on the model in the chart, since the voltage has increased substantially for the higher frequency processors, naturally, following the chart downwards in frequency, we should see far lower power draw given the lower slope and, in turn, lower power levels for the lower frequencies.
To reiterate, the second chart has nothing to do with frequency and per-core performance. Period. I do not know where you are extrapolating this from, but IPC (instructions per clock) does not change with frequency rate. IPC is a constant value, or, in other words, for a given unit of frequency, you get a given amount of performance throughput. As to the chart, all it is saying is the lowest stable voltage an eight-core Ryzen processor can be supplied by for a given clock frequency. This is to say, it is telling the lower the stable voltage, the lower the power draw. Nothing more, nothing less.
Now, the calculations I made were based solely on the 52% IPC improvement. This is why I added the qualifier "and possibly more." In other words, this is without also taking into account Ryzen's improved power efficiency and any additional transistor-level optimizations that will inevitably be made for the mobile versions of the architecture. These upcoming Raven Ridge APUs will be able to reach higher clocks as well at the same power level as Bristol Ridge. Even a meager 15% increase in clock rate compared to that Bristol Ridge processor, thanks to increased efficiency and transistor-level optimization on the top of the already 52% IPC improvement, would mean an overall single-core performance improvement of 60%. That would equate to a single-thread Cinebench score of 120 using the 76 reference score of the current Bristol Ridge processor. Note that this is a conservative estimate, meaning more performance is also possible.