They calculated the difference in energy consumption but failed to factor in the orders-of-magnitude difference of NAND performance via HDDs. To wit, an HDD will spend much more if its time at peak power consumption during I/O vs SSDs, by function of how much longer it takes complete the I/O. And that doesn't even factor in the additional power consumption from keeping the system running longer to complete I/O-bound workloads for a given user task.