I would ass-u-me that this would mean that over a period of time X, a current generation chip would process Y commands consuming N units of energy.
The new chip would perform 2Y commands over X time while only consuming
Or that each command execution would take 80% of the energy of a current gen chip, but that it could complete twice as many of them in the same time period, meaning a net increase of ~60% energy consumption at sustained max load.
Tons of ways to play with the statistics on this one, and the 100% performance improvement and 20% energy efficiency improvement are not mutually exclusive. But the summary doesn't give any context or detail, so without RTFA, it should be considered nothing more than marketing speak.