Employing currently available quantum technology, we design and implement a nonclassically correlated SWAP heat engine that allows to achieve an efficiency above the standard Carnot limit. Such an engine also boosts the amount of extractable work, in a wider parameter window, with respect to engine's cycle in the absence of initial quantum correlations in the working substance. The boosted efficiency arises from a trade-off between the entropy production and the consumption of quantum correlations during the full thermodynamic cycle. We derive a generalized second-law limit for the correlated cycle and implement a proof-of-principle demonstration of the engine efficiency enhancement by effectively tailoring the thermal engine on a cloud quantum processor.