Smart vehicles require constantly running heavyvehicular computations with their limited computation/energyresources. 5G vehicular networks have potential to resolve theissue, by letting the vehicular tasks offloaded to 5G mobile edgecomputing (MEC) servers. To better support vehicular computa-tion offloading, this paper proposes a road-side 5G infrastructureconsisting of multiple millimeter-wave (mmWave) small-cell basestations (BSs) and a cellular mid-band based macro-cell BS whereeach BS is equipped with an MEC server. Then, the vehicleswith mmWave/mid-band dual interfaces can decide which BSto choose for offloading. We propose a decentralized offloadingdecision mechanism where each vehicle tries to minimize thetime-energy joint cost with three choices: local computing,offloading to a small-cell MEC, offloading to a macro-cell MEC. In particular, we model the problem as an ordinal potentialgame, derive its potential function to ensure the existence ofand finite-time convergence to a Nash equilibrium (NE), analyzeits Price-of-Anarchy, and develop an iterative offloading decisionupdate algorithm. In doing so, we also consider slicing the globalgame into multiple non-overlapping smaller games and runningthem in parallel, to investigate the best slicing strategy. Ourextensive simulations show the game’s real-time convergence toan NE, reveal the NE’s near-optimal performance, and presentthe efficacy of the proposed game slicing.