This paper proposes a direct data-driven design approach for multivariable control of rotational speed and torque dynamics in a dynamic internal combustion engine test bench. The controller parameters are directly identified from data using a suitably extended version of the algorithm introduced in [1], allowing closed-loop experiments to be employed. The proposed controller is designed and validated using an accurate simulator of the internal combustion engine test bench on which a comparison with other model-based control strategies is also shown.