Active noise control (ANC) technique has been successfully implemented by various automotive manufacturers to interior powertrain noise cancellation and/or sound management. For the powertrain noise control, the ANC system only requires one reference signal (typically using tachometer signal) due to the fact that the powertrain response is functionally related to the engine rotational speed. Instead, the multi-reference ANC system is often required for the road noise control since the multiple-coherence between the reference accelerometer signals and the target noise must be guaranteed to yield an acceptable control performance. However, the adoption of traditional multi-reference filtered-x least mean square (FXLMS) algorithm for road noise ANC often leads to a large computational burden and slow convergence rate due to the long length of adaptive filter and a huge dynamic range of the filtered reference signal. To address these problems, road noise ANC systems configured with the time-frequency domain FXLMS (TF-FXLMS) algorithm and subband FXLMS (SFXLMS) algorithm are developed, where the adaptive algorithm either calculates the gradient estimate and filtered reference signal in the frequency domain or subband frequency domain to reduce the computational requirement and also updates the control signals in the time domain to avoid delay in the reference path. Then, a comprehensive computational complexity analysis is conducted to exhibit the computational cost for different algorithms. The effectiveness of the different algorithms will be demonstrated through an ANC system with eight accelerometers, one secondary loudspeaker, and one error microphone in the driver side to control actual road noise.