The frequency estimation performance of several modified discrete cosine transform (MDCT) domain algorithms is investigated for single real sinusoidal signal in white Gaussian noise. Two discrete Fourier transform (DFT) domain algorithms are included to make comparison. It is demonstrated that most of the MDCT domain algorithms performs alike when the signal-to-noise ratio (SNR) is low. Only at sufficiently high SNR, these algorithms show their own characteristics. But compared to the DFT algorithms and the Cramer-Rao bound (CRB), there should be opportunity to develop a better estimator in the MDCT domain.