The ever-growing machine-to-machine traffic in data centers has stimulated the increase of transceiver data rate from 25 Gb/s/λ to 100 Gb/s/λ and beyond. It is believed that advanced modulation formats and digital-to-analog converters (DACs) will be employed in next generation short-reach transceivers. Digital pre-emphasis techniques are widely employed in DAC-based transceivers to compensate for the high frequency roll-off due to the RF and optoelectronics components in optical transceivers. However, digital pre-emphasis essentially reduces the magnitude of the signal low frequency components for a flat frequency response, which unavoidably increases quantization error, reducing the overall signal-to-noise ratio. This trade-off between SNR and bandwidth conflicts with the high SNR requirement of advanced modulation formats such as the Nyquist-shaped pulse amplitude modulation (PAM). To mitigate the quantization error induced SNR degradation, we show that analog pre-emphasis filters can be used in conjunction with digital pre-emphasis for improved system performance. To understand the impact of the analog pre-emphasis filter on system performance, we analyze the relationship between the flatness of the system frequency response and the SNR degradation due to digital pre-emphasis, and demonstrate 1.1-dB increase of receiver sensitivities, for both 64-Gb/s and 128-Gb/s intensity-modulation direct detection (IM-DD) PAM4 signals, respectively employing a directly modulated laser (DML) and an electroabsorption modulator (EAM).