A newly reconstructed global sea surface temperature (SST) analysis dataset (CMA-SST), available on monthly 2º×2º resolutions since 1900, has been produced in this study. It is constructed from a newly developed integrated dataset with denser and wider sampling of in situ SST observations, and follows similar analyzing techniques to ERSST.v5. Assessments show that the larger observation quantity of the input data source is beneficial to making the reconstructed SSTs more realistic, especially in the China's offshore sea area. Apart from that, a specific parameter for bias correction is upgraded to be self-adaptive to the input data source, and plays as a mediator to improve the accuracy of the reconstructed SSTs. Generally, the reconstructed CMA-SST is comparable to the currently congeneric products. Its biases are approximate to those of ERSST.v5, COBE-SST2, HadISST2 and HadSST3, especially closest to ERSST.v5 and lower than HadISST2 and HadSST3 at the high latitudes of the Southern Hemisphere with limited in situ observations. Besides, its temporal characteristics, such as year-to-year variations of globally averaged SST anomalies and time-series of Niño3.4 and Atlantic Multidecadal Oscillation indexes are also identical to those of congeneric products. Although the warming rates of the data are a little bit higher in many regions over 1900-2019 and 1950-2019, it is proven to be sensible and within the quantified uncertainties of ERSST.v5. However, the noticeable differences of the strength and stability of spatial standard deviations among various datasets, and low correlations between CMA-SST and the other products around 60ºS with very limited in situ sampling necessitate further investigation and improvement of CMA-SST.