Studies of the correlation between different diffuse interstellar bands (DIBs) are important for exploring their origins. However, the Gaia-RVS spectral window between 846 and 870nm contains few DIBs, the strong DIB at 862nm being the only convincingly confirmed one. Here we attempt to confirm the existence of a broad DIB around 864.8nm and estimate its characteristics using the stacked Gaia-RVS spectra of a large number of stars. We study the correlations between the two DIBs at 862nm ({lambda}862) and 864.8nm ({lambda}864.8), as well as the interstellar extinction. We obtained spectra of the interstellar medium (ISM) absorption by subtracting the stellar components using templates constructed from real spectra at high Galactic latitudes with low extinctions. We then stacked the ISM spectra in Galactic coordinates (l, b) - pixelized by the HEALPix scheme - to measure the DIBs. The stacked spectrum is modeled by the profiles of the two DIBs, Gaussian for {lambda}862 and Lorentzian for {lambda}864.8, and a linear continuum. We report the fitted central depth (CD), central wavelength, equivalent width (EW), and their uncertainties for the two DIBs. We obtain 8458 stacked spectra in total, of which 1103 (13%) have reliable fitting results after applying numerous conservative filters. This work is the first of its kind to fit and measure {lambda}862 and {lambda}864.8 simultaneously in cool-star spectra. Based on these measurements, we find that the EWs and CDs of {lambda}862 and {lambda}864.8 are well correlated with each other, with Pearson coefficients (rp) of 0.78 and 0.87, respectively. The full width at half maximum (FWHM) of {lambda}864.8 is estimated as 1.62+/-0.33nm which compares to 0.55+/-0.06nm for {lambda}862. We also measure the vacuum rest-frame wavelength of {lambda}864.8 to be {lambda}0=864.53+/-0.14nm, smaller than previous estimates. We find solid confirmation of the existence of the DIB around 864.8nm based on an exploration of its correlation with {lambda}862 and estimation of its FWHM. The DIB {lambda}864.8 is very broad and shallow. That at {lambda}862 correlates better with E(BP-RP) than {lambda}864.8. The profiles of the two DIBs could strongly overlap with each other, which contributes to the skew of the {lambda}862 profile.