The identification of the red giant branch bump brightness in metal-poor globular clusters is important for low-mass stellar evolution. The release of Gaia DR2 prompted us to revisit the red giant branch bump (RGBB) in galactic globular clusters. We apply a popular nonparametric density estimation approach, kernel density estimation (KDE), to explore the position of RGBB in 7 metal-poor globular clusters (GCs). The G and V magnitudes of the RGBB according to our clustering algorithm, G(B, K) and V-B,V-K, respectively show the RGB bump magnitude detected by the KDE method in G band and V band. They show the overdensity location in the luminosity function of the RGB stars. Based on the results derived by KDE, a maximum-likelihood analysis via a Markov Chain Monte Carlo (MCMC) approach is adopted to detect the RGB bump feature and obtain more accurate RGBB brightnesses in G band and V band for the samples. We find that the red giant branch bump brightness becomes fainter as the global metallicity increases in clusters with [M/H] <= -1.4. We present the empirical relation between the global metallicity [M/H] and absolute magnitude MV of the red giant branch bump for clusters with [M/H] <= -1.4. We verify that discrepancies between observations and theory for metal-poor globular clusters with [M/H] <= -1.4.