The aim of this study was to evaluate the reproducibility of voxel-wise temporal signal-to-noise ratio (tSNR) on repeated scans across runs, sessions, and days. A group of 21 participants was scanned 16 times (4 runs per session, 2 sessions per day, 2 separate days) in a functional magnetic resonance imaging (fMRI) study on a 3T Philips Achieva scanner. For each run, we calculated t-value and tSNR maps. To ascertain that the results were not specific to the scanner, one volunteer was scanned with four fMRI runs in a single session on the above 3T Philips scanner as well as a 3T Siemens Prisma scanner. The coefficient of variation of voxel-wise tSNR across the 16 repeats was up to 25%, while the range relative to the mean of all observations was up to 80%. The voxel-wise variability of tSNR on the two different scanners was similar, indicating a general issue. Despite its use in evaluating the quality of fMRI data, we found only a weak relationship between tSNR and t-values. There is very high variability in voxel-wise tSNR, which should be considered while planning future studies that aim to identify small and focal fMRI effects or the benefits of incremental improvement in methods.