Did you mean it’s taking 3 minutes for the export to complete? The entire export (point encoding, history exports) run asynchronously so depending on the state of the system, it may be slow if it is already very busy.
You can also increase the number of threads you allow the exporter to have. Assuming you only have a single export in your system, this is how many simultaneous points will be encoded so if you have a larger point count in your hierarchy, you may wish to up this to about 10. More threads does mean heavier CPU usage during exports so be careful not to make this too high. I would not recommend going much higher than 20 threads depending on the system as there are diminishing returns at a certain point.
Additional time will be added if you are also exporting historical data. The initial export generally takes longer and subsequent less time due to it only updating new records vs the full data set. The smaller your export interval is, the less historical records get exported which slows the export time. The history size on the Niagara station also effects the speed since the larger the historical database gets, the longer it takes to query it.