Use the following filter. The Tops don't work. You can only filter for base members. If you need to filter the 'None' for UD7 and UD8 add them.
Filter = "UD2='LE00037' AND UD3='200020' AND Account In ('GrossRevTime', 'BillExps')"
As Henning said, CVs use paging precisely to avoid this sort of performance issues, as well as other fine-tuning. Trying to compare them to a BR is not very useful. What will be useful, is to try and speed up your BR; there might be things you've not optimised...
For example, FDX calls allow you to specify the level of parallelization they will employ; did you try tweaking those? If they're too low, they are going slower than they could; if they're too high, they're slowing down the server.
Also, extracting data in memory is only half-the battle - data serialization, where you save the output to file, is likely to be slow. If your BR is saving data to file, did you try timing how long that takes, versus how long it takes for the FDX call to return? There might be ways to speed up serialization, depending on output requirements, but if you need One Big File chances are that it's always going to be fairly slow - you're outputting a lot of metadata names with those records (unlike the CV...).
