The number that floats around somewhere is that 5000 to 10000 records at a time
will work the best. Too few at a time takes too long; too many at a time uses
too much memory.
I would import the CSV file first, and make it an Access table. This will let
you wrestle with any data conversion or missing data issues that might arise,
and you can also index it.
Assuming you have a key in both tables by which to match them up for the update,
you can make an update query that accepts two parameters: a lowest key and
highest key, for selecting the range to update. Then run the query repeatedly,
entering 1 and 5000, then 5001 and 10000, 10001 and 15000, etc. This can be done
in code, of course, with a nice little status bar update showing progress...