Postgresql – CSV or DB is quicker to read several lines from a file with 1,5 billion lines

If you have your first CSV loaded into a table, you can just as easily load the other one into a staging table (presumably with the same structure as the 'real' one). Then you can get the new rows by

SELECT * FROM staging_table
EXCEPT
SELECT * FROM real_table
;

Rows missing from the new CSV can be get reversing the two sides around EXCEPT. However, given the lack of a key on the staging table (and hopefully not on the real one - it's not clear from your question), deleting rows based on this comparison can be painful, especially with so many rows.

You can drop the staging table once you have finished.

(As far as I see, this approach will work in any RDBMS.)

Load the CSV file into a "staging" table. From there you can easily do the UPDATE/INSERT/DELETE in a single statement for each one

Something along the lines:

update real_table
  set ...
where exists (select 1 from staging_table where ...)

insert into ...
select 
from stage_table
  left join real_table 
where real_table.some_col is NULL

delete real_table
where not exists (select 1 from staging_table)

Then delete the data from the staging table.

Probably the fastest way to do it. The only drawback is that you have the data for one table twice in the database (until it is deleted from the staging table).