Splitting tables or databases? That's the problem

On a 128MB RAM, 450 MHz pentium 3 server with linux gentoo and postgresql
7.4.6 on an office lan we can manage satisfactorily a db containing few
tables with a cumbersome amount of data (each table is around 650 thousand
records with 98 columns) relating to the year 2002.
We use M$-Access and ODBC to access those data.
Now similar data and amounts of records for the year 2003 need to be added.
Then queries to contrast 2003 data vs. 2002 ones will be needed.

In view of the poor hardware at our disposal, is it better from the standpoint
of efficiency to:

1) Create a new database for 2003 with the same structure of that for 2002;
2) Appending new 2003 data to 2002 data in the same tables;
3) Creating in the original database new, separate tables to contain data
for 2003.

Thanks
Vittorio