anda dapat menggunakan fungsi windows seperti density_rank :
select *,dense_rank() over (order by b) group_number from t order by b
jadi kamu bisa :
select * from (
select *,dense_rank() over (order by b) group_number from t order by b
) a where group_number<=2