alter sequence g restart with 0;
recreate table t1(id int primary key, x int);
recreate table t2(id int primary key, x int);
commit;

insert into t1 select gen_id(g,1), 1000 + rand()*1000
from rdb$types,rdb$types,rdb$types
rows 50000;
commit;

insert into t2 select gen_id(g,1), 1100 + rand()*2100
from rdb$types,rdb$types,rdb$types
rows 5000000;
commit;

create index t1_x on t1(x);
create index t2_x on t2(x);
commit;


select count(*)
from t1
where exists (select x from t2 where t2.x = t1.x); --  7s 437ms, 456973 fetches

-- vs --

select count(*)
from t1
where exists (select x from t2 where t2.x = t1.x order by t2.x); -- 313ms, 390911 fetches

PLAN (T2 ORDER T2_X)
PLAN (T1 NATURAL)