T-SQL: Exists versus In – Performance Comparison

In this test I’m going to query which numbers exist in Table1 but not Table2, using “Exist” and “In”.

Exist Query:

select * from Table1 A where not exists
( select * from Table2 B where B.col1 = A.col1 )

In Query:

select * from Table1 where col1 not in
( select col1 from Table2 )

There will be 2 tests:

• Test 1: Table2 is large
• Test 2: Table2 is small

I believe that in Test 1 “exist” is a lot faster, but in Test 2 “in” is slightly faster than “exist”

Test 1: Table 2 is large

Data Setup:
Table1 contains numbers from 1 to 100,000.
Table2 contains numbers from 10 to 100,000.

Expected results: Exist is faster than In.

Script to populate both tables with data:

truncate table table1;

declare @i int
set @i = 1
while @i <= 100000
begin
insert into Table1 values (@i, @i)
set @i = @i + 1
end;

truncate table table2;

declare @i int
set @i = 10
while @i <= 100000
begin
insert into Table2 values (@i, @i)
set @i = @i + 1
end;

Results:

• Exist Query: average of 167 ms.
• In Query: average of 1892 ms.

This is as I predicted.

Test 2: Table 2 is small

Data Setup:
Table1 contains 100,000 rows. Col 1 contains numbers from 0 to 50.
Table2 contains 49 rows. Col 1 contains numbers from 2 to 50.

Expected results: In is slightly faster than Exist.

Script to populate both tables with data:

truncate table table1;

declare @i int
set @i = 1
while @i <= 100000
begin
insert into Table1 values (@i%50, @i)
set @i = @i + 1
end;

select * from Table1

truncate table table2;

declare @i int
set @i = 2
while @i <= 50
begin
insert into Table2 values (@i, @i)
set @i = @i + 1
end;

select * from Table2

Results:

• Exist Query: average of 219 ms.
• In Query: average of 253 ms.

So I was wrong. Even if Table 2 is small, “Exist” is still faster than “in” (slightly). But if Table 2 is large, “Exist” is a lot faster than “in”. So it is better to use Exist. Always.