I have posted a blog post about Different ways to find DISTINCT values where I have shown six different ways to find DISTINCT values. Here is the analysis of how long each method takes.

Create a temporary table named #sales_details

create table #sales_details(item_id int not null)

Populate one million data into this table

insert into #sales_details (item_id)
select 
	top 1000000 
		checksum(newid())%10000
from 
	sys.objects as so1 cross join
	sys.objects as so2 cross join
	sys.objects as so3 cross join
	sys.objects as so4	

Test various methods to find DISTINCT values

SET STATISTICS TIME ON 
print '======================================================'
Print 'Method 1 starts'
print '======================================================'
select distinct item_id from #sales_details 
print '======================================================'
Print 'Method 2 starts'
print '======================================================'

select item_id from #sales_details 
Group by item_id
print '======================================================'
Print 'Method 3 starts'
print '======================================================'
select item_id from #sales_details 
UNION
select item_id from #sales_details 
print '======================================================'
Print 'Method 4 starts'
print '======================================================'
select item_id from #sales_details 
UNION
select item_id from #sales_details where 1=0
print '======================================================'
Print 'Method 5 starts'
print '======================================================'
select item_id from #sales_details 
UNION
select 0 where 1=0
print '======================================================'
Print 'Method 6 starts'
print '======================================================'
select item_id from
(
	select row_number() over (partition by item_id order by item_id) as sno,* from #sales_details 
) as t
where sno=1
print '======================================================'
Print 'Method 7 starts' Pointed by Dwain camps 
print '======================================================'
;WITH    RecursiveCTE
AS      (
        SELECT  data = MIN(T.item_id)
        FROM    #sales_details T
        UNION   ALL
        SELECT  R.item_id
        FROM    (
                -- A cunning way to use TOP in the recursive part of a CTE :)
                SELECT  T.item_id,
                        rn = ROW_NUMBER() OVER (ORDER BY T.item_id)
                FROM    #sales_details T
                JOIN    RecursiveCTE R
                        ON  R.data < T.item_id
                ) R
        WHERE   R.rn = 1
        )
SELECT  *
FROM    RecursiveCTE
OPTION  (MAXRECURSION 0);
SET STATISTICS TIME OFF

Here is the output shown in tabular format

The performance of the method 7 that involves in recursive CTE depends on the index as well as the nature of the data. Without index, it is running for several minutes. With index it is running fast. But if there are lesser number of distinct values, it runs very fast. For example if we use checksum(newid())%100 in place of checksum(newid())%10000, method 7 outperform every other methods