create table tb
(
id int
,date datetime
,dvalue int
)
go

insert into tb  
select 1,'2011-12-30 19:29:00',1
union all select 1,'2011-12-30 19:36:00',1
union all select 1,'2011-12-30 20:00:00',1
union all select 2,'2011-12-30 20:29:00',3
union all select 2,'2011-12-30 20:31:00',2
union all select 3,'2011-12-30 21:00:00',5
union all select 4,'2011-12-30 22:00:00',6
union all select 5,'2011-12-30 23:00:00',7

这里需要精确到分钟，取分钟最大值；
/*-------显示结果-------
1,'2011-12-30 19:36:00',1
1,'2011-12-30 20:00:00',1
2,'2011-12-30 20:31:00',2
3,'2011-12-30 21:00:00',5
4,'2011-12-30 22:00:00',6
5,'2011-12-30 23:00:00',7
*/

select * from tb t where not exists (select 1 from tb where id=t.id and datediff(hour,date,t.date)=0 and date>t.date)

------解决方案--------------------
以什么分组不是很名确
取分钟最大值,那应该的结果是:
/*-------显示结果-------
1,'2011-12-30 20:00:00',1
2,'2011-12-30 20:31:00',2
3,'2011-12-30 21:00:00',5
4,'2011-12-30 22:00:00',6
5,'2011-12-30 23:00:00',7
*/
还是:
/*-------显示结果-------
1,'2011-12-30 19:36:00',1
2,'2011-12-30 20:31:00',2
3,'2011-12-30 21:00:00',5
4,'2011-12-30 22:00:00',6
5,'2011-12-30 23:00:00',7
*/


------解决方案--------------------
SQL code
create table tb
(
id int
,date datetime
,dvalue int
)
go

insert into tb  
select 1,'2011-12-30 19:29:00',1
union all select 1,'2011-12-30 19:36:00',1
union all select 1,'2011-12-30 20:00:00',1
union all select 2,'2011-12-30 20:29:00',3
union all select 2,'2011-12-30 20:31:00',2
union all select 3,'2011-12-30 21:00:00',5
union all select 4,'2011-12-30 22:00:00',6
union all select 5,'2011-12-30 23:00:00',7
go


select t.* from tb t where not exists(select 1 from tb where id = t.id and date > t.date) order by t.id
/*
id          date                                                   dvalue      
----------- ------------------------------------------------------ ----------- 
1           2011-12-30 20:00:00.000                                1
2           2011-12-30 20:31:00.000                                2
3           2011-12-30 21:00:00.000                                5
4           2011-12-30 22:00:00.000                                6
5           2011-12-30 23:00:00.000                                7

（所影响的行数为 5 行）
*/


select t.* from tb t where date = (select max(date) from tb where id = t.id) order by t.id
/*
id          date                                                   dvalue      
----------- ------------------------------------------------------ ----------- 
1           2011-12-30 20:00:00.000                                1
2           2011-12-30 20:31:00.000                                2
3           2011-12-30 21:00:00.000                                5
4           2011-12-30 22:00:00.000                                6
5           2011-12-30 23:00:00.000                                7

（所影响的行数为 5 行）
*/


drop table tb