sql server 2005 - What would make a table "slow?" -

what make 1 table substantially slower another? easiest illustrate:

query 1:

select top 1000 * call c join call_task ct on c.call_no=ct.call_no left join memo_clt m on m.doc_ref=ct.record , m.doc_type='clt' , m.line_no=1 left join memo_clt m2 on m2.doc_ref=ct.record , m2.doc_type='clt' , m2.line_no=2 

query 2:

select top 1000 * call c left join ext_document_detail edd   on edd.doc_type='clh'                                              , edd.doc_ext_no=21                                             , edd.doc_ref=c.record left join ext_document_detail edsource  on edsource.doc_type='clh'                                                 , edsource.doc_ext_no=22                                                 , edsource.doc_ref=c.record 

the structure of tables similar, , i'm accessing ext_document_detail similar join compared memo_clt table. yet second query takes 40 seconds, while other 1 takes 0 seconds.

both have clustered index on 3 keys i'm using join. memo_clt table has non-clustered index on it's record column though... that's difference can spot , don't think make big difference.

so why difference in speed here?

edit: since martin asked, here results of set statistics io on query 1:

table 'memo_clt'. scan count 2000, logical reads 6454, physical reads 0, read-ahead reads 0, lob logical reads 2385, lob physical reads 0, lob read-ahead reads 0. table 'call_task'. scan count 1, logical reads 39, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. table 'call'. scan count 1, logical reads 25, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 

query 2:

table 'worktable'. scan count 0, logical reads 0, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. table 'ext_document_detail'. scan count 1001, logical reads 1507004, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. table 'call'. scan count 1, logical reads 24, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 

right off bat 2 things striking me. first there no such table "worktable." second absolutely huge number of logical reads... cause that?

it's not tables causing differences in speed. structures of joins , supporting indexes on tables being queried.

to give reason difference in speed i'd need see execution plan. suspect 1 query utilizes indexes better another.

a place start see if have table scans. if have these , can optimize see increase in performance.

i give this article read. it's worth checking out , understanding..