Interesting.
Did you try counting the key as well?

I read about something similar a couple weeks ago. Seems you must have nulls in column 1 position of the query. http://stackoverflow.com/questions/59294/in-sql-wh...

Wow, that's pretty crazy... Are you sure there aren't caching effects at work here? I had a DBA who was crazy about about SELECT * being horribly inefficient, and got on our case about this with SQLServer 2000, but after several benchmarks, SELECT COUNT(*) was indistinguishable from SELECT COUNT(1).

I know that I tested the same in SQL Server 2008, and the results were identical. You might want to check your query plan, and make sure there's not some weird joining going on with the query optimizer.

Also, a more efficient way to write the SELECT COUNT(...) would potentially be to do the group on a smaller record set, i.e.:

SELECT TotalOrders, OrderVendors....
FROM (
SELECT COUNT(*), OrderVendor_ID
FROM Orders
WHERE Orders.TypeID IS NOT NULL
AND Orders.CreateDate > ?
GROUP BY OrderVendor_ID
) TotalOrderCounts INNER JOIN OrderVendors ON OrderVendors.OrderVendor_ID = TotalOrderCounts.OrderVendor_ID

etc...

I've observed the opposite in several cases, on SS2008. Queries with select(1) were in the hundreds of milliseconds, and replacing with select(id) got them down to tens of milliseconds.

What about during meetins? Is it ok to write SELECT * on the whiteboard?

I've found it more efficient in meetings to write SELECT (fieldlist) so that someone doesn't feel obligated to point out that SELECT * is inefficient.