My understanding is that a query executing at DOP 4 will never use more than 4 CPU cores at a time during execution, unless you're doing something like using UDFs, CLR, or other odd things. This implies that the CPU time for a DOP 4 query will never be greater than four times the elapsed time. One could say that a perfectly efficient query will have CPU time = 4 * elapsed time and any lower ratio suggests that a wait event could be eliminated to improve query execution time.

Today I happened to notice a DOP 4 query reporting a CPU time greater than 4 times the elapsed time. This has been reproduced on SQL Server 2019 and 2017 on five different machines by three different people. On my home computer I see the following:

> SQL Server Execution Times:  
> CPU time = 10594 ms,  elapsed time = 2355 ms.

I don't think this is a measurement problem with `SET STATISTICS TIME ON`. I see the same ratio when looking at QueryTimeStats in an Estimated Plan Plus™. An example plan is [here](https://www.brentozar.com/pastetheplan/?id=r1AhxvSOI).

It also seems to be backed up by Resource Monitor during query execution, although I admit I can't tell which workers are using the CPU. The following screenshot is from my machine with 8 physical cores and no hyperthreading:

![aaaa.PNG](/image?hash=46d7ec5e8fb193064cd8c4e6c61e7dee467f8f0dc4f3785db41acbaf01b796f5)

How can this MAXDOP 4 query have CPU time greater than 4X the elapsed time? Is it a possible bug in SQL Server? This [answer](https://dba.stackexchange.com/questions/205643/degree-of-parallelism-and-scheduler-usage) by Paul White suggests that SQL Server 2000 had a bug in this area.

Code to reproduce the issue is below. It may not work on all machines. I suggest having as little other work running as possible on the server and having a physical core count that exceeds 4.


    -- data prep
    CREATE TABLE [dbo].[N](
        N [int] NOT NULL,
    PRIMARY KEY CLUSTERED
    (N ASC) WITH (DATA_COMPRESSION = PAGE)
    );
     
     INSERT INTO [dbo].[N] WITH (TABLOCK)
    SELECT TOP (1000000) ROW_NUMBER() OVER (ORDER BY (SELECT NULL))
    FROM master..spt_values t1
    CROSS JOIN master..spt_values t2
    OPTION (MAXDOP 1);
     
    CREATE TABLE dbo.TopAnswers (ID BIGINT, INDEX C CLUSTERED COLUMNSTORE);
    
    GO 
     
    -- the problematic query
    INSERT INTO dbo.TopAnswers WITH (TABLOCK)
    SELECT 2
    FROM (
        VALUES
        (1), (1), (1), (1), (1), (1), (1), (1),
        (1), (1), (1), (1), (1), (1), (1), (1),
        (1), (1), (1), (1), (1), (1), (1), (1),
        (1), (1), (1), (1), (1), (1), (1), (1)
    ) v(x)
    CROSS APPLY (
        SELECT TOP (1048576) 2 val
        FROM (
            SELECT N
            FROM dbo.N
     
            UNION ALL
     
            SELECT N
            FROM dbo.N
        ) ca2
    ) ca
    OPTION (FORCE ORDER, LOOP JOIN, NO_PERFORMANCE_SPOOL, MAXDOP 4);