I recently purchased a new server and created a 2012 R2 VM running on a 2012 R2 Hyper-v host to replace an old server we have for accounting software.

I decided to do some performance comparisons between the old server and the new server by running a 6 month trial balance report on identical copies of the DB.  I was surprised when my tests showed the old server performing much better than the new server.  After further testing, I found the new server performs best with only one virtual processor.  In fact, with one virtual processor the new server outperforms the old.  As soon as I allow the vm to have more than one virtual processor, performance is impacted.

Can anyone shed some light on what I'm experiencing here?  It would be such a waste of hardware to run the VM with one virtual processor.

In case anyone wants to know:

Old Server

6 Month Trial Balance Test (60 Seconds)

Server 2003 R2

2 x Xeon 5120

4GB of Ram

7200 RPM Sata Drives

New Server

6 Month Trial Balance Test  With 1 Virtual Processor(43 Seconds )

6 Month Trial Balance Test  With 4 Virtual Processor(1:22 )

6 Month Trial Balance Test  With 32 Virtual Processor(1:11)

2012 R2 VM running on 2012 R2 Hyper-V host (No other VMs on host and BW is the only thing installed on VM)

2 x Xeon e5 2640 v3

32GB of Ram

10K SAS Drives in raid 10