Multi-processor virtual machines?

[konsultimedia]

Hello,
I'm trying to compare multi-CPU virtual machines with physical ones.

The problem is that the virtual machines get so bad numbers that I'm starting to doubt that PerformanceTest works properly on them.

It's mostly the CPU mark that seems odd.
Examples:
Physical Core 2 Duo 2,4GHz dual core: 1484
Physical Xeon 5110 1,6GHz dual core: 1890
Physical Xeon 5130 2,0GHz quad core: 2374
Physical Xeon 5420 2,5GHz quad core: 3202

Virtual Xeon 5440 2,8GHz dual core: 800
Virtual Xeon 5520 2,5GHz quad core: 902

As you can see, the virtual machines get abysmal results, compared to their reported specification.

I'm certainly expecting a performance drop for VM's, but not this much. Obviously. I'm doing something wrong here (or is it that PassMark is not suited for virtual machines?).

Best regards,
/Tomas

[Michael (Passmark)]

It looks like these virtual machines are only utilising one core in the tests.

Firstly you should check to make sure your virtual machines do in fact have access to all the cores on the CPU. Some VM packages will limit you to using only 1 of the cores.

Secondly, under preferences in PT make sure the number of processes is set to equal or greater than the number of cores available to the machine. This should be automatically set to the correct number during the first run however it's possible PT incorrectly detected the number of available cores under the virtual environment.