virtualbox.org

I recently ran the "Windows Experience" benchmark (accessible via Control Panel) for 32-bit Windows 7 on VirtualBox 3.2.8, Fusion 3.1.1, and Boot Camp 3.1.
I have a MacBook Air running OS X 10.6.4 with 2 GB RAM and 1.8 GHz Intel Core 2 Duo CPU. I ran VirtualBox and Fusion guests on an external USB hard disc; Boot Camp on the internal SSD.
For VirtualBox and Fusion, I allocated 0.75 GB RAM to the Windows 7 guest and 1.25 GB RAM to the OS X host.

The results:

VirtualBox 3.2.8:
Processor 5.6
RAM 3.9
Graphics 1.0
Gaming Graphics 1.0
Primary Hard Disk 5.9

Fusion 3.1.1:
Processor 3.5
RAM 3.9
Graphics 3.0
Gaming Graphics 3.0
Primary Hard Disk 6.2

Boot Camp 3.1:
Processor 4.3
RAM 4.7
Graphics 3.5
Gaming Graphics 3.1
Primary Hard Disk 5.8

Note: VirtualBox graphics scores were the same both with and without the "experimental" 3D graphics guest addition enabled. 64-bit Windows 7 via Boot Camp gave the same scores as 32-bit except Processor improved from 4.3 to 4.6 and RAM improved from 4.7 to 4.8 . Puzzling that 32-bit Windows performed better via VirtualBox than 64-bit via Boot Camp.

Excecutive Summary: VirtualBox seems good, except for the graphics. Does this matter to anyone except gamers?

Has anyone else run similar tests?

George

I don't currently use VirtualBox, However, I found similar differences between Fusion 3.1.1 and Bootcamp using the Windows Experience Index. The general consensus in my reading about WEI is that it is not a reliable benchmark/indicator for real world performance. However, when I ran GeekBench the performance for single-thread processes was almost identical for Fusion and Bootcamp. Note that I was running Fusion on a single core which is why I am only comparing single thread processes. This is fine for me because the key apps I use in VM are not multi-threaded (e.g., ESRI ArcGIS).

My machine is a MacBook Pro i7 2.66GHZ, 8GB RAM. My fusion VM has 4GB of RAM and a single CPU core.

I can provide the Geekbench results if anyone is interested.

It would be interesting if you include VB in the comparing benchmarks. It's of little use here to see benchmarks from competitors without our own product to compare it with.