Parallels 6: assigning more cores -> slower Guest OS performance

[John BrianK]

I recently installed Parallels Desktop 6 and a Windows XP VM on my new 2011 Macbook Pro with the following specs:

2.2GHz quad-core i7 w/ HyperThreading
8GB RAM
128GB SSD
OSX 10.6.7

I'm using this machine in part to do development in Microsoft Visual Studio 2008. Running under Parallels, I've noticed some very odd behavior: beyond two, the more CPUs I assign to my VM in the Preferences pane, the slower my compiles seem to go.

With a single CPU assigned, compiling pegs the Windows Task Manager CPU meter at 100%, and I see a single core in OS X's Activity Monitor pegged as well.

With two CPUs, Task Manager shows both CPUs running around 80-90% capacity, and I see all four of my physical cores running at various utilizations between 20% - 70% in Activity Monitor.

With four CPUs, Task Manager shows between 20% - 40% CPU utilization across the cores, while Activity Monitor shows four cores running around 40% - 65%.

I haven't tried with more (I have up to eight CPUS available in the Parallels configuration). Changing RAM amount doesn't seem to change anything.

Can anyone help me understand what's going on here? Compilation is far slower than it ought to be on a machine with these specs, and I'd really like to be able to take full advantage of the cores in this machine to speed things up. Anybody else experiencing weird slowdowns with CPU heavy stuff and > 1 core assigned?

Thank you!

 

[MarcCa]

Hello,

I have installed Parallels on my new MacBook Pro and I have experienced the same problems that you have. Have you solved it?

I have created a Windows XP virtual machine SP3 with all updates to compile with Visual Studio 2008 and I have experienced that when more cores I added to the virtual machine slower is visual studio to compile.

My Visual Studio project takes about 10 minutes to compile with two cores, 22 minutes to compile with 4 cores and about 35 minutes to compile with 8 cores.

I have tested the virtual machine with benchmark software and the machine with 8 cores is 4 times faster with integer and float operations than the machine with two cores, except for md5 hash operations in this case the machine is little slower than 2 cores machine.

I would like to know if this behavior that compiling Visual studio is normal? People that you have older MacBook Pro models have you experienced this?

Thanks

My Macbook Pro specs are:
2.2GHz quad-core i7 w/ HyperThreading
8GB RAM
128GB SSD
OSX 10.6.7

 

[JakubW]

Hi,

I'm experiencing similar issues with my Windows XP installation on similar MBP i7 2.2. There is a huge speed difference when running system with 1, 2 or 4 cores.

I wonder whether this is Parallels issue or Windows XP on SandyBridge issue (though I haven't found any evidence of the latter in google).

Kuba