virtualbox.org

Hi,

I did some benchmarking to eke out as much CPU performance from my 4 core desktop system as possible, and was surprised by the hyper threading results - written up at

http://envobi.com/post/virtualbox-hyper ... -surprise/

Cheers

Interesting info, WillShak! One thing mentioned in the forums when the subject of benchmarks in a guest is that the timers needed by the benchmarks aren't connected as well in the guest as they are in the host, so they may not fire properly to give the benchmark program accurate results. The best way to time the performance of a guest is an old-school stopwatch. Run a few iterations of a non-benchmark program that makes use of the desired host or guest functions, and which will take enough time in running to allow reasonable use of the stopwatch. Change the processor setup between runs and see what you get. It would be interesting to see if the results were similar or different.

One other thing, taking all the cores, physical or hyperthreaded, from the host make allow the host and guest to run together just fine for a while - Virtualbox doesn't actually take the core away, best of my knowledge, it allows the guest to run processes on the core. But if a guest thinks it can run on eight cores and all the host has is eight cores, then the guest may one day start a multi-core process and see that, hey, I have eight cores, let's use them all! Then the host chokes...