virtualbox.org

There have been several discussions around various parts of the forum about using 32 bit SMP in VBox and the fact that it is faster if one uses a single CPU. I decided to test this out and have some results for you to digest below. First I wanted to see if there was any speed difference between 32 bit 1 and 2 cpu and also see what difference there was with 64 bit . I used for this test a Ubuntu 64 bit host 4 Gig of ram on a laptop. The results mean little as far as actual performance because it is all subjective, but it does show a clear difference. I did not install the 3D effects so the actual grade is going to be lower.

Test 1 was Windows XPpro 32 bit SP3 single CPU 512 meg Ram
Test 2 was Windows XPpro 32 bit SP3 Dual CPU 512 meg Ram
Test 3 was Windows XPpro 64 bit Sp2 Dual CPU 1024 meg Ram

For the Test 3 I used the same software (32 bit) so it will reflect hyper-threading and not the actual 64 bit dual core, but as I said that was subjective to real world where you run 32 bit apps on both types.
I have found that converting a Windows machine that was previously installed as a single processor to a multi-processor is hit-and-miss and usually works better if you select a single CPU. Also just for completion I am using the PUEL version of VirtualBox version 3.0.8
You must have at least SP2 for Windows XP or you need to get a patch from Microsoft to be able to use SMP

Comments and other test results are welcome, but what I am interest in is the difference between single and multiple cores preferably 32 bit, and if this was a clean install or an upgrade to SMP.

What I'd like to see compared is smp with the 'old way' (no smp) as in 2.2.4, or, does pre smp (cpu loadbalancing) work faster or slower then smp (both with 1 or 2 cores) ?

Ok here is an XPpro that has never seen IO APIC, VT-x, or SMP. It was being used on a 32 bit PC and I moved it to the same PC that I did the other test with so it would be comparable.

Funny huh? Better that a single CPU with all the bells and whistles, but still below the SMP using (2) CPUs.

Perryg wrote:Funny huh? Better that a single CPU with all the bells and whistles, but still below the SMP using (2) CPUs.

Its exactly what I expected..... for fun run this again on 2.2.4, it might just slightly outperform 3.08's 2 smp's.

Ha ha you dog you! I don't want to go to the trouble of down grading anything. Snag the test program performance test 7.0 from the net and run it.
I know that running the 32 bit program worked just fine, but with all the hoopla of the SMP (1) verses (2) CPU I wanted to see if there was a difference. Which I must say there was.

The really interesting part is the CPU mark in all of these since that is what we are really after see below:

• 32 Bit SMP CPU 1 = 750.2
• 32 Bit SMP CPU 2 = 1408.6
• 64 Bit CPU2 = 1427.9
• 32 bit single processor = 723.4

For reference the processor speed and specs are below:
Intel Core2 Duo P8600 @2.40GHz with 3072 KB L2 cache