Vista vs XP Benchmark puzzle

I have a Vista Business VM and an XP Pro VM, both on a separate striped RAID volume in my Macpro 2.66 with 5Gb RAM. Both VMs have 700 MB RAM, and Virtual Memory paging in both machines is set to be controlled by the PC (not the very low default values they started with). The only difference in set up is that the XP is an expanding disk and the Vista one is a fixed 32Gb. Parallels is v3188.

For what I do, I am very happy with the speed of Vista, which I use most of the time, though it does not feel quite as snappy as XP.

The geek in me wanted to measure and compare the performance so I put Sisoft Sandra Lite on both and ran it several times, varying virus on and off, and "adjust for best performance". Apart from the virtual memory paging both XP and Vista are pretty much default installlations with the same software installed.

None of the adjustments make much difference, but always XP scores MUCH higher on the CPU scores for the "arithmetic" test. eg

XP 33,572
Vista 10,253

The "multimedia" score is similarly much higher for XP, as is the "Filesystems" score, but by a smaller factor (about 50%). The "memory bandwidth" score is about the same.

The version of SiSoft Sandra is the latest, stated to be Vista compatible.
Is there a setting in Parallels or the VMs which can cause this?
Is there a more appropriate benchmark tool?

Benchmarks with SiSoft Sandra on PCs found on Google show that XP and Vista are within a very few percent of each other, so why is there such a big difference in Parallels?

Anyone got any comment on this?

Thanks

Mike

 

Seems about right. Vista, like all OS releases is more bloated than it's precessor. It may not be as big of a difference as indicated by the numbers, but in my experience every MS OS release is much slower than the previous on equivalent hardware.

Surely you weren't expecting Vista to be faster??

 

No, but nor was I expecting it to be over three times slower!

I don't actually believe it is. I just did some comparisons of my own, timing the same photoshop processes on both and they are about the same.

 

FWIW, earlier today I was reading a VMWare white paper on benchmarking and VMs. I don't recall the exact URL (look at the PDFs associated with their new Workstation 6), but I do remember reading that SiSoft Sandra was considered unreliable for VM benchmarking.

Apologies for mentioning the competition here. It seems to me that this info might be vendor-neutral and potentially useful to us Parallels users who are interested in tweaking our VMs.

 

Certainly seems to be my experience. Might be useful for assessing tweaks on one OS, but not comparing OSs.