For kicks, I downloaded the common benchmarking program superpi: http://www.techpowerup.com/downloads/125 Some review sites run the program on its setting to calculate pi to 2 million digits 20 times. I did that several times in the Parallels VM (Beta 2) and the result (verified by wall clock) is consistently around 1m 12 seconds (I had a bunch of other stuff open, but it wasn't loading the processor at all). I'm running it on system 10.4.6, MacBook Pro, 2.0GHz. Now check out the results from this recent notebook.com review of the Dell laptop that uses the same Core Duo chip at 2.0G: http://www.notebookreview.com/default.asp?newsID=2827 Notebook Time Dell Inspiron e1505 (2.0GHz Core Duo) 1m 16s Acer TravelMate 8204WLMi (2.0GHz Core Duo) 1m 15s Sony VAIO FS680 (1.86 GHz Pentium M) 1m 53s Lenovo ThinkPad T60 (2.0GHz Core Duo) 1m 18s IBM ThinkPad Z60m (2.0 GHz Pentium M) 1m 36s Fujitsu LifeBook N3510 (1.73 GHz Pentium M) 1m 48s Dell Inspiron 6000D (1.6 GHz Pentium M) 1m 52s Dell Inspiron 600M (1.6 GHz Pentium M) 2m 10s HP Pavilion dv4000 (1.86 GHz Pentium M) 1m 39s HP DV4170us (Pentium M 1.73 GHz) 1m 53s Sony VAIO S380 (1.86 GHz Pentium M) 1m 45s So, how can the parallels virtual machine be 4 seconds faster than a real Windows XP box with the exact same chip, and way faster than any real Pentium-M? Have I entered the Steve Jobs reality distortion field?
The Parallels VM execute CPU instructions directly on the host CPU without having to do emulation, so it's not surprising that the performance would be comparable with native performance.
