Integer performance; when > 2 cores?
A simple test of integer performance is to run a Windows chess program and see how many nodes-per-second it computes. (Indeed, the Crafty program was part of SPECint 2000.) Comparing Parallels 3.0 and VMWare Fusion beta VMs each using the same Boot Camp partition on my 2.6Ghz 4-core MacPro with 3GB RAM, and remembering to set the chess engines running under Fusion to run just 1 processor only, I get:
() Deep Fritz 10 is 33-40% faster on VMWare than on Parallels
() Rybka 2.3.2a is 40-50% faster on VMWare
() My own Borland C++ compile of Fruit 2.1 is only about 15% faster on VMWare.
(All engines have 512MB hash, and both VMs have about 1.3GB RAM assigned.)
Of course I'd like my chess programs to use all 4 cores, the way they do under Boot Camp proper, so: when will Parallels go multi-processor, and when will Fusion allow more than a dual-core VM?
Also, the need to reactivate my Boot Camp partition seemingly whenever I switch from one to the other is a royal pain... !