Hyperthreading makes a difference when using 3Dmark06's CPU test. As shown here, with HT on, CPU score is higher. In effect you are comparing 8threads of i7 to 4 threads of the PhenomII.
Although, I'm sure if it were also 4GHz, it would match or surpass the P2... but still that's damn impressive assuming the P2 benchmark is legit.
Also, in real life usage, HT would still be enabled. It's not Intel's fault that its CPUs can execute two threads per core and AMD can't.
When I get back from work, I will try and get 4GHz with no HT and do another run. Oh, and I forgot to mention, for both of my tests, turbo mode is disabled. Again in real life, it would probably be enabled.