To be a little bit more fair you should write the Squeak benchmark exactly like the C++ one:
| a | a := Array new: 1. 1 to: 10000000 do: [:i| a at: 1 put: 3.]
Using #to:do: gives a speed-up by 100% here!
-- Bert
PS: of course, we should change the implementation of #timesRepeat: accordingly
No, it is not "more fair". All Smalltalks run the same code, C++ was here just for... I don't know exactly why... just for fun?
Viktor