The processors are actually faster on other many levels.
1) The new Mac Pro uses Nahelem 5500 series Xeon processors. These processors have a turbo boost functionality that can adapative overclock the processors. The 2.66 GHz will top out at around 3 GHz and the 2.93 will top out at around 3.33GHz - so they are faster. When not doing heavy processing, there is no real point in having a cpu running at max speed, it consumes less power and produces less heat running at a lower speed. For more information about turbo boost - see:
IT @ AnandTech: The Best Server CPUs part 2: the Intel "Nehalem" Xeon X5570
2) Nahelem are not just quad core processors, they are quad core hyper threading processors. Hyper threading allows for multiple threads to be run simultaneously per core (multiple meaning 2). This means that a single quad core has a performance factor of about an 8 core machine (this is theoretical, actual is a bit less performance in some areas, and bit more in others). This also means that an 8 core machine has up to 16 virtual cores with HT - this is twice the capability of the 8 core '08 mac pro (although actual performance boost is less - this is due partially from it not being a true 16 core, and partially from you loose a certain amount of performance as you add more cores due to the workload of distributing threads across the cores. I don't remember the figures but I want to say the boost in performance was between 1.6 and 2x the capability of the old machines depending on the work being done)
3) You have a very different architecture for i7 based cpus (the nahelem cpus are i7 baesd). Traditional motherboard design has required for an external memory controller that was part of what was known as the front side bus. Intel has done away with that, embedding the memory controlling logic within the cpu (the other functions of the FSB have been offloaded to a controller having a different name of which suddenly I can't remember). This improves the overall ram transfer capability
4) i7 based cpus have the capability of tripple channel ram (DDR3's in sets of 3) which inrease data throughput
These are just some of the things that make the chips faster even tho their clock speed is slower.
It's been a long arguement that clock speed is not the end-all be-all of system capabilities. There is just a lot more to operation then brute force speed.
Another simple example (that is going way beyond the scope of this discussion but oh well
) is the design of 2 fictitious chips for purposes of discussion, ie:
if you have 2 cpus
both 32 bit
1 is a 2 GHz (2 billion clock cycles per second) cpu
1 is a 1.5 GHz (1.5 billion clock cycles per second) cpu
on the surface, the 2 GHz is faster
Then you find out that it take on average 5 clock cycles to do 1 instruction (average based on many factors being not described for sake of brevity) on the 2GHz machine, and on average 2 clock cycles to do 1 instruction on the 1.5 GHz machine. This means that you will at most get 400mips on the 2GHz machine, but 750Mips on the 1.5 GHz machine (mips = Million Instructions Per Second). Potentially the 1.5GHz machine is faster because it takes less time to execute instructions.