I'm not sure how the 2006 and 2007 Mac Pro's have bus speed disadvantage when they have a 1333MHz FSB which is higher than the CPU clock speed of the Apple A4 which is only 1GHz. I don't think video decode ability is very good as a pure performance measure. All that matters for video decode is that there are no dropped frames for high-bandwidth h.264 video which should be the case for a 4 core 2006 Mac Pro or 8 core 2007 Mac Pro even without the dedicated h.264 decode hardware that the Apple A4 presumably has. If you are measuring CPU usage or power consumption on video decode than the Apple A4 would be superior.I think the primary differentiating factor between a Geekbenched MacPro and a Geekbenched iPad will be Apple specific optimizatons in the silicon we have yet to have even a basic understanding of. GB has several different tests all combined to a final figure, so that gives my prediction a leg up since there will be heavy emphasis on video decoding capacity.
Finally I have a card in my pocket for practical speed. The MacPro (and iMac) I referenced use moderately slow busses and HD's. The iPad uses flash which in effect is a RAMDISC, and despite the low CPU Ghz, it will have a speedy bus.
In any case my prediction is on the record. My overall results are pretty damn good.
Rocketman
how they interact with disc.
Video encoding would be more interesting, but there's no info yet on whether the Apple A4 has dedicate video encoding hardware and with what limitations. Similarly, I haven't seen comparisons on the efficiencies of ARM's new NEON SIMD instruction set compared to up to SSSE3 that 2006 and 2007 Mac Pros support. Even with dedicated video encode hardware, it's hard to see the Apple A4 come even close to the performance of an 8 core 2007 Mac Pro.