Also, you need to realize that for ARM to increase performance, they will need to also increase power consumption, and thus lose the only aspect of their so called ISA that made them popular -----
No. ARM (the company) are keeping within the power budgets for mobile devices. ARM (the company) uses the extra transistors that a given process (node shrink) allows to increase performance. This is why the (QCOM's) S4 *increases* it's performance/watt profile, the same will be true of ARM's A15. The A7 goes the other way, it actually *reduces* performance and power.
Then there are design/process improvements made by the foundries/licensees, for example TI's 4470:
"For starters, the 4470 is rated at a CPU speed of 1.8GHz, 20% faster than the 4460 (which TI originally announced as the 4440 but later renamed the 4460 in typically capricious fashion). TI’s circuit designers have spent the past six months hammering on critical speed paths in the Cortex-A9 CPU to shave off a few dozen picoseconds here and there from the already reoptimized 4460 layout. These guys must be good, as this speed is attained in the same TI-defined 45nm LP process that the other OMAP4 products use."
Other ARM ISA designs may take a different tack, and in which case your statement is correct. Nvidia have suggested that their Denver core will have high single thread performance in which case they will use more transistors and thus power in their design. ----- I can't wait to see actual product comparisons in the near future. -----
Neither can I:) In the meantime if you want your benchmark fix there is always Geekbench: