Rob>Yes an IOP 1.2GHz will *not* give the same performance as 3-4GHz 64bit x86 CPU when running the same task natively. However we're not comparing horses for courses here - an ARM CPU run at 1.2GHz with the larger cache (L2) offers the means of running *natively* ARM RISC OS code that'll outpace an x86 based ARM emulator and probably by a margin too.
Yes I'll also concede that they may cost more than an equivalent x86 set up - but then you're paying for the ability to run RISC OS natively (i.e., at it's most efficient and with no reliance on another OS to run).
As to multiple cores yes RISC OS can't natively use them - but surely some "background" or applications specific tasks can be so done. (Ok you won't get x2 times the performance of a single core version of the same CPU and it may not be as transparent or as flexible - but a gain is a gain so let's pocket it eh ).