I actually do hope Intel keep their xScale in the Wireless, network, mass storage controller space - because (guess what) they often are specced higher than common or garden ARM processors that are used in PDA's and the like.
The xScale used in the Iyonix (the IOP321) is an Input/Output processor - so actually has rather better PCI support, support for a range of interfaces and functionality that no self respecting PDA driving ARM would require. The successor to the IOP321 - the 332 (covered previously by Drobe) has a lot of bells and whistles that a typical PDA wouldn't have (or need) - such as 800MHz clock rate, support for faster PCI variants with a PCI-Express to PCI-X bridge, faster memory with DDR2 support (even ECC/Error Correction) and an improved interrupt controller.
By contrast the ARM9 is anemic, probably hitting around the performance level of a StrongARM (ah back to 1995 again - wonderful). Intel currently *do* actually produce and release the fastest ARM processors *and* the fact these processors are intended for high speed coms/networking or mass storage should be seen as a big PLUS because the performance demanded of them (and acchieved by them) is somewhat better than StrongARM, ARM9 or other PDA drivers. The IOP range have more relaxed power budgets (they don't have to be battery powered), they do communicate with faster devices than a PDA would ever need to so the interfaces provided are more in line with what one would expect of a desktop processor (or as close to one as we're likely to see in ARM format for the foreseeable future).
Given the frugality of RISC OS the performance the user experiences often exceeds what the raw MHz figure would suggest. Yes these (IOP) processors are not targetted at the home market - but do they need to be if (after all) the role as I/O processors become more demanding to the point where these ARM's become virtually desktop processors in all but name ?