Clients looking like servers -- who like clients who like servers
To those nay-sayers I would reply that there are two aspects to the argument that ARM's A15 can win for the company in both its traditional applications with the likes of Nokia and with servers and wireless infrastructure.
The first point is that the A15 is a core not a processor and is itself a configurable set of potential processor cores. The silicon partners themselves – Texas Instruments, Samsung, ST-Ericsson and no doubt others to follow – will choose where in that cloud of potential A15s they wish to pin their instantiation and set their performance. The chip makers could produce a dual-core or four-core A15 processor optimized to work at about a 1.5-GHz clock frequency and suitable for next-generation client computers. Or it could be a 2.5-GHz many-cored and high value processor designed to serve pages.
The second point is that there is a convergence of requirements across the computing sector. The mobile sector wants to gain access to the rich processing abilities traditionally associated with the desktop adding multimedia and HDTV to voice. And the clients of tomorrow will have to take on the functions of today's infrastructure even down to such things as deep-packet inspection (see Clients to become more like servers, says ARM's CTO). But all this must be done without decreasing battery life even as we are on those mobile devices more and more of the time.
And on the infrastructure side the population and the manufacturers have become aware, relatively suddenly, that they are power limited. They'll take more performance if it comes but most of all they want more power-efficient computation.
Indeed it can be argued that the main thing that characterizes a server processor is the large memory spaces and very high speed interfaces that are necessary to move massive amounts of data around. These are things that ARM's semiconductor partners will be able to add to an A15 fabric as their value adding contribution.