> As for Atom for Apple: > > Price. Intel can't make money on an Atom based SoC shipping at $25. > Performance. It is unlikely that Atom @22nm would beat Arm @28nm across a wide > range of benchmarks. > Power consumption. It is unlikely that Atom @22nm will be competitive (even against > 45nm parts, let alone 28nm). > Apple control. Why would Apple give both ISA control and foundry control to Intel?
Price: 32nm Cedar Trail parts are in the low-40s range, wouldn't be surprised if Intel manages to get 22nm in the low to mid-30s or even lower to get into the same price range.
Performance: Seems like a huge guess on your part since there are no CPUs out there to benchmark.
Power Consumption: Did you notice "Near Threshold Voltage" demos at IDF SFO? Not sure when Intel can/will apply NTV to production CPUs, but I am sure they are driving hard to get it done. And check out the power slide at this URL. http://www.anandtech.com/show/4345/intels-2011-investor-meeting-intels-architecture-group-14nm-airmont-atom-in-2014 Medfield power draws are already in the lower part of the range of ARM-based SOCs...so I don't think Intel is too far behind. Even more interesting is the accelerated roadmap that Intel is moving towards for Atom...they plan to have the same node-level process for Atoms as for their advanced Core CPUs by 2014. That should make Apple and other OEMs sit up and take notice. If they feel Intel can achieve this, they better consider moving to the architecture that is going to take a 2-3 year process lead.
Apple control: Yes, this is the biggest thing that will hold Apple up...but if it becomes evident that x86 architecture will take the lead in power/performance, Apple may be compelled to come around. Over time, I think pragmatism will trump hubris!