Apple took three cores for winning benchmarks
In Soc development, Apple's mantra appears to be "low frequency". Apple is reasonably happy to sacrifice die area to achieve that.
Hence their swift cores are quite big (in relative terms), but launched @ a lowly 1Ghz, and then later @ 1.4Ghz. And of course they went 64-bit architecture with cyclone ahead of the rest,again gaining performance at the low-ish clocks (1.5Ghz, when Nvidia/Qualcomm are clocking 1Ghz higher), at the expense of die area.
Need more performance ?, lets not redesign for higher clocks, rather lets use more die space and install another CPU core.
Apple have historically had this same approach to their implementation of IMG's GPU IP. When everyone else was clocking multi-core series 5 543/544 @500Mhz+ (including Intel), Apple rather went with the bigger 554MP4 design in the ipad4, and clocked it at 280Mhz. Apple were the ONLY licencee of that SGX554 GPU core.
Interestingly, unless they are doing something really special with that A8X that has yet to be revealed, I don't see that trend continuing with the series 6 GPUs from IMG.
I am guessing when you are getting unbelievable high margins from end devices, and have multiple other sources of related income (services,apps,music) you can be comparatively price agnostic in ih-house soc design decisions, and when you can be, it looks like the best socs are ones that use Apple's design philosophy of using die space over frequency
When on the other hand you are selling chips and chips alone, die area becomes a much more crucial factor in the soc design process.
