There was
an interesting article today about the new Apple TV's processor. Essentially, the Apple TV now has a dual-core A5 with one core disabled. But that's not the interesting part; it's actually a fairly common strategy in dealing with a not-100%-reliable process, where one core may end up unusable. Those weird three-core AMD processors, for instance, are reportedly four-core parts where one core did not work correctly, and was disabled. All PS3 chips have one SPE (a glorified DSP; the Cell chip has a number of them) disabled, likely to improve yield.
The interesting part is that, unlike all previous A5s (and A4s), which were made on a 45nm process, this one was made on a 32nm process. In general, you'd expect an equivalent 32nm chip to have lower power consumption (and thus heat output) than the same design on a 45nm process.
The even more interesting part... When they looked inside a new iPad 2 (the cheaper one announced at the same time as the new iPad), they found a 32nm A5, with both cores enabled. It'll be interesting to see if the new iPad 2s have better battery life than the old ones; you'd certainly expect it.
And now the borderline conspiracy theory part. The A5X (new iPad SoC) looked at by Chipworks
was a 45nm part. But are all A5Xes 45nm parts? Surely, if you have a 32nm process available to you, you'd want to use it for your largest and most expensive, and hottest, chips? Is it possible that some iPad not-3s have 45nm SoCs, and some 32nm?
In
a now-notorious test, Consumer Reports found that the new iPad gets hot,
so hot that it could melt a cat well, warm; about 46 degrees, when playing a strenuous game and plugged in. They also found that the charger was unable to keep up with it in this condition; the battery kept draining. But, while some were able to reproduce at least some of this, others failed, with some tests finding it getting only marginally warmer than the old iPad 2, and cooler than many competitors. Could there be a 32nm version in the wild?
