Opinion – Apple’s confirmation that it would acquire PA Semi prompted a wave of speculation that Apple has decided to drop the idea of an Intel Atom iPhone and go instead with PA Semi. But I wonder: How can Apple consider using a processor design that uses more than 12 times the power of Atom? Why use a processor design that is not nearly small enough to fit even in MIDs? And how can Apple drop the idea of an Atom iPhone, if Atom has never been considered for the iPhone in the first place? There must be a completely different idea behind the PA Semi acquisition.
It is easy to get excited about acquisitions that Apple makes. They don’t acquire very often and if an acquisition happens, well, yes, it is something special and worth to look at. But this one is blown out of proportion.
What struck me is the fact that everything-Apple is about the iPhone these days and if Apple makes a move it is concluded that it must be about the iPhone. Today, it is a general conviction that PA Semi will provide the architecture for the next-generation iPhone processor. Seriously? Has anyone spent some time looking at PA Semi’s architecture?
PA Semi’s product line is currently limited to a dual-core, 2 GHz PWRficient 64-bit processor. Based on a PowerPC design (enabled through an IBM license) The platform includes two DDR2 memory controllers, 2 MB of L2 cache as well as a flexible I/O subsystem-is that is tailored towards “high-performance computing and embedded applications”. The fabless developer never aimed its design against ARM or ultra-mobile devices and has very few features that make it suitable for such applications. Instead, this chip was positioned as a competitor against embedded versions of Intel’s Core 2 Duo and AMD’s Athlon X2 processors.
“Because the 1682M achieves dramatically better performance per watt than do competing multigigahertz microprocessors, equipment manufacturers can use the processor as the basis for products that are cool and efficient yet high in performance,” PA Semi states on its website.
Consider that the current iPhone runs on a 620 MHz ARM processor and is rated at a power consumption somewhere between 180 and 450 milliwatts. It is a fairly small design with very little heat dissipation, neatly fitting into the iPhone’s form factor. The 65 nm PA Semi chip is several times larger and uses more the 50 times the power of the current ARM chip (up to 25 watts, 5-13 watts typical).
At least in its current state, the PA Semi chip is the wrong design, it is too power hungry and too large to even fit in the iPhone. And I haven’t mentioned yet that 2 GHz and dual-core is overkill for a smartphone anyway (the PA Semi design scales to eight cores and 2.5 GHz).
So what about the claims that Apple disses Intel’s Atom processor? Well, they really can’t because Apple and not even Intel have intended to use Atom for the iPhone. Intel itself says that Atom in its first generation is too large (even at a size of the size of your finger nail) and uses too much power (Atom tops out at 2 watts). Instead, Intel aims to position the 2010 Moorestown SoC (which includes a graphics engine) as a potential iPhone candidate. Given the fact that Intel has done 4 years of research in this field already and will have 32 nm manufacturing in place for Moorestown, this could be the best product on the market at the time and an easy choice for Apple.
On an application basis, Intel’s x86 approach has the edge in my personal opinion, but, of course, Apple’s Mac OS X-based applications could easily be ported to the PA Semi design as well.
But why would Apple try to go out of its comfort zone and try to compete with Intel in that space? To me, it would be very surprising if they have decided to do that. Apple is more of a consumer electronics company than an IT vendor and I find it very unlikely that Apple will now compete against Intel, AMD and possibly ARM.
PA Semi could have many other implications. After all, it is a processor for embedded applications. What about an Apple TV powered by this chip? Whatever it will be, Apple will have a lot of work to do with this design and results may not show up for several years.
It just won’t show up in an iPhone. I am quite certain of that.