Apple acquires PA Semi; Press proves lack of understanding. [UPDATED]

Wednesday 23rd April 2008, 02:00:00 PM, written by Arun

It's official: the iPhone's future CPU will have a TDP of 25W! Or so mindlessly speculate Forbes and a variety of other websites following Apple's stunning acquisition of P.A. Semi, a manufacturer of power-efficient PowerPC-based CPUs.

Of course, it should be obvious that such a claim is absolute nonsense. The company's dual-core processor, which is manufactured on a 65nm process (likely at IBM), sports typical power consumption of 5 to 13W at 2GHz. Even if shrunk to 45nm, stripped of its second core, synthesised for a lower target frequency, and binned even more aggressively for power efficiency.. It'd still be take substantially too much power for a mobile phone. The idle power consumption of 1W would also need to be reduced by more than two orders of magnitude.

So clearly this acquisition isn't for the iPhone or the iPod, except perhaps in the very-long-term with a different processor architecture developed by the same team. Which doesn't seem very likely either; they'd need to integrate multimedia capabilities on the same chip and so forth, so effectively Apple would be designing the entire system except the baseband. They'd be competing with a proprietary solution against every single handheld SoC manufacturer in the world. Once again, nonsense!

So what's the point behind the acquisition, then? Well first of all, it likely represents a complete rejection of Intel's Atom architecture for ultraportables and mobile devices between the iPhone and the Macbook, such as a previously rumoured Apple Tablet PC. Secondly, it also represents a possible rejection of Intel for more embedded applications such as the Apple TV. And finally, it might mean trouble for Intel in the Mac Mini and Macbook Air segments of the market.

Those who claim this represents a 'missed opportunity' for Intel with its Atom architecture are completely missing the big picture; this isn't just about new business Intel might have captured. It's also about existing business they're likely to lose now, and obviously the financial consequences of that are more problematic.

One intriguing question, of course, is who's going to deliver the multimedia and connectivity capabilities for future P.A. Semi-based Apple products. After all, Apple still doesn't have any Southbridge or GPU capabilities. P.A. Semi claims their CPU integrates both the northbridge and the southbridge, but that's not really true; it's really a SoC aimed at embedded markets, which don't need SATA or USB or PCI or ... - and they pretty clearly don't have that IP either.

The only two companies manufacturing modern southbridges and who don't have a x86 license are NVIDIA and SiS. However, that's not the only alternative. For example, they could use a third party chip for SATA and integrate USB (or use yet another third party chip for that, or request a custom chip combining the two). So clearly a traditional southbridge isn't the only possibility, although it does remain relatively likely. In terms of GPUs, they can use a single-chip GPU+Southbridge from NVIDIA, or a discrete solution from NVIDIA or AMD (such as in the Apple TV). Or they might even license PowerVR's high-end 3D and Video IP cores if they really want to do more in-house.

Note that we're not saying every single press coverage of this story got everything wrong; that's obviously not the case, but the vast vast majority did and most didn't even go into any detail, so hopefully this rectifies that.

!UPDATE4! [24th of April]: The Wall Street Journal quotes Steve Jobs as saying the deal is indeed for the iPod and iPhone, but from a talent perspective and to a much lesser extend a technology perspective (and certainly not for their current or upcoming products based on PWRficient). So indeed, much of the speculation above is simply incorrect; what is actually correct is the scenario I proposed in paragraph 3, which I described as 'once again, nonsense!' - clearly Apple doesn't seem to think so. As we said, it seems that Apple might have been doing semiconductor R&D substantially before this acquisition, which goes completely against common wisdom. So you'll have to excuse us for not properly taking into consideration data that has, quite simply, never been made public and very few people knew about - and which would still massively surprise just about everyone if all of its implications are indeed what we think they are.

UPDATE3: Looks like it might be the case that Apple did do some semiconductor R&D in-house even a long time before this acquisition, especially for handhelds. So it is possible (but very far from certain) that they are indeed only interested in the engineers and patents, not the  architecture. In which case the mainstream press coverage would still be wrong, btw: obviously PWRficient is never going to be used in the handheld space. Clearly the picture is getting muddier by the minute...

UPDATE2: A new article at makes the big picture substantially harder to understand: apparently, P.A. Semi isn't even sure whether they'll be able to renegociate a PowerPC license, and Apple was "not interested in the startup's products or road map, but is buying the company for its intellectual property and engineering talent".

However, it's worth pointing out they do not explicitly say they are not interested in P.A. Semi's current architecture; so, it may simply be that what Apple wants to do with it (i.e. the chip roadmap) is different from P.A. Semi's previous plans. It would seem extremely strange for Apple to acquire a company with so many short/mid-term contracts and design wins (see: EETimes article) just to put it back in stealth mode for several years to work on a new architecture.

UPDATE1: It has been pointed out to us that P.A. Semi has been working on a solution with integrated SATA and USB. This was mentioned for example by The Inquirer in February 2007, although obviously it got delayed since then. It would seem to make the 'discrete GPU' option by far the most likely. The article also claimed the company was working on a quad-core derivative, that might be appealing for future server products if Apple keeps it on their roadmap.

Discuss on the forums


apple ± intel, silverthorne, arm

Related apple News

3G iPhone using same 90nm Application Processor as 2G iPhone?
iPhone 3.5G to support 1080p H.264 High-Profile Dual-Stream Decoding?
Apple picks AMD Radeon HD GPUs for iMac
EA games for Mac announced at WWDC not native
NVIDIA and Intel power the Apple TV