x86 Processors: Is This Year The Entrée To The Era Of ARM?

This blog post references my feature article ‘x86 Processors: Continued Innovation Is A Welcome Contradiction‘ in EDN’s February 4, 2010 edition. It’s one of a series of web addendums to the print writeup.

My review of the Google Nexus One earlier this week reminded me of a piece that I’ve had planned since January’s Consumer Electronics Show, specifically on the handset’s Qualcomm Snapdragon CPU, and more generally on the processor’s ARM-based origins. CES marked the coming-out ceremonies for numerous ARM partners’ products; Nvidia showcased the latest dual-core Cortex A9 iteration of the Tegra line, for example, while both prior to and at the show Marvell unveiled a blizzard of families and proliferations (including a quad-core teaser). More recently, Marvell has kept the announcement treadmill smoothly humming, Freescale has discussed an eBook-targeting product, Texas Instruments (with OMAP4) and the partnership of ST-NXP Wireless and Ericsson Mobile Platforms both launched chips at Mobile World Congress, and of course Apple has alluded to an internally designed (or not) A4 SoC first employed in the upcoming iPad tablet.

Apple’s been mum on the details of its first not only branded but also (supposedly) self-developed ARM semiconductor offering, and we may now know why. EDN contributing blogger Steve Leibson alluded to 1 GHz clock-derived architectural confusion a month ago, Linley Gwennap suggested a possible Intrinsity source at that same time, and Ars Technica’s Jon Stokes more recently revealed that his confidential sources tell him the A4 is a relatively humble chip; a single-core Cortex A8 mated to an Imagination Technologies PowerVR SGX graphics-and-video core. Much of the I/O and other resources that might be found on a mass market SoC have been stripped from the Apple A4, thereby leading to a fiscally attractive smaller die size. And especially if Apple plans to proliferate the A4 beyond the iPad to future iPhones, iPod touches and other products, thereby securing sufficient production volume to further drive down costs.

CES also enabled me to check out innumerable mobile systems based on both x86- and ARM-architecture CPUs. Qualcomm’s Snapdragon was dominant in the latter case, and I must say that I’m mightily impressed with how quickly the company has entered the ARM space and ramped up product developments, design wins and shipments. More generally, my product perusal provided an opportunity for me to test my previously published hypotheses on the comparative success of each processor in notebooks (and netbooks), in cellphones and smartphones, in in-between products such as MIDs/UMPCs and tablets, and in dedicated-function devices such as eBook readers. I was quite relieved to discern that the abundance of data points closely aligned with the relative market ramp vectors I’d previously forecasted.

For notebook and netbook PCs, I remain convinced that x86 (and Windows as the corresponding O/S) will remain dominant in times to come, at the expense of ARM and Unix-based O/S alternatives. If it looks like a laptop PC, people expect it to work like a laptop PC, and for the bulk of the world that means a full-blown Windows experience. Low-to-no entry cost carrier-subsidized netbooks such as a $29 Verizon offer I heard about a few weeks ago are (as I suggested more than a year ago) an anomaly; the minimal-risk pricing entry point encourages folks to take the plunge on a system they might not otherwise consider, but I don’t seeing the practice being pervasively embraced by consumers.

The only factor in traditional computing that admittedly gives me pause is the economy-be-damned market share increase for Mac OS X-based Apple systems in recent years at the expense of Windows-fueled units, with yours truly as a case study. Here, I suspect, you’ve got a population of folks who are fed up with Windows-targeting viruses, and who are willing to endure the learning ramp pain of a new platform in order to flee and be free (at least temporarily) of malware. Again, thought, I doubt there are enough folks sufficiently motivated (and with sufficiently-sized wallets) to dent Windows’ market share to a substantial degree.

Conversely, in traditional mobile phone designs, I suspect that ARM will be dominant far into the future, if not forever. Again quoting Drucker’s Rule, which I most recently mentioned a week ago, "An emerging technology must be 10x better than the incumbent approach in order to have a reasonable likelihood of replacing that incumbent". x86 CPUs don’t come close to satisfying that criteria, even in the ascendant Moorestown era. The relative ease with which folks switch from one handset to another shows me that x86 code compatibility largely is a don’t-care here; whatever applications and functions are most important to consumers will be ported to whatever operating systems and microprocessors are pervasive. Although I’m a big fan of physical keyboards, I remain unconvinced of the mass market potential for MIDs as a mobile phone killer. The reputed lack of interest that Intel and TSMC have received for Atom core licenses is suggestive to me of the validity of my stance. And the only factor that gives me pause in this particular case is Intel’s partnership with Nokia; as I brainstormed last summer, are the two companies really working together on a SoC design, and if so where does Nokia plan to use it?

Speaking of MIDs, what about tablets? I’m not speaking here of so-called "convertibles", which adapt an otherwise conventional netbook or notebook PC design with a flexible hinge to a touch-augmented display and are therefore subject to the same prognostications I offered above. I’m talking about touch-only large-screen devices such as the iPad and Microsoft’s long-rumoured Courier (potentially pictured here). In this case, the field’s wide open, because the form factor is so revolutionary. Folks don’t have preconceived notions of what such a system should be or do, as in the case with traditional computing form factors, so operating system and processor options correspondingly expand. But as I suggested in my iPad analysis a few weeks back, battery life in both operating and standby modes is a critical consideration.

So I largely give the nod to ARM over x86 here, and the relative processor proportions in prototype platforms I saw at CES (the bulk of the ARM-powered ones also running Android as their O/S) suggested that my stance is on the right track. The same logic applies, I think, to dedicated-function devices such as eBook readers. With that said, I believe that history bears out my perception that such focused products have at best a limited lifetime. Most people have sufficient tolerance (and budget) for two widgets, one that goes in the pocket (the cellphone) and another that goes in the purse or satchel (the mobile computer). An eBook reader (for example, or a portable multimedia player) may deliver sufficient advantage (readability, battery life, etc) to justify its standalone existence for a while, but eventually one or both of the traditional form factors will suck up the function in the form of a software application.

p.s…Speaking of Google O/S-fueled tablet computers, and in acknowledgment of my longstanding desire to entertain my readers with videos on end-of-workweek afternoons, check out this Google Chrome O/S-centric proof of concept clip recently produced by the Chromium open source development team: