Mad Dog 21/21: Will You Still Need Me When ARM’s 64?
March 3, 2014 Hesh Wiener
At the end of February there was a big technology trade show, Mobile World Congress, in Barcelona. MWC attracted more than 70,000 visitors. It spanned mobile gadgets, electronic components, apps, and services–stuff for consumers, and stuff for the industry itself. This year’s unofficial theme was the transition from 32-bit to 64-bit processing, chosen last September 10. That is when Apple unveiled its 5S phone with a 64-bit A7 processor. Apple’s circuit, manufactured by Samsung, is based on ARMv8A architecture. It runs 64-bit iOS 7. And, for now, it defies competition.
In the iPhone, as in other computers, whether clients or servers, 64-bit technology pretty much removes any logical barriers imposed by limitations on memory. Physical memory, of course, depends on the configuration of the particular device. In a 64-bit environment, it becomes possible for programmers to hide most distinctions between main memory and data storage peripherals and to leave issues that might involve virtualization to hardware, firmware and systems software. The result can be an improved user experience.
In the case of the ARM processor family, the move from 32 to 64 bits provided other benefits, because ARM’s engineers got a fresh chance to optimize the instruction set, to add more registers that boost performance, and to smooth multitasking. Combined with glue circuitry, 64-bit ARM processors will take better advantage of advances in memory technology, peripherals, sensors, communications coprocessors and graphics engines. But client technology is only part of the picture.
ARM Holdings also wants its chips to be adapted by server makers. With 64-bit technology, ARM can foster the development of similar hardware platforms at both ends of the wire, something currently only available in the case of the X86. Similar hardware allows the use of similar software, as is the case in X86 country, where there are client as well as server versions of both Linux and Windows.
Many industry observers now see the processor market as a battle between ARM and X86 with other options, including Itanium, Sparc, Power, and System z, as pretty much irrelevant, legacy chips supporting legacy apps for dwindling customer bases. This view of the processor market weighs heavily on users of i, Unix, and mainframe software. Understandably, they worry that they are in some kind of technological backwater and that their companies and their individual careers may suffer regardless of the quality of computing they deliver day in, day out via their legacy boxes.
Apple, by contrast, is relentlessly moving ahead, encouraging its wealthiest and most enthusiastic customers to adopt the latest products.
Last October, a month after the debut of the iPhone 5S, Apple announced the iPad Air and iPad Mini tablets, also powered by A7s. The chips used in Apple’s tablets are packaged differently than those in the iPhone. They have more pep but they use more power and generate more heat. They support the same 64-bit iOS 7 operating system. They run the same apps.
Still, Apple has always taken care not to disenfranchise users of its products who don’t have the latest gadgets, and it did so again as part of its most recent Christmas season product rollout.
Ten days after announcing the iPhone 5S, Apple said it would market a new smartphone based on its prior generation engine, the A6 processor. The 5C, like the 5S and Apple’s latest tablets, runs iOS 7, but does so in 32-bit mode. It has most but not all the features of its predecessor, the iPhone 5, plus a better back camera and a lower price. It is Apple’s long goodbye to the 32-bit world.
Apple’s front line is 64-bit, but its clear promise to support a new 32-bit product running a version of its latest systems software reinforces one of the great strengths of the company product strategy. Apple tries very hard to protect past customers from undue erosion of the value of their devices. It has succeeded admirably at this, preserving customers’ investments in iPhones, iPads and other products not only in economic terms but also in technical terms. Apple’s customers have hardware that they know will last and that retains high residual values, allowing customers to exercise a trade-in, trade-up strategy. A byproduct of Apple’s product preservation plans is a large base of users with iPhones that are one, two, and three generations old, who feel that they are as much a part of Apple’s world as people using the latest 5S or iPad Air.
Apple is not alone in this. Amazon has done an outstanding job preserving the value of its Kindle Fire e-readers and tablets. By contrast, Windows does not provide very good legacy support, as customers with ancient apps often find out, to their dismay.
Nevertheless, everyone knows that it is much nicer to have the latest Apple device than an older one. The most powerful reasons for this, apps that excel on 64-bit hardware, have yet to emerge, but they will. There are bound to be some killer iPhone and iPad features or apps that pretty much demand entry into the new world, features as compelling as Siri.
Apple’s rivals, companies that offer smartphones and tablets that are for the most part based on ARM and Android, know they cannot stick with 32-bit technology, even though they are yet to find any specific reason to move up to 64 bits. So, at Mobile World Congress, vendors showed off their latest 32-bit offerings but kept on talking about their hopes and plans for a new generation. They are very nervous, because there are no 64-bit ARM processors ready for use in mainstream products, just promises, at least not from the kinds of vendors the big phone and tablet makers turn to. Soon, though, Qualcomm, Nvidia, and Samsung will deliver in commercial quantities. In addition, Intel is trying to persuade vendors to put their 64-bit products on its Atom architecture chips instead of ARM.
But there is yet no sign that Google is ready to deliver a 64-bit version of Android and Microsoft, which would love to grab some market share in the mobile device game, has been frustrated by a market that doesn’t want phones or tablets unless the devices have the kind of vast choice of apps that iOS and Android provide.
And then there’s the 64-bit ARM server world, which, when I looked just now, was missing something: commercial-grade operating systems. There isn’t a server version of Android. There are some Linux distros that could run nicely on 64-bit ARM boxes, and Red Hat and Canonical are working on them. But the mainstream commercial market, the market that for servers at companies that are not themselves dot-com type outfits, has yet to find a credible source of software. And of course, there is no 64-bit server chip available on the market yet unless you are talking about development kits from Advanced Micro Devices and Applied Micro.
There are two very good candidates for providing ARM systems software, however, yet both of them have good reasons to avoid the ARM race. The two are, of course, Microsoft and IBM. Windows on ARM servers could be a commercial success, but ARM and X86 are different enough to present Microsoft with quite a challenge if it wanted to do a first class job on both platforms.
IBM is not wedded to X86. That would be easier than its tie-in with Power and its wealthy cousin, z mainframe. The conflict would not be so much an issue with customers, many of whom don’t particularly care what kind of hardware IBM puts under its logo, just so the software runs. This is particularly true in the case of IBM’s most traditional products, i and z.
IBM i customers have no reason to be concerned about the hardware under their systems. If IBM offered both Power and ARM iron, with ARM at the low end and Power at the top, customers might well rejoice, because IBM-brand ARM servers would undoubtedly be priced to compete with the ARM servers that other vendors will be rolling out. The reception ARM hardware might get from IBM’s mainframe users would be even more enthusiastic. Mainframe shops have shown, time and again, that that they are perfectly happy to use alternative hardware whenever IBM lets them, just as long as the stuff works and the price is right.
Power Systems users might not rush to embrace ARM boxes running AIX, should IBM offer such products, but if IBM offered ARM servers for small i shops, it would very likely offer a port of AIX, too, and very possibly an IBM-on-ARM distro of Linux that included hooks to make superior linkage to legacy mainframe and Power back ends.
Of all these possibilities, the i opportunities might be the best for IBM and for customers, too. ARM-i hardware could give IBM its first ultra-small business system since the days of the System/34. IBM would have to find a way to price software and services at levels that were competitive with onsite Windows/X86 systems and cloud alternatives. But the low cost of ARM technology, the enthusiasm that will undoubtedly accrue to a number of 64-bit ARM servers as they hit the market, and the vast base of small businesses that would prefer a friendly, inexpensive computing solution that is not in the cloud suggest that ARM hardware will spark a boom. And, it turns out, IBM i software is just about the only operating environment that could be easier for low tech small businesses to run than Windows or Linux.
It’s a shame IBM suffers from a chronic failure of imagination. Which, we’ve noticed, starts with i.