Hypothesis or not, If the PC and Intel are joined at the hip, then Intel is hobbled, with Fiscal 2013’s 10 percent drop in PC sales marking “the worst decline in PC market history,” according to Gartner.
ARM seems to give the impression to have picked a better partner. Its chip designs have found their way into a staggering 99 percent of mobile devices, which in Fiscal 2013 saw increases of 50 and 38 percent in sales of tablets and smartphones, respectively.
Picture This! It tends to sound like a classic David-and-Goliath saga: Intel is the reeling giant clinging to its crumbling PC pillar, while ARM is the sprightly mobile player with perfect aim that owns the future of computing. But then again there are problems with that narrative.
Let’s start with the size at the first place. In the last 12 months, ARM pulled in a little over $1.1 billion in revenue, while Intel netted nearly $53 billion. The two companies could scarcely be more different: Intel is vertically integrated, both designing and manufacturing its own chips, while ARM develops intellectual property, period. ARM licenses its processor designs to such stalwarts as Qualcomm, Nvidia, or Samsung, which in turn contract with companies you’ve never heard of — such as GlobalFoundries, United Microelectronics, and TSMC — for the actual manufacturing.
As of October 2013, you can add Intel to that last list. That’s right — Intel actually manufactures ARM chips.
So ARM and Intel are not direct competitors in the classic sense, though it’s absolutely true that ARM’s approach to the mobile market has been vastly more successful than Intel’s. Moreover, although PCs may be on the decline, Intel x86 processors still own the higher-margin server market, where ARM has barely entered the room. It’s also clear that the mobile game is not over, as Intel gears up with a fresh line of mobile processors and unique new advances in processor manufacturing.
Why ARM has triumphed in mobile
ARM-based chips have dominated the designs for mobile processors in much the same way that Intel’s x 86 families of processors locked up the Windows PC market a few decades ago.
A big part of ARM’s success, however, has little to do with better technology. It’s the company’s business model. Whereas Intel builds chips, ARM only designs them, leaving the physical fabrication to its customers/licensees. Whether by accident or shrewd planning, ARM pretty much launched the idea of semiconductor intellectual property licensing.
That was important to Nokia, Ericsson, Motorola, and other early adopters of the ARM processor architecture because those companies were making newfangled devices called mobile phones that needed to be small and run on batteries. All of the traditional microprocessor companies’ chips were too big to fit, and they were designed to run on AC power. Some needed fans or heat sinks to stay cool.
What sealed the deal, however, was that mobile device makers needed to design their own ASICs intended to power specific devices — and realized it worked better to integrate a CPU into those ASICs rather than soldering a CPU alongside.
Enter ARM and its CPU-for-hire business model. Want to mix a CPU in with the other ingredients of your new ASIC or SoC? ARM has your recipe. Like few other companies at that time, ARM was willing to let customers bake its CPU into their own devices.
From there it was all momentum. Once ARM gained a toehold in the mobile market, others followed. As with PCs, servers, and most processor-based systems, later products were based on the same processor family as the earlier products. After all, why switch? If the previous generation worked, make the next generation similar but faster. Gutting a design and starting over with a new CPU family, including new software and new development tools, is a surefire career killer.
ARM’s efficiency edge
The fact that ARM’s CPU architecture was comparatively simple made it a good choice for integration. A more complicated CPU would’ve been difficult for outsiders to get right — and might not work.
Imagine a skilled but inexperienced engineer trying to fabricate his first Pentium processor based on nothing more than schematics. How well do you suppose that would work? The early ARM designs, in contrast, used static timing and simple clock trees, and they had no dynamic state. The entire CPU could be stopped dead (0 MHz) and still work when it sped back up. That was a huge boon for ASIC designers trying to concentrate on their own logic, not someone else’s.
ARM uses a RISC architecture for its chips. RISC is perceived as more modern, more elegant, and more appropriate for new designs than CISC. But the differences today are more semantic than technical. RISC was supposed to be a streamlined, stripped-down approach to CPU design. Jettison all those messy old instructions you’re not using, the thinking goes, and what’s left is a lean, mean computing machine.
The original RISC designs were indeed very Spartan. That stopped almost immediately. Little by little, RISC processors like MIPS and SPARC started clawing back many of the features they’d eschewed earlier. (Hardware multipliers? Gosh, maybe that’s a good idea, after all.) ARM, like most other RISC-inspired designs of its day, started out simple and gradually grew more complex. Still, it’s cleaner and neater than the barnacle-encrusted hull of the Good Ship X86.
As it turns out, few anticipated the real payoff of this simpler design: lower power consumption. Contrary to popular belief, RISC processors are neither faster nor slower than their doddering CISC forebears. But they are generally more power-efficient, which turns out to be a key differentiator when you’re making battery-powered gizmos.
It’s tough to compare one CPU circuit design to another, but roughly speaking, an ARM design has about one-third the number of transistors of an x86 design. That’s leaving out the likes of cache and bus interfaces, which can easily consume more transistors than the CPU “core” itself. Indeed, most processors today — RISC or CISC — are about three-quarters cache, with a little CPU core lurking in one corner of the chip.
ARM can get away with slashing its transistor budget because it doesn’t support as many architectural features as x86. It has a smaller repertoire of functions — a smaller assembly-language vocabulary, so to speak. Intel chips can handle BCD-encoded numbers; ARM does binary only. Intel has a built-in loop counter; ARM relies on software-defined loops. Intel enforces and defends boundaries on code, data, and stack segments, and can automatically trap accidental (or malicious) violations. ARM processors have nothing like that.
All x86 processors since the ‘386 have implemented a four-level privilege hierarchy intended to prevent low-level code from contaminating high-level code. This is the kind of stuff complex operating systems do in software, yet the chips implement it entirely in hardware, no code required. It’s a remarkable feature, but it requires several million transistors to make it work, and that means power.
The list goes on and on, but the features that the x86 architecture has accumulated over the past three decades don’t always show up in the benchmark results. System-level functions help with operating-system code or with obscure corner cases, or they were added to enhance security. All of those add-ons make the chips bigger, hotter, and more expensive.
As ARM finds itself designing processors for ever more complex systems like multicore Android tablets, microservers, and 64-bit machines running hypervisors and multiple secure operating systems, it has to add in more big-boy features that it had originally omitted.
Virtual memory? Check. Security features? Check. Wide registers, floating-point extensions, pseudo-DSP instructions? Check, check, and check. Little by little, ARM is leaving behind RISC philosophy and becoming more like the complex processors it upended. The company has redefined its instruction set at least twice, sometimes sacrificing binary compatibility on the altar of performance, capability, or scalability.
Going with what they know
As mobile devices become more and more complex, does that leave an opening for Intel? Probably not, at least not beyond a certain point. For one thing, Intel is late to the party, and we’ve seen how inertia drives this business.
As it stands, Intel’s mobile Atom processors are barely as good as the leading ARM-based alternatives, and the ARM army has an obvious head start when it comes to software, ecosystem, and experience pool. It’s a tough sell to convince an engineering team to change its processor family, software, and development tools in order to buy a chip that’s “barely as good” as the one they’re already using.
By the same token, ARM in the PC business makes no sense at all, at least if you define “PC” as a system running Windows applications. People don’t buy a PC for the lovely Windows user interface; they buy it for access to their existing PC applications. The ignominious fate of Windows RT has proven that.
ARM in servers makes a bit more sense, mostly because a server doesn’t run very much shrink-wrapped code or have a user interface that anyone but its handlers will see. After all, Linux rules the data center, and Linux can run on ARM processors. Plus, there’s increasing demand for low-power solutions in data centers, because over time the cost of power can outstrip the cost of the hardware itself.
But here again inertia rears its head. Most servers are x86-based and there’s presently no compelling reason to switch. If ARM-based server chips manage to offer a significantly better price/power/performance ratio, they can probably make inroads. Certainly, some new server makers are eyeing ARM-based designs eagerly, if only because it sets them apart from their generic x86-based competitors. But it will be a tough slog.
Intel’s manufacturing advantage
Intel is traditional smokestack industry: Everything from design to manufacturing to sales is done under one roof. ARM is more like a downtown architectural firm, doing white-collar work and licensing its blueprints to others. ARM doesn’t make chips, and Intel doesn’t collect royalties.
That means Intel chips come only from Intel, whereas ARM-based chips could come from any of a number of different vendors — theoretically. In reality, every ARM-based chip is unique, and there are no second sources. That makes ARM processors every bit as vendor-specific as Intel’s or anyone else’s.
Intel’s emphasis on manufacturing means the company is fanatically focused on volume. When it costs upward of $5 billion to build a new chipmaking plant that will be obsolete in just a few years, you don’t need an advanced economics degree to see that amortization is an enormous burden. Every single chip is weighed down with the crushing cost of its birthplace.
Thus, Intel is all about high-volume products — not specialty niche products. That excludes Intel from a lot of narrow (read: emerging) markets. It also prevents the company from making special spin-off processors with unique I/O, different configurations, or application-specific features. One size has to fit all.
That $5 billion investment has its benefits, however. It pays off handsomely in terms of faster speed, smaller die size, and lower power consumption. The laws of physics apply equally to everyone, regardless of their processor architecture, and Intel’s current 22nm semiconductor lithography is a full generation ahead of most everyone else’s. Smaller transistors mean smaller voltage swings, shorter signal-propagation times, and less heat to dissipate. It also means smaller circuit dimensions, which means smaller die sizes, which translates into more chips per silicon wafer. And that means more chips to sell.
Intel is one of the few companies left with the financial resources to invest in state-of-the-art manufacturing R&D. Everyone else — including all the ARM licensees — have to make do with shared manufacturing, mainstream technology, and less-aggressive physics.
Slugging it out
Intel and ARM couldn’t be more different commercially, but their respective architectures are gradually converging technologically. After all, computer scientists have been studying and simulating the best ways to design a processor for decades, and those same findings apply to everyone. ARM has grown more CISC-ified over the years, while Intel’s x86 has become streamlined in other areas.
As it stands, Intel’s manufacturing prowess very neatly offsets ARM’s architectural advantages, at least in terms of energy efficiency. In the performance race, Intel is the undisputed leader, but that may be because ARM doesn’t want to design ultrafast processors, not because it can’t.
To break ARM’s inertia-driven upgrade cycle and crack the mobile market, Intel needs a new product category. That may come in the form of low-cost Android handsets now being fielded by a handful of vendors. These new “starter” handsets have little to no established software ecosystem to protect; they’re virgin territory, apps-wise. Android itself runs on x86 as well as ARM (or MIPS), so Intel’s Bay Trail and forthcoming Cherry Trail devices might gain a toehold. An all-in-one SoC with x86 core, wireless interface, and integrated peripherals could make a good single-chip handset platform if Intel prices it aggressively enough.
At that point, history may repeat itself, as Intel sneaks up on ARM from the low end of the market. It would be an ironic twist in a long-running, high-stakes game.