Intel Advances on Multiple Fronts Against ARM
If you’ve been developing embedded systems for a while, chances are you’re working with either an ARM or 8051 device, the latter being Intel’s bequest to embedded developers before abandoning them over 20 years ago for the more lucrative PC market.
ARM, of course, went after the embedded market from the beginning with a low-power, high performance RISC core. Today Intel has over 90% of the market for PC processors but less than 2% of the embedded market, where ARM is the king of the mountain. That’s about to change.
Build It and They Will Come
Unable to gain a foothold in the wildly popular handset market, in 2006 Intel set out to create a market for its powerful, if power-hungry processors—the Ultra-Mobile PC, which mostly ran the powerful, if power-hungry Windows XP; unlike handsets, UMPCs can hold batteries large enough to tolerate power-hungry processors. The UMPC turned out to be a flop, but it’s been resurrected recently as the Mobile Internet Device (MID). Thanks to the (wildly popular) ASUS Eee PC—based on Intel’s new Atom processor—this category is really taking off, even bifurcating into MIDs and Netbooks.
The betting to date has been whether Intel could flog that category hard enough for it to take off and be a vehicle to sell Atom chips. Here they’ve clearly succeeded.
With the PC market maturing, and with Intel’s market share there about as high as it can go, the company needs to get back into the embedded and portable markets, which is where the action is. ARM is clearly in their cross hairs.
ARM’s currently shipping Cortex-A8 beats the current generation of Atom hands down in terms of low-power performance. The next generation Atom may leapfrog the A8, but ARM’s follow-on application processor—the Cortex-A9—will still be ahead. However, ARM’s inability to run Flash—critical for displaying web sites—has given Intel more breathing room before A9-equipped chips ship some time next year.
Attack on Multiple Fronts
Over the last few months Intel has quickly made four moves that clearly demonstrate just how serious they are about getting into the mobile and embedded markets:
1. Intel partners with TSMC on Atom processors
On March 2 Intel announced that it was porting Atom CPU cores to TSMC’s technology platform, including processes, IP, libraries, and design flows. According to Intel, the collaboration is “intended to expand Intel's Atom SoCs availability for Intel customers for a wider range of applications.” It also takes direct aim at ARM’s embedded market, making it easier and more affordable for system engineers to incorporate Atom processors into their designs.
2. Intel Buys Wind River
On June 4 Intel announced its acquisition of embedded software vendor Wind River Systems. Intel was quite explicit that this was “part of Intel's strategy to grow its processor and software presence outside the traditional PC and server market segments into embedded systems and mobile handheld devices.”
The Wind River acquisition isn’t without risks. It does get Intel into markets where it has long struggled, including consumer electronics, automotive, medical and defense. But the acquisition is likely to scare off some Wind River customers, many of whom compete directly with Intel. Going forward will VxWorks continue to support ARM, Freescale and TI processors as adequately as Atom? Expect Integrity, ThreadX, Nucleus and other RTOSs to gain market share.
Even with Wind River in its camp, Intel’s ecosystem is still heavily skewed toward PC applications, in contrast to ARM’s, which is all about embedded and mobile devices. All told this was an excellent move, but it’s hardly checkmate.
3. Imagine, ARM and Intel
On June 19 Intel increased its investment in Imagination Technologies, the leading graphics processor IP maker outside of ARM. This had to be a particularly threatening—not to mention galling—move to ARM, who long incorporated Imagination Tech’s IP for graphics processing into ARM cores before dumping them to develop their own graphics IP, Mali. Imagination Tech’s POWERVR graphics cores are the leading alternative to ARM/Mali; licensees include TI, Marvell, Freescale, Renasas, Samsung, NXP and Intel.
This move essentially counters ARM’s advantage in graphics, an area where Intel has long been weak. Imagination Tech’s POWERVR SGX520 is currently available in TSMC’s 65LP process, making it straightforward for SoC designers to integrate with the forthcoming Atom IP.
4. Intel and Nokia
Last week Intel and Nokia signed an agreement that will enable Intel to license Nokia's HSPA/3G modem technologies with the aim of developing “advanced mobile computing solutions that deliver a powerful and flexible computing experience”. While long on implications and short on details, the move is a powerful shot across ARM’s bow.
Intel on the Move
Intel’s move into the handheld and embedded markets is a predictable response to Deep Throat’s dictum, “Follow the money.” The handset market in particular is much too lucrative to ignore, even if the margins on processor chips don’t begin to match those in the PC world. But with a near monopoly in the mature PC market, Intel had little choice but to make the move.
Intel’s acquisition of Wind River adds strong underpinnings to its software infrastructure. Its further investment in Imagination Technology buttresses another critical area in which it has been lacking. Both actions go against Intel’s historical corporate “not invented here” culture, which has apparently been bested by the need to move as quickly as the market.
For the past several years semiconductor companies have been focusing more and more on becoming systems providers, not just chip makers, in order to capture sockets. OEMs and IDMs are now driving the consumer electronics market, and semiconductor manufacturers are forced to supply entire system-level solutions.
Intel’s partnership with TSMC is its most radical move, a complete turnaround on its traditional business plan. Intel is a chip maker, not an IP supplier—but that was then, this is now. Intel’s business plan for Atom looks to be identical to ARM’s, and ARM’s licensees will have a choice when they go to TSMC. Intel and ARM are now both working from the same playbook. It’s a high stakes game, and Intel is a well-financed and focused player.
ARM has been extremely successful in the face of relatively weak competition. That is to take nothing away from their brilliant design and marketing teams, which have long since made the ARM architecture an industry standard. ARM is still a good generation ahead of Intel’s Atom architecture, and—if they continue to execute well—likely to remain so for some time.
But the elephant has now entered the room. The game has just gotten a lot more interesting. Stay tuned.