backtop


Print 64 comment(s) - last by marsovac.. on Jan 9 at 7:56 PM


  (Source: Marvel and DC Comics)
Meanwhile Intel strikes back with smartphone chips

The war between Intel Corp. (INTC) and the alliance of chipmakers using ARM Holdings, Plc.'s (LON:ARM) titular core designs is heating up.  Intel is the world's largest personal computer CPU maker, while ARM Holdings is the largest core licenser on the general CPU market, which includes everything from the chip in your dishwasher to automotive CPUs.  The stage is set for the pair to duke it out at the 2012 Consumer Electronics Show.

I. Intel Bets on Its Process Technology

At CES the hot topic for Intel will be Medfield.  Medfield is Intel's crack at an ultra-mobile x86 system-on-a-chip, a type of design that incorporates a CPU, GPU, RAM, and I/O controllers all into a single die.  Intel's hope is to eventually cut the platform's power consumption low enough to put it on smartphones [source; PDF].  

Intel CEO Paul Otellini is expected to unveil smartphone designs in his keynote on Jan. 10.  The pressing question is whether this technology is production ready, or still in the prototype phase.

Even if it doesn't achieve the goal of production smartphone chips in 2012, it has a good shot at achieving it in 2013 when its 22 nm 3D FinFET transistor technology is brought onboard Medfield's successor.

Jim McGregor, chief technology strategist for research firm In-Stat, in an interview with Bloomberg emphasizes how critical this push is for Intel, stating, "For Intel, it's a 'we have to be there.'  Never bet against a computing device that fits in your pocket. I do more on my smartphone than any other device."

Intel's x86 is a complex instruction set computer (CISC) architecture.  Intel has struggled to match ARM -- a reduced instruction set computer (RISC) architecture -- in power efficiency.  

Intel's first foray into the mobile space -- tablets based on its Moorestown (Android-aimed) and Oak Trail (Windows-aimed) -- was largely a flop with very low sales over 2011.

But Intel looks to be much more competitive soon.  Two things have become increasingly clear in this more competitive market.  The first is that despite all Intel's work ARM is likely to maintain a slight power edge due to advantages of its architecture.  The second thing is that Intel is on the verge of gaining a substantial an edge of its own, through its dedication to developing the best process (chip manufacturing) technologies on the market (e.g. 3D FinFET).

In the world of computer chips smaller transistors means less power, so it's a big deal that Intel is about to unleash 22 nm, even as its competitors reportedly are struggling with the transition to 28-32 nm.  When you add in the proprietary power-sipping "fin" gate design, in theory Intel could compensate for its architectural disadvantages and beat ARM at its own game.

II. For ARM it's All About Sticking to its Guns -- Architecture

For ARM chipmakers like NVIDIA Corp. (NVDA) and Qualcomm, Inc. (QCOM), there isn't the luxury of coasting on a process lead.  For them it's all about the architecture.  While much of the architecture's baseline performance is dependent on the designs produced by ARM, a fair amount of the performance also falls on the partner's ability to tweak and maximize its finished die.

The ARM alliance isn't exactly resting on their laurels.  According to a report by Bloomberg, Qualcomm earlier on Jan. 10 will show off its counterstrike against Intel -- high-end ARM CPUs aimed at laptops.  

Qualcomm toyed with this idea in the netbook realm as early as two years ago.  But it lacked the powerful CPUs that people expected of full-sized laptops.  Now that it's got those CPUs, it's waiting on the second key part of that equation -- support from most peoples' OS of choice -- Windows.  

Windows with ARM
Windows 8 is finally supporting ARM chips, as shown here in this CES 2011 shot -- Qualcomm is expected to unveil some powerful notebook at CES 2012. [(c) DailyTech/Jason Mick]

That support has arrived with Windows 8, currently in its beta phase.  Finally breaking its "Wintel" tradition -- of targeting Windows primarily at Intel chips, Microsoft is throwing its weight behind ARM as well.

Windows 8 offers full support for ARM CPUs, although legacy software products will have to be recompiled to work on the new architecture.  Given Windows' shift to app market software distribution model, though, legacy software may be less of an issue than some have previously speculated.

Even analysts who are mildly optimistic about Intel's mobile prospects are concerned about this fresh blood in the laptop space.  Daniel Amir, an analyst at San Francisco, Calif.-based Lazard Capital Markets estimates that Intel will reach around 13 percent share in the smartphone market by 2015, of an estimated total market of 1.1 billion devices that year (2011's market was 300 million devices).  

But he also estimates than ARM's purpose-driven designs will steal a third of the mobile computer (laptop/netbook) market by 2015, up from 8 percent in 2011.  That market is expected to also see growth, expanding from 275 million units in 2011 to 340 million units in 2015.

The key problem for Intel, though, says Mr. Amir is that laptop chips fetch a higher price -- around $107 USD on average -- where as smartphone chips are typically priced around $20 USD.  In that sense Intel has more to lose in the laptop market where it currently has an 80 percent market share.  Mr. Amir states, "[Intel] needs to be sure that [they're] not losing the notebook."

He predicts $2.2B USD in lost sales for Intel by 2015.

Qualcomm and others still have a lot to prove, however.  Much like Intel must prove that it can scale its powerful chips to smaller, lower-power smartphone designs, Qualcomm must prove that it can scale its smaller, lower-power chips to higher power laptop designs.  

The first major test for Qualcomm will come with its Windows 8 laptop and tablet developer platforms, which it plans to release shortly.  In a bit of a sign of favoritism to its traditional x86 partner, Windows 8's reference tablet and notebook designs are still sporting Intel chips.  But Qualcomm isn't concerned about this, saying that it's ready to put its own reference designs in competitors’ hands.

III. Servers -- the Third Tier

With ARM invading the laptop space and x86 invading the smartphone/tablet space, it's easy to forget that there's four pillars of the modern computing chip market:

i. Embedded
ii. Mobile Devices
iii. Servers
iv. Personal Computers

While Intel hasn't shown much interest in squeezing x86 into the low-volume embedded market, the third tier -- servers -- is a crucial bread and butter sector of Intel's business.  Intel can sell server chips to businesses at much higher prices than its consumer chips command.

But ARM is preparing to invade this market as well.  Hewlett Packard Comp. (HPQ) is working on the world's first mass market server chip designs.  We may not see these designs at CES, or at least not on prominent display.  But keep this final frontier of the conflict in the back of your mind.

Calxeda dense server
[Image Source: Calxeda/ARM Holdings/HP]

If ARM penetrates the server market -- another power consumption sensitive application -- Intel may be impacted far more severely than if ARM penetrates the slightly lower margin laptop/netbook market.

Source: Bloomberg



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Nothing to prevent Intel from making ARM
By vol7ron on 1/3/2012 12:40:45 PM , Rating: 2
Well, they would have to license it and being Intel, I bet they would be charged more than competitors.

Also, Intel already has went RISC before with controller chips (i960) and the StrongARM CPU and I think the Itanium is RISC-based.

That said, if Intel wanted to go RISC, I'm not so sure they'd go ARM, they might fall back on their own patents, or re-develop.


RE: Nothing to prevent Intel from making ARM
By Shig on 1/3/2012 1:05:57 PM , Rating: 2
From what I've read about the industry roadmaps in fabrication, Intel moving to 22nm 3-D transistors was multiple years ahead of anyone else. Similar roadmaps can keep up with the lithography dimensions, but they're 2-3 years behind on FinFet, and that's optimistic. FinFet by itself is nearly worth an entire node shrink in performance, even more so when you're dealing with lower voltage performance.

If Intel pulls off good yields with Ivy Bridge, things will not look good for other fabs and the companies that rely on them.


RE: Nothing to prevent Intel from making ARM
By french toast on 1/3/2012 1:59:46 PM , Rating: 2
Fair point, however early ivy bridge indications point to the lower end chips having very simular tdps as the sandy bridge parts with only minimal performance improvement.

Of course this new node tech will help it alot, but x86 akready starts at a massive disadvantage due to legacy bloat, so it may take 22nm finFET just to stay competitve in 2011..let alone 2013 when it MAY release..
By that time arm will be moving to 22nm themselfs..
I just see it working out some how


By vol7ron on 1/4/2012 3:13:49 PM , Rating: 2
I'm confident that Intel will make the right decision when competition closes in. I admit, I'm not an expert in this, but from my understanding a lot of x86 bloat is due to deals Intel has made with companies that still run that old software.

My question is, can you take some of those instructions away and still have today's programs (Office/games) that don't rely on those instructions, to operate unaffected. If so, they might be x86-based, so that most of the x86 compiled programs today are still functional, but still be reduced to improve efficiency.

Either way, Intel has shown a history of relying on its production means until competition ramped up and caused them to switch. Take, for instance, the Pentium CPU. They kept increasing Hz, not efficiency. Competition came in, and no longer was it about clock speed, but the biggest improvement was doing what AMD had done (Intel's card up its sleeve): moving the memory controller off the board and on die.


RE: Nothing to prevent Intel from making ARM
By geddarkstorm on 1/3/2012 2:45:24 PM , Rating: 1
The fact ARM is resoundly beating Intel in the mobile space, and now posed to invade into laptops (and servers), all while having a massive disadvantage process and manufacturing wise, says a lot about the ARM architecture (or conversely says a lot negatively about x86).

Just imagine paring ARM's architecture prowess with Intel's manufacturing, and we could have CPUs that are jawdropping.


By encia on 1/4/2012 8:48:28 AM , Rating: 2
CPU alone doesn't complete a laptop i.e. the GPU factor.


RE: Nothing to prevent Intel from making ARM
By bruce24 on 1/3/2012 2:23:14 PM , Rating: 2
quote:
Well, they would have to license it


They are ready have a license. While they sold most of their Xscale line to Marvell back in 2006 they kept Network and Media processors based on Xscale, which they have since moved to x86, but still have the license to design/manufacture ARM processors.


By Shig on 1/3/2012 5:21:36 PM , Rating: 2
This is a little off-topic, but I want to commend ARM on their recent initiative of 'Big-Little' processing. They are making great strides in power gating technology and pairing 'big-iron' processing with super efficient processing.

In Intel terms this would be like a Sandybridge processor coming together with an Atom processor with graphics on the same die. Whenever you needed a lot of power, sandybridge turns on, whenever you don't, atom turns on, and the other turns off, and vise versa.

This is apparent in the Nvidia Tegra 3 quad core. This processor is actually a 5 core CPU that has 4 'big' cores and one 'little' core, with nearly full power gating. The goal is to use exactly as much cpu power that you need at any one point totally maximizing performance and battery life.


RE: Nothing to prevent Intel from making ARM
By Jeffk464 on 1/3/2012 6:13:27 PM , Rating: 2
With intel's talent and R&D budget it would be interesting to see what they could come with if they dropped all legacy and came up with something entirely new.


By Calin on 1/4/2012 4:35:02 AM , Rating: 2
Itanium?


"Intel is investing heavily (think gazillions of dollars and bazillions of engineering man hours) in resources to create an Intel host controllers spec in order to speed time to market of the USB 3.0 technology." -- Intel blogger Nick Knupffer














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki