backtop


Print 12 comment(s) - last by GatoRat.. on Jan 16 at 7:55 PM

The war between x86 and ARM is almost upon us

One of the most compelling storylines to emerge from the 2012 Consumer Electronics Show was the struggle between ARM and x86, two architectures vying for CPU global domination.  ARM’s proponents argue that the reduced instruction set computer (RISC) architecture used in ARM is inherently more power efficient than x86.  x86 proponents disagree, often arguing that ARM cores are too weak for serious computing.  x86's proponents also brag about their lead in semiconductor manufacturing processes, which they feel will yield more efficient features, putting x86 ahead of ARM in power efficiency.

I. Where Are we at Now?

In order to understand where the consumer CPU business is headed, it's important to first understand where it's at now.

i. ARM

"ARM CPUs" is a rather ambiguous term in that it encompasses two unique kinds of chips.

The first kind of ARM CPU is an ARM core that licensed IP-core designs for the CPU from the UK's ARM Holdings plc (LON:ARM).  The chipmaker then combines that IP-core with graphics core(s), audio processing circuitry, I/O circuitry, and more to form a system-on-a-chip (SoC).  Samsung Electronics Comp., Ltd.'s (KS:005930) Exynos chips, NVIDIA Corp.'s (NVDA) Tegra chips, and Texas Instruments Inc.' (TXN) OMAP CPUs all fall under this category as they use full ARM Holdings IP-cores.  In the case of Samsung's Galaxy S II, a graphics IP-core is even licensed from ARM.

The second type of ARM chipmaker is the handful of companies who have opted to rather than licensing IP-cores to license the ARM Holdings' instruction set, but instead design their own CPUs from the ground up.  Qualcomm Inc. (QCOM) is one such player.  It primarily targets the high-end American market.  It is the exclusive CPU SoC maker for Windows Phone.  Another example of an ARM instructions set licensee is Marvell Inc., a a company who primarily has targeted budget/emerging markets like Indonesia and China with its fully custom designs.

Apple, Inc. (AAPL) is also rumored to have recently obtained an ARM Holdings instruction set license, pointing to that they may be leveraging their recent chipmaking acquisitions to build a next generation fully custom in-house Ax SoC, versus relying on rebranded designs from Samsung and others.

So when you hear "ARM CPU" or see Qualcomm chips compared to NVIDIA chips be aware that you're dealing with pretty different implementation scheme, which deal with the same base level instruction sets

Also, to be clear none of this information on who licenses IP-cores versus who licenses the instruction set is official -- ARM Holdings does not publicly reveal its licensees.  However, the above information comes from industry sources close to ARM Holdings.

ARM currenty dominates the majority of embedded markets (think dishwashers, microwaves, etc.).  It also owns a previously unchallenged monopoly on smartphone chip architecture, and a near-complete monopoly on tablet SoCs.

ii. x86

The x86 CPU dates all the way back to Intel Corp.'s (INTC) Intel 8086 in 1978.  Since then almost all Intel consumer CPUs have been x86.  Advanced Micro Devices, Inc. (AMD) also adopted x86 for their purposes.  However, Intel is by far the bigger player, still, owning somewhere around 80 percent of the x86 CPU market, according to recent estimates.

The x86 sales model is simpler -- there's no tangled web of licensing.  You simply produce and sell what you design.

Intel is the world's most used personal computer chipmaker.  That means x86 chips can be found notebooks, laptops, netbooks, and even Intel's favorite ultraportables -- "Ultrabooks".

Ironically, Intel currently does reportedly own an ARM Holdings IP-core licensing, as it uses a small ARM processor in its subsidiary Infineon's baseband chipsets.

II. Intel Moves to Stay Relevant

Given the long delays, the arrival of completed Medfield silicon at CES 2012 was a big deal.  Medfield -- Intel's smartphone chip -- had been in the works since 2009 at least.

Lenovo Group, Ltd. (HKG:0992) announced the world's first commercial Medfield phone -- the K800 -- at CES 2012.  The phone will launch exclusively in China in Q2 2012, on China United Telecommunications Corp., Ltd. (SHA:600050).  The phone features:
  • 1.6GHz Intel Atom (Medfield) Z2460 chip (single-core)
  • 4.5-inch TFT scren (720p capable)
  • Intel GMA Graphics
  • 8 MP rear-facing camera
  • front-facing camera (resolution unknown)
  • WCDMA 21 Gbits/s HSPA+ 3G modem
  • Bluetooth 3.0
  • 802.11 b/g/n
  • GPS
  • Android 2.3 "Gingerbread", upgrade to Android 4.0 "Ice Cream Sandwich" is coming shortly.

Intel also aired a similar handset as a reference design for OEMs who might be interested in toying with the idea of an x86 smartphone.

These devices are buoyed by razor-thin, attractive packaging -- a pleasant surprise. However, it's important to exercise caution as heat and battery life for the new platform still need to be examined in full depth.  Various publications were unable to test these traits as Lenovo K800 lacked access to the Android Market, which is banned in China.  Without taxing apps it was hard to get a good feel for exactly how good or bad battery life/heat are.

Aside from the two concrete devices (the K800 and the reference handset) there was only one other decidedly more vague promise of production of Medfield smartphones.  That promise came from Google Inc. (GOOG) subsidiary Motorola how promised Intel, "a multi-year, multi-device strategic relationship."

Of course this could mean two devices in 2013, or 20 devices in 2012 -- the statement is very ambiguous.  What is clear is that Motorola is interested in the long term prospects of Medfield.

III. 2013 -- the Real Battle

This year may see a small fraction of x86 uptake in tablets and PCs.  But it seems apparent that Intel merely wants to put its foot in the door, in preparation for 2013.

In 2013 Intel will shrink Medfield's successor to the 22 nm node, bringing in the 3D FinFET transistor design, which dramatically reduces leakage in order to boost power efficiency (on the mobile end) or processing power (on the desktop end).

Of course Intel is currently only on the 32 nm node for its mobile chips, while some ARM players are already moving to 28 nm.  What way the 2013 battle goes remains to be seen, but it is clear that the specialized 3D FinFET feature will drive some mobile sales.

Further, it will be a big year for ARM, as well.  ARM chipmakers are currently playing with Windows 8 latpop prototypes.  While some may go on sale over the holidays, it's clear that ARM's notebook PC position will be more mature in 2013 than in 2012.  The same goes for ARM's efforts to deploy servers.

IV. Bad Blood Ignites

We've long heard Intel boast about how it's going to wipe the floors (at some point) with ARM in terms of power efficiency and processing power.  Intel claims to have "at least a two year" production process lead over its ARM rivals.

While this may be a bit exagerrated, it does look Intel has arguably a one year process lead or so -- only slightly less that Intel's bold claims.

By constrast, when we met with ARM, we found ARM's managemento be careful not to attack x86 too hard.  Instead they dropped humorous observations -- like that "Moorsestown" was originally aimed at smartphones, but only later reduced to a smartphone push only, for the brand.

The ARM folks also pointed to Medifield's long delays, as a humorous observation on the platform's struggles.

At his 2012 CES press conference, CEO Warren East stepped up the rhetoric, commenting, "It's inevitable Intel will get a few smartphone design wins -- we regard Intel as a serious competitor.  Are they ever going to be the leaders in power efficiency? No, of course not. But they have a lot more to offer."

He adds, "They (Intel) have taken some designs that were never meant for mobile phones and they've literally wrenched those designs and put them into a power-performance space which is roughly good enough for mobile phones."

Harsh criticism indeed.  

But as Intel invades the smartphone space, and ARM invades the notebook/server markets, it should be interesting times for tech enthusiasts.  But just remember, the true tests -- in terms of the quality -- will likely mostly land in 2013, when Intel and the ARM chipmakers are compelled to modify their own chips.

Sources: Reuters, Intel



Comments     Threshold


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

By mackx on 1/14/2012 5:33:17 PM , Rating: 2
isn't quality a factor in this as well? regarding nodes i mean? i thought intels 32nm was better than TSMC/GFs 28nm as an example?

intel can use finfets but i read that the competition won't have that tech for their 22nm nodes? if that's true, then TSMC/GFs new nodes won't be as efficient as intels right? then 14nm and it's even more expensive and even trickier.


By lol123 on 1/15/2012 5:40:00 PM , Rating: 2
"isn't quality a factor in this as well? regarding nodes i mean? i thought intels 32nm was better than TSMC/GFs 28nm as an example?" That's absolutely accurate. Have a look at Figure 11 in this article for example: http://preview.tinyurl.com/72sxf79

The SRAM density of Intel's 32nm process is lower than that of 28nm TSMC, but logic density is actually higher.


"DailyTech is the best kept secret on the Internet." -- Larry Barber














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