backtop


Print 25 comment(s) - last by michael2k.. on Oct 21 at 4:58 PM

Samsung and TSMC are hindering Intel's mobile dreams

Analysts polled by Bloomberg and Thomson Reuters predicted Silicon Valley giant Intel Corp. (INTC) would take a beating fiscally in Q3 2013 as a sagging PC market was expected to badly dent its traditionally strong bottom line.  The Reuters crowd estimated earnings of 53 cents per share, which works out to about $2.69B USD in total profit, while the Bloomberg survey saw a prediction of 54 cents per share ($2.75B USD, total) in profit.

I. Brian Krzanich : "[Intel Has] an Increasing Focus on Mobile"

Instead Intel surprised, delivering earnings of 58 cents per share -- $2.95B USD total -- which allowed profit to stay relatively flat on a year-to-year basis, down just slightly (3 percent) from $2.97B USD in Q3 2012.  Operating income and net income  -- $3.5B USD and $3.0B USD -- were also higher than expected, despite Intel spending $4.7B USD on research and development.

Intel's revenue tied analysts expectations, drawing in $13.5B USD.  And it was able to reward shareholders with $1.1B USD in dividends and a share repurchasing effort that exchanged $536M USD for 24 million shares (a price of $22.33 USD per share).

Intel CEO
Intel CEO Brian Krzanich [Image Source: Intel]

Intel's data center group posted double digit growth (12.2 percent) as Intel's server chips continue to be top dogs in this market, which has experienced explosive growth with rising cloud storage and software use.

Facebook Data Center
A Facebook data center.  Data center growth was a major factor in Intel's ability to break even in Q3. [Image Source: AFP]

CEO Brian Krzanich, who took the reins from veteran Intel CEO Paul Otellini late this spring, expressed optimism about these results, remarking, "The third quarter came in as expected, with modest growth in a tough environment.  We're executing on our strategy to offer an increasingly broad and diverse product portfolio that spans key growth segments, operating systems and form factors. Since August we have introduced more than 40 new products for market segments from the Internet-of-Things to datacenters, with an increasing focus on ultra-mobile devices and 2 in 1 systems."

II. 22 nm Intel Smartphone chips May Face 16 nm ARM Chips Next Year

That's not to say that the news is all good.  

Overall the semiconductor market is booming, as evidenced by top chipmakers Samsung Electronics Comp., Ltd.'s (KSC:005930) record profit.  But increasingly much of the growth in demand is in the mobile sector, where Intel has struggled ever since the advent of the smartphone in the middle of the last decade.

A perfect case study is Intel's long awaited 22 nm tablet chip, that was supposed to deliver big power savings to the tablet market, via its 3D FinFET design, which saw its PC debut in Ivy Bridge chips.  

But tests of the quad-core 1.5 GHz (2.4 GHz turbo) Atom Z3770 Bay Trail chip -- while impressive versus old Qualcomm Inc. (QCOM) and NVIDIA Corp. (NVDA) CPUs found in various Android smartphones and tablets -- fell short against the Apple, Inc. (AAPL) A7 chip which is built on Samsung's 28 nm gate-first high-κ metal gate (HMKG) process.  The A7 uses ARM Holdings Plc.'s (LON:ARM) proprietary licensed ARMv8 instruction set and is found in the new iPhone 5S.
 
Bay Trail -- chipset

What is impressive about the A7's win in AnandTech's recent thorough testing is that it's actually a lower clocked chip (1.2 GHz) and has fewer cores (dual-core), and is built on a larger 28 nm node, yet somehow -- incredibly -- rivals-cum-partners Apple and Samsung (along with ARM Holdings, of course) have managed to squeeze out a chip that packs more processing power than Bay Trail.


There is also word that Intel's 14 nm mobile smartphone stock won't launch until early 2015, with a H2 2014 launch for 14 nm tablet chips.  Given that TSMC hopes to ship 16 nm chips in devices in H1 2014, it's quite possible that Intel's new 22 nm smartphone chips will have to compete with 16 nm ARM chips after seeing their tablet brethren lose to an ARM chip that was manufactured on what appears to be a far cruder (28 nm) process.

Historically TSMC has had production problems with die shrinks, but it managed to go from 65 nm (starting in 2007) to 28 nm (starting in 2010), so it's clearly advanced down the die-size ladder at a pace competitive to Intel's.

III. Broadwell Delayed, Intel Finds Itself in a Lonely Place

Intel announced during the earnings call a slight delay to Broadwell (which was supposed to launch in Q4, but will launch in Q1 2014) due to higher than expected defects.  Taken alone, this might be easy to dismiss.  But look at Intel's market situation as a while, its a bit troubling.

Broadwell
Intel's 14 nm Broadwell will launch in Q1 2014, followed by tablet counterparts, and (in early 2015) smartphone chips. [Image Source: PC Gamer]

Intel desperately needs an ally in its fight against ARM, but it's find one increasingly hard to come by -- partially because some key supporters of x86 or Intel's hardware are losing faith in both of these quantities amid poor mobile results.

For now Intel is relatively safe in that it has an entrenched market share in the PC space and an even more deeply entrenched lead in the server space.  But if ARM chips can beat Intel chips in power it may convince OEMs to start moving away from x86 -- particularly when Advanced Micro Devices, Inc. (AMD) switches its line slowly to ARM cores, which is expected to happen over the next few years.  That move could leave Intel in a lonely position as the only defender of x86.

ARM engineers
Everyone seems to be in ARM's camp except for Intel and its OEMs. [Image Source: My Statesman]

Even Microsoft Corp. (MSFT) and Intel -- long the sellers of the glorious "Wintel" solution -- have seemed at odds of late, with both companies seeming to view the other as the crucial obstacle to their mobile aspirations.  Intel insists its mission is to be the best for "every operating system", including pointedly asking Microsoft what it was doing to improve Windows 8's poor sales at its recent trade show.  Microsoft meanwhile is swallowing the financial costs of distancing itself from Intel by maturing the Windows RT platform (which runs on ARM processors), which has thus far been slow to take hold.

Of course Intel could always make an ARM core -- and may even has the licenses to do so via its acquisitions (ARM Holdings does not disclose all of its licensees).  But Intel's market dominance is built on x86, by locking demand via software compatibility.  If it loses the fight to keep PC software primarily x86-exclusive -- which could soon be a one company fight against an army of ARM chip producing companies -- it will likely lose the war.

Sources: Intel, Bloomberg [estimate, commentary], Reuters [estimate]



Comments     Threshold


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

How is Baytrail falling short of the A7?
By Mint on 10/17/2013 12:17:10 PM , Rating: 4
Your own Anandtech link shows the A7 falling behind Baytrail in 3 out of 4 benchmarks (note that for Sunspider, Anand says, "Intel responded with a Bay Trail run under IE11, which comes in at 329.6 ms"). Those are all primarily single thread tests, so if Anand included a couple multithreaded tests you'd see Baytrail curbstomp the A7.




RE: How is Baytrail falling short of the A7?
By michael2k on 10/17/2013 12:35:48 PM , Rating: 2
There's an entire section dedicated to the iPhone v Bay Trail.
quote:
I couldn't help but run Intel's current favorite mobile benchmark on the iPhone 5s. WebXPRT by Principled Technologies is a collection of browser based benchmarks that use HTML5 and js to simulate a number of workloads (photo editing, face detection, stocks dashboard and offline notes).


The iPhone wins half the tests and in aggregate is one point higher than Bay Trail!

quote:
It's not a clean sweep for the iPhone 5s, but keep in mind that we are comparing to the best AMD and Intel have to offer in this space. I suspect part of why this is close is because both of those companies have been holding back a bit (there's no rush to build the fastest low margin parts), but it doesn't change reality.


RE: How is Baytrail falling short of the A7?
By ritualm on 10/17/2013 2:58:42 PM , Rating: 3
Those goalposts are not static. Neither the Atom (Bay Trail) nor the A7 is fast enough. That the A7 beats BT in half the tests means nothing - I wouldn't use either to run a laptop, much less a desktop.


RE: How is Baytrail falling short of the A7?
By michael2k on 10/17/2013 3:45:37 PM , Rating: 1
It's a good thing you aren't running Intel or Apple.

The point isn't whether the A7 or BT is fast enough to run a laptop or desktop, it's whether they are fast enough to run a smartphone or tablet.


RE: How is Baytrail falling short of the A7?
By ritualm on 10/17/2013 8:20:38 PM , Rating: 3
You're missing the point. Both A7 and BT are too slow as they currently stand. The only reason the A7 seems like the fastest of all is either

1) you are an Apple fanboy

or

2) it's designed to run on what is basically a content consumption device.

Apple can design an ARMv9-based SoC that thrashes anything short of Sandy Bridge-E in benchmarks, and it's completely meaningless for everyone else. All it does is give you the ability to wave your imaginary epeen at those who don't subscribe to your Cupertino bias.


By A11 on 10/18/2013 7:46:22 AM , Rating: 1
You make absolutely no sense.


By michael2k on 10/21/2013 1:46:35 PM , Rating: 2
What's wrong with a consumption device?

Do you eschew a smartphone because it's not powerful enough? So you stick to even less powerful camera phones, or even older landlines?


RE: How is Baytrail falling short of the A7?
By Reclaimer77 on 10/17/2013 3:47:30 PM , Rating: 2
Those benchmarks are highly misleading because they are just browser benchmarks. They aren't telling you which SoC is faster, they are just telling you that differences in the platforms have Apple rendering browser pages faster.

Notice the actual synthetic benchmarks, Google Octane and Kraken, that really challenges the hardware beyond simple webpage rendering, Bay Trail wins easily.


RE: How is Baytrail falling short of the A7?
By michael2k on 10/17/13, Rating: 0
RE: How is Baytrail falling short of the A7?
By Reclaimer77 on 10/17/2013 5:10:12 PM , Rating: 2
All Apple did was take an existing SoC from someone else, and design a few customizations and tweaks.

Big freaking deal. That hardly puts them on par with Intel!


RE: How is Baytrail falling short of the A7?
By michael2k on 10/17/2013 5:36:53 PM , Rating: 2
Um, did you read about the A7?

They took an existing instruction set (ARMv8) implemented ahead of ARM. If you read about the A6, they did something similar with the ARVv7 architecture.

The, as you put it, existing SoC was only completed this year, and Apple beat everyone else to market.

AMD, Samsung, and others aren't going to ship anything until 2014. Source is Anandtech:

http://www.anandtech.com/show/6420/arms-cortex-a57...


RE: How is Baytrail falling short of the A7?
By Reclaimer77 on 10/17/2013 10:50:26 PM , Rating: 1
Okay fanboi, whatever you say.


By Mint on 10/18/2013 4:43:24 PM , Rating: 2
He's actually right about that.

But you're right about the A7 benchmark wins all being browser based, and Apple is running a different browser. Intel's best result will probably come with IE11.

On top of that, Intel fit a quad-core into the that power envelope. All these benchmarks are single-threaded.


By michael2k on 10/21/2013 4:58:54 PM , Rating: 2
Is that all you can add to the discussion?

You don't even care that Apple, like Qualcomm, are designing their own CPUs, even if they are using an existing instruction set?

This is akin to how AMD designs their own CPUs but stick, mostly, to the x86 instruction set (the big difference being when they implement their own extensions).

You think I'm being a fanboi for bringing this up, but think you're acting acceptably in saying this is irrelevant?

According to Strategy Analytics, Apple is the #2 multicore SoC vendor the first half of 2013, and it wasn't util Q4Y12 that Qualcomm surpassed Apple:
http://blogs.strategyanalytics.com/HCT/post/2013/0...


RE: How is Baytrail falling short of the A7?
By A11 on 10/18/2013 8:05:38 AM , Rating: 2

It's not just Apple that Intel is up against but Apple and ARM, another pure CPU company.

Anyway I'd think it more relevant to compare Intel to Qualcomm's/NVidia's offerings, Apple is more or less playing its own game but those three are in direct competition for design wins.


RE: How is Baytrail falling short of the A7?
By ritualm on 10/18/2013 7:48:52 PM , Rating: 2
Cute you think that, because ARM never actually makes its SoCs, never mind CPUs. It does the architecture work and then sells licenses to anyone who wants to use it / entire chips for drop-in solutions. The actual chip fabrication needs to be done elsewhere e.g. GloFo, Samsung, TSMC, etc, so any ARM licensee still needs to deal with a third party for the actual chips.

Apple doesn't fab its own ARM-based chips, either - that stuff currently comes from Samsung. All it has is its chip design, nothing more.

Both GloFo and TSMC are just chip fabs. They make chips for everyone else, that's it.

Meanwhile, Samsung is in an unique position because it owns both the ARM licenses, and additionally the massive multi-billion chip fabs needed to actually make those SoCs (amongst other chips that make a phone/tablet/computer!).


RE: How is Baytrail falling short of the A7?
By A11 on 10/19/2013 5:50:16 AM , Rating: 2
I'm not quite sure if you're trying to make a case for ARM not being a CPU company here?

But besides that, thanks for telling me what I already knew about architectures and foundries.


RE: How is Baytrail falling short of the A7?
By ritualm on 10/19/2013 2:45:14 PM , Rating: 2
You're implying that ARM is a pure CPU maker. Maybe, but ARM doesn't own/run any chip fabs, so the only thing ARM has to worry about is process defects hurting their chip designs.

Of course, none of this is relevant to your "ARM is the good guy, Intel is evil!" irrational discourse.


By A11 on 10/21/2013 9:44:58 AM , Rating: 2
Designing chips and producing chips are two different things and ARM not being a foundry is no argument for it not being a pure CPU company. All they do is designing CPU's.

As for your ARM is the good guy straw man then I'd like to know where I've expressed any kind of corporation favoritism.
I don't worship companies like certain other people do but I do think intel got caught with its pants down when the mobile train left. However, I'm also aware that Intel is a giant in chip design and they are throwing everything they got at catching up so I'm far from counting them out.


RE: How is Baytrail falling short of the A7?
By mjv.theory on 10/17/2013 2:53:00 PM , Rating: 2
quote:

Apple A7 1.3GHz 878.9 ms 1831.4 ms 436.1 ms 604.6 ms
Atom Z3770 1.46GHz 693.5 ms 1557.0 ms 542.9 ms 737.3 ms


By Mint on 10/18/2013 4:49:00 PM , Rating: 2
Those are all sub-results from the WebXPRT benchmark, which is yet another browser test.


By mjv.theory on 10/17/2013 2:53:01 PM , Rating: 2
quote:

Apple A7 1.3GHz 878.9 ms 1831.4 ms 436.1 ms 604.6 ms
Atom Z3770 1.46GHz 693.5 ms 1557.0 ms 542.9 ms 737.3 ms


By mjv.theory on 10/17/2013 2:53:01 PM , Rating: 2
quote:

Apple A7 1.3GHz 878.9 ms 1831.4 ms 436.1 ms 604.6 ms
Atom Z3770 1.46GHz 693.5 ms 1557.0 ms 542.9 ms 737.3 ms


Bay Trail vs, A7
By avishaybennatan on 10/17/2013 1:55:12 PM , Rating: 2
As these benchmarks test JS performance they mean almost nothing if not running the same JS engine




I'll believe it when I see it...
By Nagorak on 10/17/2013 4:40:41 PM , Rating: 2
Intel's experience with fabs is unparalleled. If Intel is having some difficulties moving to 14nm, I have serious doubts that TSMC is going to seamlessly transition to 16nm. They're going to have similar problems, if not more.

I also think the thing to keep in mind is how much improvement Intel made from Clover Trail to Bay Trail. With Bay Trail they've essentially achieved parity with the ARM SoCs. Intel is basically coming at the problem from the other direction than ARM (scaling high performance down), and now they've met in the middle. Now that they're equal, the next step is really the one that will be interesting to see.

I think if Intel can roll out 14 nm next year, they have a shot at really putting it to the competition. With Bay Trail it looks to me like graphics is the weakest link, with general processing power being top notch. It remains to be seen how power use stacks up though.




"I modded down, down, down, and the flames went higher." -- Sven Olsen














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