BannerforSemiWiki 800x100 (2)
Semiconductor Expert Forum

Posted on

Apple’s iPAD 3 Kicks Off a Season For Silicon Valley Re-Alignments

Apple’s iPAD 3 Kicks Off a Season For Silicon Valley Re-Alignments
by Ed McKernan on 02-13-2012 at 3:43 pm
Categories: Foundries, Intel Foundry, IP

The Corporate PC may finally be on life supports and with it the purveyors of the hardware and software solutions as consumer adoption of new technology outpaces the carefully planned IT budgets that guarantee positive 3 year Returns on Investments (ROI). The Apple of the consumer’s eye will soon feast on the iPAD 3 with its retina display and 4G LTE that untethers the productivity class for good from centralized IT departments. It’s “Ash Heap of History Time.”

If it were just the iPAD3 that was rolling out, you could push back and say, Please! However, quick to follow will be redesigned MAC Air’s and MAC Book Pro’s that will feature Intel Ivy Bridge processors and 4G LTE as well. Then the IT department has to admit that the “Cloud” is everywhere and that it no longer makes sense to deal with all the complexity of software versions and updates between the various versions of Dell and HP PCs.

The pain will soon start to get extreme for Microsoft and Google as the train that left the station years ago in terms of the mobile hardware that rests in the users hands on a 24/7 basis will be available from just one vendor: Apple. A sign that chaos reigns is when those on the defense start coming up with half answers. For example, Microsoft is rumored to be working on a version of Office for iOS. That’s nice and what is the price? Surely much less than a version delivered on a Microsoft Cloud. By relying on Nokia and Apple to deliver the hardware, Microsoft falls into a two degrees of separation gap from the end customer. Meanwhile Intel ships Xeon’s at ever increasing prices to Microsoft’s data center and to those of Google and Facebook. The Intel bacteria have glommed onto the Data Center Host.

Is the Motorola alliance with Intel for Smartphones an all encompassing one that precludes the use of nVidia or Qualcomm. What about tablet and Ultrabooks? When will Google/Motorola wake up and see that their main revenue stream of search can be redirected by Siri to the iCloud? Time to consider Andy Grove’s dictum: “Only the Paranoid Survive.”

It is truly historical to observe the alliances that are breaking under the duress of Apple’s assault and the resulting ones to form as a means to stop the blitzkrieg. Intel’s client price and delivery model now revolves around the Apple MAC Air. The pricing every other OEM gets has to be higher than Apple’s. And so, every other OEM and partner is a 2[SUP]nd[/SUP] class citizen sitting at the back of the bus. Apple data centers will receive favorable Xeon treatment. However, what if Google went to Intel and promised to up the server buy while also planning to build 10M tablets a quarter based on Atom. Intel would need to consider a front of the bus seat unless Apple A6 is already taped-out in 22nm.

An Intel no bid to Google then leads to the next scenario – a Google/AMD joint venture on x86 Server Processors and possibly a low cost x86 mobile processor for ultrabooks, so that the search bar remains “Google”. The same scenario could play out for Microsoft and AMD. This is what will make the semiconductor industry extremely interesting over the next several years.

On the semiconductor side, Qualcomm has to consider scaling dramatically in 2012 as they leverage their 4G advantage. Paul Otellini, Intel’s CEO claimed in the last earnings conference call that the communications silicon is being pulled into the x86 platform. In one way or another, Intel is forcing Qualcomm to retreat into an IP mode or to man up in a Jerry Sanders fashion. Furthermore, Qualcomm has to realize that if Intel seeks to own Ultrabook down through tablets to the smartphones then Qualcomm has to make a maximum thrust into the ultrabook space and it can’t be with just ARM. A Qualcomm – AMD “x86 Snapdragon” would allow them to increase the penetration of the communications content just as the x86 value piece is waning. In reality the x86 and ARM cores are both diminishing as a percentage of the overall mobile platform.

We live not only in interesting times but one that is about to have incredible amounts of upheaval. Not all of the Paranoid will Survive.

FULL DISCLOSURE: I am long AAPL, INTC, QCOM, ALTR

Posted on

Smartphone shipments in Q4’11, the Milky Way the new limit!

Smartphone shipments in Q4’11, the Milky Way the new limit!
by Eric Esteve on 02-13-2012 at 3:51 am
Categories: IP

In a previous blog, I was forecasting smartphone shipment to reach 430 million units in 2011. I was wrong! Apparently, I am not better than other analysts, as the actual figures for the smartphone shipment in 2011 are more than slightly different, as these have topped 491.4 million units! After 304 million in 2010, that’s a 61.3 percent year over year increase, according with IDC. IPNEST has revised the smartphone shipments forecast for 2012- 2016, taking into account these latest figures for 2011: we expect almost 1 billion smartphoneto be shipped in 2016.

If we come back to the actual shipments, in 2011 one out of three wirelesses handset shipped was a smartphone. This is to be compared with 2010, when it was one out of five. What is the impact on our day to day business, IP and Semiconductor?

Application processor market: this is the obvious first segment to look at, even if it’s not the only one. With an Average Selling Price (ASP) in the $20 range in 2011 for an application processor dedicated to smartphone, this segment has represented a $10 billion market in 2011. If we look at the Total Addressable Market (TAM), it is slightly smaller… this is an understatement, as the difference between the total market and the TAM is made of the A5 (A4) shipments. Apple shipping application processor to… Apple. As Apple has shipped about 60 million iPhone during 2011, the application processor TAM is equal to (490 -60)*20 or $8.6 billion, or almost 3% of the total SC market of $300 billion.

PMIC market: those who have read my previous post about the Power Management IC market have discovered like I did that this segment, even if it’s not as impressive as application processor, is generating pretty decent sales, as the PMIC content has been evaluated around $8 by smartphone (and $20 by Media Tablet!). The PMIC market in the smartphone segment only can be evaluated at $4 billion. If we add the Media Tablet segment, with 63 million units shipped in 2011 according with IDC, we add $1.26 billion to the PMIC segment. Thus, for the smartphone and Media Tablet segment, the PMIC TAM was above $5 billionin 2011. Pretty impressive for a SC segment almost neglected by the IDC, Gartner et al., as they focus on the more sexy application processor segment!

MIPI IP: IPNEST analysis is that MIPI, as any emerging technologies, is firstly used in the high end market, or smartphone segment. As with any new technology, integrating one of the MIPI specifications (CSI, DSI, DigRF, UniPro, LLI and more, see MIPI organization websitefor more information) represent an investment to make, in order to understand the specification, design the function or acquire the IP externally, integrate it into the IC (it can be the application processor or one of the many peripheral IC), and so on… But the many advantages generated by MIPI usage will certainly bring a high return on this investment, especially to the system integrator (manufacturing the smartphone) and ultimately to the end user, as MIPI allow for:

  • Lower power consumption compared with other interfaces
  • Standardization in the way to interface the various peripheral devices with the application processor, like camera controller or display controller IC for example, allowing to build the complete system in an easier way, thanks to the guaranteed compatibility between IC compliant with the same interface specification.

If the smartphone shipments are rocketing, that should be good for MIPI adoption, as more than expected MIPI powered IC will be manufactured. Then, when the production volumes are growing, it’s usual to see the IC price going down. When the price is going down, the over-cost associated with the integration of a new technology tend to disappear. It’s a kind of virtuous cycle: more smartphone shipped means more MIPI powered IC manufactured at a more affordable price for the customer, then more customers supporting the technology. Look at Qualcomm: a couple of years ago, the company was not supporting MIPI (but MDDI, a homemade concurrent specification). Now, Qualcomm has dropped MDDI and fully support MIPI!

For more exaustive information about MIPI, see MIPI Orgor “MIPI IP Survey“

By Eric Estevefrom IPnest

Posted on

AMD Financials are Puzzling!?!?!?!

AMD Financials are Puzzling!?!?!?!
by Daniel Nenni on 02-12-2012 at 4:00 pm
Categories: General

As I mentioned in my previous blog, AMD and GlobalFoundries,it is rather difficult to interpret AMD’s financial results. At the end of the quarterly earnings conference calls analysts get to ask questions. It is the comedy relief portion of the call usually, so I blog this today to better prepare these so called analysts for their job of helping investors understand what is really going on insideAMD.
Continue reading “AMD Financials are Puzzling!?!?!?!”

Posted on

DFM Industry Survey

DFM Industry Survey
by Beth Martin on 02-10-2012 at 1:28 pm
Categories: EDA, Siemens EDA

As part of the DFM Conference at the SPIE Advance Lithography symposium, the DFM committee is conducting an informal survey on the current state of Design For Manufacturability in the Semiconductor Industry.

Please take this anonymous 16 question survey to identify critical Design for Manufacturability (DFM) issues facing the semiconductor industry.

Posted on

Yes, there is such a thing as a free…model

Yes, there is such a thing as a free…model
by Paul McLellan on 02-09-2012 at 8:18 pm
Categories: EDA, Synopsys

I have been saying for years, ever since I started working at VaST, the biggest barrier to adoption of virtual platform technology for what I like to call virtualized software development is the availability of models. If models do not already exist when they are needed there are two issues: it takes money to develop them but, probably more importantly, it takes time. Since a large part of the attraction of virtual platforms is that they can potentially be available before reference boards, chips, real cell-phones etc, anything that soaks up that time makes the approach less attractive. A month developing models is a month slip in the software development schedule.

There have been a number of attempts to address this problem, the most serious of which is Synopsys’s TLMCentral. When this started last year, it was focused on being a central location to find models that were commercially available either through the site, or, more often, elsewhere. There are now over 800 such models available.

This year they are adding a push to encourage model developers to upload open-source models. Obviously nobody will do this for something proprietary like a processor, nor something expensive to develop that therefore can easily be sold. But there is really little downside to sharing a model for a UART or a timer, and there is little point in every company developing their own generic timer model.

To encourage people to get started, Synopsys have uploaded 16 or so models that can be freely downloaded. Some have already been downloaded 30 times so there is definitely demand there. These are open-source models with no licensing fees.

And as a more direct form of encouragement there are several competitions to win an iPad2 (and I’m sure they’ll switch it to iPad3 if Apple announce a new one as some people expect).

  • submit any model. One of the first 50 submitted will win (although confusingly the front page of TLMCentral says upload the 50th model and win; maybe you win another one for that too)
  • be the person who uploads the most models by the end of March
  • upload the model that gets the highest vote by end of April
  • upload the model that gets the most downloads by the end of May
  • or upload a model of a sensor interface such as accelerometer, gyroscope, magnetometer or proximity sensor (deadline February 17th)

Of course there are some conditions, the models need to be written in System-C TLM 2.0 being the main one.

More details about all of this, including the competitions, here.

Posted on

DFM at SPIE Advance Litho show

DFM at SPIE Advance Litho show
by Beth Martin on 02-09-2012 at 6:40 pm
Categories: EDA, Siemens EDA

This year’s SPIE Advanced Lithography is loaded with interesting keynotes and sessions. To help me narrow down what to see, I spoke with John Sturtevant. John is co-chair of the Design for Manufacturability through Design-Process Integration conference, and the director for technical marketing for RET products at Mentor Graphics.

12-15 February, 2012
San Jose Convention Center
All the cool kids will be there

I asked him how it’s changed since first introduced six years ago. John has been co-chair of the DFM conference for the last five years, so he has perspective. Here’s what he told me.
The conference saw a steady upsurge in number of papers in years 1-4, reflecting to a certain extent the hype that was DFM early on, as startups appeared on the scene, and “DFx” appeared on the business cards of an increasing number of engineers and VPs. As we separated the wheat from the chaff, most startups disappeared, and SPIE paper submissions to the conference dropped off somewhat to a steady state of around 40 papers. This was consistent with the predictions of Joe Sawicki in his invited paper in 2004, the second year of the conference.


The wheat which is left now grows in the fields of

    [*=1]Multi-patterning implications to design and manufacturing
    [*=1]Implications of EUV
    [*=1]Design variability effects in manufacturing
    [*=1]Litho-friendly design
    [*=1]Restricted design rules

John recommends these papers in the DFM conference at SPIE:

Wednesday 10:40a, “Layout optimization through robust pattern learning and prediction in SADP gridded designs.” UC Santa Barbara and Mentor researchers present their study of placement-level optimization, including how to build a predictive model for layout pattern classification, and applying the model to find and eliminate printing hotspots. [8327-04]

Wednesday 1:50p, “Fully integrated litho aware PnR design solution.” STMicroelectronics and Mentor engineers present the STMicroelectronics back-end CAD solution for litho hotspot search and repair that is based on pattern matching and local re-route abilities in place and route tools. [8327-09]

Wednesday 2:30p, “Smart double-cut via insertion flow with dynamic design-rules compliance for fast new technology adoption.” Mentor and GLOBALFOUNDRIES engineers introduce an automatic redundant-via insertion flow. [8327-11]

Thursday 1:40p,“Thickness-aware LFD for the hotspot detection induced by topology.” Samsung and Mentor engineers present a method for advanced process window simulations with awareness of chip. [8327-24]

Thursday 2:00p, “The complexity of fill at 28nm and beyond.” Mentor and AMD engineers discuss modern fill challenges and advances in technology for 28nm. [8327-25]

And, if you’re interested to see how triple patterning will work for 14nm designs, go see “14nm M1 triple patterning”at 5:10p on Wednesday. [8326-38]

Posted on

Why X-Fab uses 3D Resistance Extraction and Analysis

Why X-Fab uses 3D Resistance Extraction and Analysis
by Daniel Payne on 02-09-2012 at 11:18 am
Categories: Uncategorized
10 Comments

At DAC in 2011 I visited an EDA company called Silicon Frontline Technology because they offered some 3D field solver tools used to create the highest accuracy netlists that can then be simulated with a SPICE circuit simulator to predict timing, power and IR drop. A recent press release with X-FAB and Silicon Frontline looked interesting so I contacted Thomas Hartung, the VP Marketing and Joerg Doblaski, the team leader of the Design Technology Group at X-FAB to better understand their IC design process and why it required a 3D resistance extractor.
Continue reading “Why X-Fab uses 3D Resistance Extraction and Analysis”

Posted on

DVCon: Hardware/software Co-design from a Software Perspective

DVCon: Hardware/software Co-design from a Software Perspective
by Paul McLellan on 02-09-2012 at 4:56 am
Categories: EDA

The EDAC Emerging Companies Comittee (would that be the EDACECC?) is organizing a free panel session one evening at DVCon. It is Monday February 27th from 6pm to 8.30pm. I don’t yet have a room but it will be at the DoubleTree Hotel where DVCon is being held.

EDA companies often address hardware/software co-design from a hardware point of view as if the software somehow is going to be put together once the chip is available, and is a relatively small part of the design of the system (real men design chips). But, in fact, software is often much longer lasting than any individual chips. Much of the software in your phone may be a decade old and running on the fifth or sixth iteration of the hardware. Apple’s iOS alone ran first on a Samsung chip, and then Apple’s own A4 and A5 chips, for example. The result is that hardware and software teams look at the importance of software and software development methodology very differently.

Michel Genard of Wind River was going to moderate it, but Wind moved their sales kickoff and so I’ve been drafted in. I was VP marketing at both VaST Systems Technology and at Virtutech, both of which were heavily involved in the space (VaST is now part of Synopsys, Virtutech now part of Wind River, in turn owned by Intel).

The panellists are:

  • Atul Kwatra, principal engineer, Intel Corporation (Intel own Wind River and Virtutech)
  • Michael James, senior staff engineer, Lockheed-Martin Space Systems Company (Michael used to work with me at Virtutech)
  • Don Williams, head of core technology, Skype (and previously headed up a software development team at Cisco)
  • Bill Neifert, chief technology officer, Carbon Design Systems (a virtual platform and modeling company)

Grab a beer and come along. I hope to see you there. The panel is free. You don’t need to be registered for DVCon, nor an EDAC member.

To register for the EDAC panel, go here
To register for DVCon, go here
For more information about the panel, go here

Posted on

Powering the Platforms: ARM’s 2012 Approach

Powering the Platforms: ARM’s 2012 Approach
by Don Dingee on 02-08-2012 at 4:30 pm
Categories: Arm, IP

A client turned me on to a great new book, “The Age of the Platform” by Phil Simon. It’s about how Amazon, Apple, Facebook, and Google have radically transformed the landscape. For me, it’s not just social networking – it’s social computing, changing how things are designed.

I’m borrowing this right from Phil’s description of why he selected these four companies as the platforms at the center of this shift:

“1) They are rooted in equally powerful technologies—and their intelligent usage. In other words, they differ from traditional platforms in that they are not predicated on physical assets, land, and natural resources.

2) They benefit tremendously from vibrant ecosystems (read: partners, developers, users, customers, and communities).”

All these platforms are making this kind of impact in large part because they run on devices powered by one company: ARM. They match the above description, as a purveyor of intellectual property with virtual design and fabrication allies. ARM processors power social computing.

ARM’s technology does one thing better right now: it puts intelligent use in a person’s hand. The ARM approach is unique in that most tasks can be done in less than 1W, and some very complex things can be done in less than 2W – which correspond to the points of interest for smartphones and tablets respectively.

A vibrant ecosystem is an understatement, and it’s building smartphone and tablet momentum. Apple’s epic success, driven by dual Cortex-A9 cores in their A5 processor, is fueling a whole new set of changes driven by BYOD and consumer trends like mHealth and the “quantified self”. Qualcomm, about to set a new bar with their Krait implementation of the ARM instruction set, has taken a lead in smartphone processing innovation. NVIDIA and Freescale have gone quad Cortex-A9, with NVIDIA rumored to have taken the socket in the next-gen Amazon Kindle Fire. TI and ST-Ericsson are on dual Cortex-A15, with first silicon just seeing the light of day. Microsoft, having swung the Windows 8 machine in the direction of ARM, is part of the equation – and we’re about to find out more about that February 29th at Mobile World Congress.

The upper right of ARM’s “Gods and Giants” roadmap shows new 64-bit cores, Atlas and Apollo, targeting 20nm implementations of ARMv8 with an eye on growing into the server market. Here the ecosystem isn’t as thriving yet, with names like AppliedMicro and Calxeda at the front. Crossing Intel’s Sandy Bridge in that direction may prove just as difficult as Intel coming across toward the smartphone side.

Where things are thriving is the center of the roadmap, for smartphones and tablets. They tipped their highly efficient Cortex-A7 core last fall, and is now looking deeply at a “big-little” processing approach where the A7 does all but the biggest tasks, with a Cortex-A15 dozing patiently until needed for brief heavy lifting. They’ve hinted at a similar big-little approach with two new graphics cores, Skrymir and Tyr, with few details so far. It seems they are looking at smooth, dynamic shifting between cores sized for the task instead of simply tossing more cores into the mix. It’s all about the 1W.

What do you think of the state of social computing, the idea of these as the new platforms, the contrast between ARM and Intel, and what’s on the horizon from ARM?

Posted on

The Old Order Changeth

The Old Order Changeth
by Paul McLellan on 02-07-2012 at 3:02 pm
Categories: Uncategorized
9 Comments

It is interesting watching as changes in technology bring giants to their knees. Far and away the best book on the subject is Clayton Christensen’s The Innovator’s Dilemma. If you haven’t read it then rush out and buy it immediately. In tech, you are not educated if you haven’t.

Two things made me think about this recently. One is Kodak filing for bankruptcy. Some commentary is basically critical that Kodak were blindsided by the digital revolution. But in fact they not only pioneered early digital cameras, but also predicted almost exactly the speed of adoption by government (think spy satellite), then business, and then the consumer. The problem was that nobody knew how to make money on digital cameras which are a one-off low-margin sale. It was classic Innovator’s Dilemma stuff: everyone saw it coming and even so, nobody knew what to do about it. Curiously, FujiFilm, Kodak’s big competitor has fared better and managed to diversify into other markets. But that is always a risk strategy. Most forest product companies that try to get into electronics do not become Nokia, one of the most amazing business transformations ever. Now in turn the standalone digital camera industry is itself under threat, at least the the low end. When your cell phone has an 8 megapixel camera and is always in your pocket, why do you want another camera that is only marginally better.

Talking of cell phones, here are a couple of anecdotesabout changes in technology leadership. Apple’s iPhone business is now larger than Microsoft (in revenue). Not just Office or Xbox, but all of Microsoft, a company that recently was regarded as so powerful that it should be broken up since it would otherwise be a perpetual monopoly. But here’s an even more amazing fact. Apple is now criticized as being too dependent on iPhone, a one-product company. But if you take away Apple’s iPhone business completely, then the rest of Apple is still bigger than Microsoft and pretty well diversified.


Of course for EDA, companies that ship huge volume are not a dream but more of a nightmare. Apple designs one or two chips per year and so have comparatively modest demands for EDA tools. Of course they also buy lots of silicon from Qualcomm, Broadcom and memory suppliers in particular. But since EDA doesn’t share in volume, it shares in design starts, it benefits from less concentrated market power. Lots of competitors all designing their own chips is the EDA dream (and EDA doesn’t really care if chips go into production). As a handful of companies become more and more dominant in the end markets, shipping enormous quantities of relatively few designs, I think it will be a challenge for EDA to maintain its business models unchanged. And, as has been pointed out on this site many times, it will also have a major impact on where the chips are manufactured and from where the capital for the fabs comes.