Wednesday, June 29, 2011

NCCAVS Junction Technology Group Meeting at SEMICON West

Be sure to check out this annual event. This year, it will be held Thursday, July 14, 8:30AM-5:00PM in Moscone Center, South Mezzanine, Room 250. The topic is semiconductor and solar junction technology. Co-chairs are John Borland and Michael Current. For more info, go to http://www.avsusersgroup.org (Debra Vogler).

Monday, June 20, 2011

Wanna be AMD's CEO? Here's your checklist

AMD's got a problem: Nobody wants to be its CEO, or at least those that have been pre-selected as prime candidates appear to be uninterested. According to reports, that list includes: Apple CEO Tim Cook, Oracle copresident (and ex-HPer) Mark Hurd, EMC COO and ex-Inteler Pat Gelsinger, and Carlyle Group managing director Greg Summe. Interim CEO (and current CFO) Thomas Seifert has said he doesn't want to be considered for the permanent job.

One thing an AMD CEO candidate will want to know is the BoD's dynamics: is there consensus on a strategic direction, and how much input will the new exec have in setting strategy? "In this case the search is not just about finding a proven CEO who fits into the culture, but much more importantly, a transformative leader," Louis Gerhardy, a former exec-search consultant and market analyst, told Bloomberg. "The selection will have a profound impact on shareholder value at AMD."

It can typically take six months to a year for high-profile position like this; "finding the right candidate is important, more important than hitting a specific time window," Siefert said in the company's 1Q11 call. But "if by the end of the summer they haven't done something, then they've got a problem," Raymond James analyst Hans Mosesmann told Bloomberg.

So what else should the new AMD CEO -- whoever it is -- expect to deal with? Uche Orji of investment bank UBS has a checklist:

Clear the air. One challenge that wraps over all others: the need for a stronger strategic view that moves AMD beyond a current perceived role (justified or not) as just an alternative to Intel, Orji says.

Mobile strategy: "religious" roadmap, beyond tablets. Its rivalry with Intel aside, AMD has not been seeing growth inroads into mobile devices (tablets and smartphones), which apparently was the key sticking point between the company's board and ex-CEO Dirk Meyer. AMD's first-gen product gives it a toehold in tablets, but it's a good start, "especially as AMD can differentiate on graphics," Orji notes. But adopting the ARM architecture (either organically or through M&A) may not be the way to go, as it "could dilute resources that would hurt its ability to compete in its core x86 market," outside of which it hasn't executed well.

Unfortunately, everyone else is pursuing those markets too, so AMD needs to also take care and not become "the umpteenth chip firm" with a me-too ARM-based processor."

Leverage graphics. Better microprocessor/graphics integration was the reason AMD took over ATI back in 2006. Now, the company needs to take its fight with Nvidia in the consumer space and push harder into professional graphics, Orji argues.

Server focus. Going fabless has had some advantages (e.g. no exorbitant spending on next-gen process and fab facilities), but at the same time there's a concerning trend: AMD's share sunk to 7% in 1Q11, down from 22%. ("The lesson: owning server marketshare is more important than not owning a new fab," summed Credit Suisse's John Pitzer.) Orji thinks the new AMD CEO needs "a more viable go-to-market strategy" that approaches server chips as "a service business," with appropriate sales and support underpinnings.

Elsewhere, get friendly. Partner on baseband, connectivity, and software.

Reduce debt. AMD's long-term debt is around $2.2B (total debt almost $2.3B; that's half of what it was from 2007-2009. But with ~1.75B in cash, that's a debt-to-equity of 143%, way way above peers (single-digits) and 5-year return-on-equity is in the red.

Thursday, June 16, 2011

ElectroIQ Moves to New Platform

The ElectroIQ website is now residing on a new platform, giving you new functionality, faster load times, improved search and easier navigation. It will also make it easier for us to post new content, so we will be able to more quickly deliver a greater array of news, feature stories, blogs, videos, podcasts, white papers, wires news feeds and more.

One of the most notable changes is how the navigation bar works. In the past, the topic centers for each of our channels (semiconductors, photovoltaics, packaging, nanotech and MEMS) appeared immediately below the nav bar.





Now, they appear in an easy-to-use drop down menu.













The new platform also implements new search functionality using a product from a company called Ramp. This product, called the MediaCloud, gives all users the ability to sort, navigate and share search results across all of our content, including video, audio, text and images. With video and audio files, MediaCloud produces time-stamped transcripts from the original files. Text articles, images and the transcripts are processed through a natural language processing technology to create a unified set of tags and categories for all of our content.

The platform itself is a web content management system from Adobe (formerly Day Communique) called CQ5. Adobe calls it simple, intuitive web publishing. It is mostly invisible to users but easier for us editors, with in-context, drag-and-drop page design; in-context content authoring; in-context digital asset editing and keyboard shortcuts for native application feel. Check out a video Adobe produced!

No software migration come with perfect results. If you notice any problems, please let us know (just shoot an e-mail to psinger@pennwell.com).

A big thanks to the PennWell digital media team for making this happen. A special thanks goes to Robert Williams, the digital product manager for the ElectroIQ website; Meredith Courtemanche, digital media editor; and to all the folks in PennWell's Tulsa office, particularly David Warren, vice president, digital media; and Leif Nevener, senior product manager. Also a special thanks to Nicholas Iervolino who worked through the night to make it possible.

Now that we are working with a greatly improved backend program/platform, we will be quickly focusing on enhancing the look and feel of the site and adding some exciting new coverage this Fall. Stay tuned for more information. (Pete Singer)

Wednesday, June 8, 2011

Designing with eWLB packaging technology

Rajaran Pendse, VP of the Technical Marketing Group at STATSChipPAC, told MEPTEC lunch forum attendees (6/8/11, Biltmore Hotel, Santa Clara, CA) that packaging technology and functional integration have progressed on three parallel fronts: package level, wafer level (fan-in/fan-out WLP), and at the silicon level (TSV). “There is a convergence towards silicon and wafer level approaches driven by I/O density trends,” said Pendse.

The company is already in HVM with standard eWLB technology (embedded wafer-level ball grid array) – a fan-out wafer-level technology that Pendse says is a breakthrough in terms of high routing density, low parasitics, and small form factors. The cost efficiency comes from the elimination of substrates, bumping, and underfill – the limiting factors in flip-chip packaging. Pendse also described work that the company is doing on more advanced versions process.

The process flow for eWLB comprise four basic steps: reconstitution of dies to an “artificial” wafer, redistribution, ball application and singulation, and test/route/scan/pack. To determine whether or not using eWLB will be cost-effective, the company has developed a design metric called “fan-out ratio” (FR). [It’s defined as: (I/O density of the die-substrate)/(I/O density of substrate-PCB).] When FR<1, the design calls for using WLCSP as the solution; when FR is between 1 and 3, the design calls for using eWLB; and when FR>3, the solutions should be either fcFBGA or fcBGA. “We have used FR as a parameter to define the application space and cross-over point among different packaging choices,” noted Pendse. (Debra Vogler)

Tuesday, June 7, 2011

Design the grid to be DC at Research @ Intel Day (6/7/11, Computer History Museum, Mountain View, CA)

According to Intel Researcher, Guy AlLee, if engineers could design the grid today, they would make it DC. According to AlLee, the world is becoming DC, e.g., 80% of the power handled by power electonics is DC, alternate energy (PV, wind) is DC, and energy storage (batteries) is DC. The key to efficiency says Intel is to eliminate conversions and higher voltage.



-- Debra Vogler

Intel demonstrates "pedal power" for emerging markets at Research @ Intel Day (6/7/11, Computer History Museum, Mountain View, CA)


Intel's Frugal Computing project enables people in emerging markets to generate power by simply pedaling a bicycle. In this example, according to Intel Researcher, Venkat Natarajan, by pedaling for 5min, a person could generate enough power to charge a cell phone for 5hrs, or provide lighting for 5hrs. By pedaling 10min, a person could generate enough power to charge a netbook for about 2hrs.



-- Debra Vogler

Intel's Smart Vehicle: from the Research @ Intel Day (6/7/11, Computer History Museum, Mountain View, CA)



Intel Researcher, Joseph Pitarresi, described what Intel is developing to enable "anywhere car access." Such access creates an extensible client-to-cloud architecture for connected cars and service. The technology also brings the IA continuum of secure, personal computing experiences inside vehicles, noted Pitarresi. With one touch, the car, cloud and phone can be paired.



Also listen to this interview with Vijay Kesavan on the topic:


--Debra Vogler

"Magic Mirror" at the Research @ Intel Day (6/7/11, Computer History Museum, Mountain View, CA)


Intel Researcher, Nola Donato demonstrated the Magic Mirror: a gesture-controlled parametric body model display using Intel's high-performance rendering engine. Applications include retail (for digital signs), online clothing sales, clothing manufacture, and entertainment. The long-term goal is a virtual dressing room with a realistic avatar of the consumer that can try on the latest fashions.



-- Debra Vogler

Photos from Research @ Intel Day (6/7/11, Computer History Museum, Mountain View, CA)



A robot getting a “shave” from another robot greeted visitors to the Research @ Intel event at the Computer History Museum (6/7/11, Mountain View, CA). The robots are a send-up of the Intel TV commercial in which a man gets a shave from a robot (Debra Vogler).






Monday, June 6, 2011

Intel as a foundry?

Intel CFO Stacy Smith is on record saying that combining other chip design IPs with Intel's architecture core "would be fantastic business for us." That, of course, is spawning all sorts of speculation about the company's future plans, and whether it could be angling for business from consumer juggernaut Apple.

With its recently unveiled 22nm trigate structure set to ramp later this year, maintaining the company's 1-2 year technology headstart over competitors, Intel could, if it wanted, throw open the doors for some foundry customers

It's not a slam-dunk decision, though. Putting Apple or Sony's IP onto an Intel core is one thing; it's another entirely if a customer wants a custom-designed core. "Then you are only getting the manufacturing margin," noted Smith, and "that would be a much more in-depth discussion."

Smith was quick to point out that no proposal from Apple or others is in the works right now, and that foundry isn't a driver of Intel's capex.

So how much of this is real?

Eric Sherman over at BNet thinks there could be a few reasons why Intel would take on foundry orders: it wants to learn more about making low-power mobile chips (unlikely given IP restrictions), or there's a lot of money available (Intel isn't hurting for sales). The only reason that makes sense, he argues, is that Intel sees, and wants to head off, a long-term trend of falling utilization due to consumer purchasing shifts.

Piper Jaffray's Gus Richard echoes the idea that Intel is on the hunt for fab-filling foundry orders, citing scuttlebutt that Intel has been talking to OEMs -- including Motorola, a Toshiba customer -- and is looking to hire ASIC engineers and support staff. He pulls in other possible customer names into the mix, including EMC, Cisco, Juniper Networks, Sony, Apple, Nokia, and others who need leading-edge logic, and the x86 architecture -- i.e. not ARM or other competitors. What it's all boiling down to, he says, is "will Intel learn how to design its way out of a PC, or will the foundries overcome the process complexity?"

And at its recent investor day, Intel "sent what we thought was a clear message that it was very interested in some type of foundry relationship with AAPL," notes Susquehanna analyst Christopher Caso. Firstly that means Apple's business is indeed up for grabs, i.e. not locked in by previous partner Samsung -- nor a shoo-in for speculated new partner TSMC. Which, secondly, is someone Intel is increasingly thinking of as a process-technology competitor, so INTC wants to keep business away from them, especially any ARM-based competition.

Update: Citigroup's Glen Yeung is adamant that an Intel-Apple foundry partnership is on the horizon, perhaps for the A4 and A5 chips (currently Samsung-made) -- but probably on the condition that Apple converts to the x86 architecture for handsets and tablets, not an ARM-based core. INTC's recently-announced trigate architecture is the main draw for Apple, he notes. TSMC also stands to win some Apple foundry business, though the challenges of shifting to 28nm may spook AAPL into siding with Intel -- and in fact that might be why Intel isn't porting its Atom chips to 22nm just yet, Yeung says.

Yeung adds that fabbing Apple's 122mm A5 chip would require ~24,000 wafers/month, which is roughly equivalent to an Intel fab, but requirements would double with Apple's next chip and any quad-core parts.



A history lesson in M&A fundraising

How much have things changed in 13 years' of M&A? That's the amount of time passed from Applied Materials' most recent offering ($400M) and what it's raking together today to help fund its $4.9B deal for Varian Semi, a mix of five-, 10-, and 30-year bonds, notes Bloomberg's Tim Catts.

Some quick data pulled together by Bloomberg:

- The series of new notes all yield more than a full percent (107-175 basis points) more than similar-maturity treasuries, with interest of 2.65% (five-year), 4.3% (10-year), and 5.85% (30-year). Average data from Bank of America Merrill Lynch shows 3.64% for A-rated debt maturing in 5-7 years and 5.08% for ones due in 10-15 years.

- Borrowing costs have sunk a full three percentage points since AMAT's last $400M offering on October 9, 1997.

- On the other hand, so have returns -- yields on investment-grade corporate bonds were 3.86% on June 1, vs. 6.82% back in 1997.