Processor Watch

January 31, 2005

Editor: Tom R. Halfhill

In this issue:

Multicore Chips Rule in 2004
Tom R. Halfhill - Senior Editor  {01/31/2005}

PC processors boast the highest clock speeds. Server processors have the fattest caches. But unsung embedded processors are at the forefront of microprocessor evolution. As we noted last year, some of the architectural advances appearing first or most extensively in embedded processors are massively parallel processor arrays, on-chip interconnect fabrics, reconfigurable logic, DSP extensions, extendable instruction sets, and hardware-assisted simultaneous multitasking. In 2004, high-performance embedded processors also set the pace for multicore designs.

While the PC market is agog at dual-core 64-bit processors, the embedded market already takes such chips for granted and will deliver processors with four, eight, and sixteen 64-bit cores this year. The rising demands of telecommunications and network processing continue to drive aggressive innovation. Only the need for power efficiency restrains embedded chips from matching the high clock frequencies of PC processors and the bloated transistor budgets of the biggest server processors.

Our year-end review of high-performance embedded processors finds that innovation in this market continued to accelerate in 2004. Looking forward, we expect most chip vendors will be preoccupied delivering the ambitious processors announced last year. However, we also expect some exciting new announcements in 2005, including the public debut of at least two vendors now working in stealth mode.

Our nominees for a Microprocessor Report Analysts’ Choice Award in the High-Performance Embedded Processors category: Broadcom’s SiByte BCM1480; IBM Microelectronics’ BlueGene/L; Stretch’s S5610; and Toshiba’s TX9956CXBG.

Microprocessor Report readers can access the full story (4 pages) here: www.mdronline.com/mpr/h/2005/0131/190502.html. To find out more about Microprocessor Report, please visit: www.mdronline.com.

Editorial: State of the Semi Industry
Kevin Krewell - Senior Editor  {01/31/2005}

In early January, I attended the Industry Strategy Symposium (ISS) in Half Moon Bay, California, run by the Semiconductor Equipment and Materials Institute (SEMI). SEMI’s ISS is an executive conference geared for semiconductor equipment and materials suppliers. The conference spent two and a half days looking at the short-term market (one to three years) for semiconductors and exploring some of the potential changes to the market over the next decade. The makeup of this conference audience was international, but Silicon Valley was of primary interest to many of the speakers and attendees.

There were some common economic trends throughout the two days I attended: a soft downturn in 2005, with a slow rebound in 2006; consumer products replacing information technology (IT) products as the primary driver of semiconductor growth; and concern over China’s impact on the industry.

In regard to the semiconductor industry, although many presenters expressed public confidence that the 64–45nm process node transition will go smoothly, there were also a few skeptics who believe the industry is underestimating the complexity of the 65–45nm transition.

Continual innovation will be required just to keep the industry on the long-term trend line of 8–9% annual growth. This will be true in the 65–45nm process node transition, as this transition will need more new materials than ever before.

Education Is the Key
One of the speakers was W. Brian Arthur of the Santa Fe Institute, and one of the topics he spoke about was how our education system would help us to stay ahead of the rest of the world. That it is our investment in education that feeds the high quality of our basic research, which, in turn, leads to new discoveries and innovations.

Unfortunately, unlike Mr. Arthur, I’m not convinced we are doing enough to stay ahead of the world on education. At the same time that we need to be investing more in education, overall government spending on education has not increased. In California, for example, the state budget crisis has put a squeeze on the California state university programs, forcing tuition increases and enrollment decreases. These are not developments that will strengthen U.S. global competitiveness.

Another speaker at the conference, Paul Saffo, related that China has had a massive increase in the number of universities (he said 400 universities are being founded) and is graduating tens of thousands of engineers every year. Although our university system is well established, whereas China’s is still in the early formative stage, just the sheer number of Chinese graduates will make a significant impact on the world’s manufacturers. Other speakers decried the present restrictions the U.S. government has put on visas, limiting companies’ ability to attract foreign talent to this country.

Aggregating the Analysts
The meeting included plenty of charts, graphs, bar charts, and arrows (up and down). We even had an industry “weather report.” None of the analyst firms represented at the conference believes the present slowdown will be as deep or as long as the 2001–2003 semiconductor recession. In-Stat’s own projections are on the more pessimistic side, however, with industry revenues shrinking 5.7% in 2005. The projections for 2005 ranged from –5.7% (In-Stat) to + 15 (Future Horizons), but the distribution centered at near zero growth (which is also the SIA projection).

As you might expect, we heard a lot about the boom-and-bust semiconductor cycle. This cycle shows no sign of going away, but there is evidence that the industry is finally beginning to deal with it more effectively. The projected downturn in 2005 will be softened, compared with previous downturns, by an earlier throttling of chip supplies with the onset of the downturn. The typical pattern in the past had been that the semiconductor industry would overbuild capacity during the boom time, only to find itself with an oversupply when demand eventually softened. The oversupply leads to a collapse in ASPs and higher production overhead, owing to the underutilization of fabs. A number of quarters are then required for the oversupply to be absorbed by increasing demand, and the industry eventually begins another growth spurt.

In the case of 2005, the general consensus of the analysts was that companies adjusted production in 2004 before reaching the brink, better anticipating the correction this time. In the past, it was the greed of missing additional business during the growth cycle that drove companies to continue building in the face of softening demand and overcapacity; now, it appears to be the fear of a painful downturn that is driving companies to be more cautious. The 2001–2003 downturn is all too fresh in everyone’s mind for them to become overly exuberant about this most recent (2H03/1H04) upturn. The exception to the rule that is scaring many at the conference is China, which is still building fab capacity at an alarming rate.

Many analysts believe the manufacturers anticipated this downturn better than previous ones, but none of the analysts would attribute the better planning to any particular factor. Maybe it was the conversion to better forecasting tools, or an accelerated supply chain, or more pervasive enterprise resource planning (ERP) software. The issue that did divide the analysts was whether 2005 would have a net positive (dollar) growth or a slightly negative one. None were expecting more than a single digit of growth or shrinkage on either side of flat. They did agree the market will grow in unit shipments, but that overall ASPs are under pressure.

The Digital Transition for Consumer Devices
David Steel, vice president of marketing for Samsung’s digital media business described that transition of consumer products like cameras, camcorders, VCRs, TV, and tape players (e.g., Walkman) to being digital devices. The issue of the convergence of many of these devices into a few multifunction devices had everyone’s attention, as CES had just ended. These new products and markets are exploding and will generate continued demand for the semiconductor technologies that made the digital revolution possible in the first place. This is where the cutting edge of the chip business is today.

The cutting edge of the digital revolution movement is Korea, where broadband penetration is the highest in the world. Korea is therefore becoming the test bed for the next generation of broadband wireless applications, which has proved convenient for Korea-based Samsung.

China on Everyone’s Mind
One of China’s policies that was the focus of some concern is the fixed Chinese exchange rate with the U.S. dollar. One analyst went so far as to call it economic warfare. The fixed exchange rate and the growing U.S. trade deficit with China are making China a major purchaser of U.S. bonds.

While many analysts were counseling companies to grow slowly in 2005 and 2006, it is evident that China is continuing to invest heavily in new fab capacity. China has only a handful of 300mm fabs today, but it is currently building tens of new fabs. With the projected slowdown of 2005, many hoped China would slow its building binge also. At present, China still represents only a small percentage of worldwide fab capacity, and it is not at the cutting edge—at least not yet.

There are also some limitations that a few noted about China. The country’s infrastructure, including roads and power, is still a work in progress. It was also noted that local Chinese materials suppliers are not up to world-class quality.

Mr. Arthur said that he believes China is the biggest economic story of the past 20 years and will be the biggest story for the next 50 years. He predicted China will dominate commodity markets, pushing the U.S. to stay ahead by innovating faster. The next waves of technology that will keep us ahead will be in biotech and nanotechnology. This is where maintaining our lead in the sciences will be essential.

Silicon Valley: Haven for Bad Management?
One of the guest speakers at the SEMI ISS meeting was Paul Saffo, director and Roy Amara Fellow from the Institute for the Future. Mr. Saffo provided the most quotable line of the conference: Bad management is not a bug, it’s a feature [of Silicon Valley]. Mr. Saffo then went on to list some of the key Valley entrepreneurs, such as Steve Jobs and Larry Ellison, whom no one would consider easy managers to work for. The defining example may well have been William Shockley, who, had he been a good manager, might not have inspired (actually driven) the “Traitorous Eight” to leave Schockley Semiconductor and start Fairchild Semiconductor. And then later, had Fairchild been such a great place to work, the founders of Intel, National, and AMD wouldn’t have been inspired to leave Fairchild. It also inspired the comic strip Dilbert written by ex-Silicon Valley engineer Scott Adams.

Furthermore, innovative technologies like Java and products like Acrobat survived being killed in their early stages only because management overlooked them. Saffo called innovation “irrational” and “deeply, deeply weird.” Which is not surprising, as innovation is a creative process, and the creative process is difficult to schedule.

When Failure Is Essential for Eventual Success
The other key to the success of Silicon Valley is that it is a place built on the edge of failure. Many of the new technologies, innovations, and products are built on the failure of pioneers. One example was the PDA business, which was built on the failure of Apple’s Newton. We could also be seeing one of the pioneers of the digital video recorder, TiVO, fail to capitalize (financially) on this new market it helped create. It could be that Silicon Valley has learned how to fail correctly. It is also because the Valley is a place where people dream they can change the world. The question Saffo posed to the audience was this: Will it continue to be that place, or will India and China become the next centers of innovation?

Despite all the talk of an industry downturn, disruptive technologies, and the high cost of new process technology, the conference was still reasonably upbeat. One thing that particularly cheered everyone at the conference was the announcement that at its earnings release, Intel announced plans to increase capital spending in 2005. Hope springs eternal.

My own take was that this is another structural change for Silicon Valley, much like the major upheaval in 1985. In that past transformation, the Valley began losing its manufacturing base but maintained the engineering and corporate functions. This next transition appears to be carving out some of the engineering and administrative functions and moving them to foreign locations. It still leaves in the Valley the higher-value engineering, marketing, and corporate leadership. Once more, the Valley needs to stay on the top of the value chain, which I believe it can accomplish. Unfortunately, I believe some jobs will be moved offshore, never to return. The education and immigration issues do concern me, because without both education and immigration, it will be harder for us to stay on the top of the value chain in the future.

To find out more about Microprocessor Report, please visit: www.mdronline.com.

Chips, Software, and Systems
Kevin Krewell - Senior Editor  {01/31/2005}

Each year we look back at the technologies that were either revealed during the previous year or made a significant impact in that year. The list of earlier winners is quite diverse, ranging from the processor and system architecture of IBM’s Power 4, to the EUV lithography developments that promise to extend Moore’s Law scaling beyond the limits of optical lithography (and well beyond the end of this decade), to the software architecture of Microsoft’s Next-Generation Secure Computing Base and Intel’s mainstream multithreading (Hyper-Threading). This year’s candidates also mix hardware, software, and systems. Each clearly cannot exist without significant cooperative efforts in all three areas.

We see three technologies making a big impact during the next few years:

• The first is heavily threaded and multicore (beyond two) mainstream processors.
• Virtualization, available for many decades in mainframe servers from IBM, will make it to mainstream PCs in 2006.
• The Cell processor, being developed by IBM, Sony, and Toshiba at the STI Design Center, promises to be a new architecture optimized for broadband media and 3D graphics performance, but with uses beyond the game console.
  

With the promise of incredible performance and scalability, and what is basically a vector supercomputer on a chip, we present the 2004 Microprocessor Report Analysts’ Choice Award for Best Technology to the Cell Processor. Congratulations to the STI Design Center and the participating employees at IBM, Sony, and Toshiba for what promises to be a very exciting challenge to mainstream processors.

Microprocessor Report readers can access the full story (3 page; 1 graphic) here: www.mdronline.com/mpr/h/2005/0131/190501.html. To find out more about Microprocessor Report, please visit: www.mdronline.com.

• Most recent Processor Watch articles
• Processor Watch archives