Semiconductor & Patent Expert Consulting

IP Litigation expert consultant and patent expert witness for process, device, and circuit of  Dynamic

 Ram (DRAM), Flash  (NAND, NOR, EEPROM), and Static Ram (SRAM) Memories,

 and Microprocessor, Logic, and Analog Devices


Eight things to know about Intel's quad-core, Penryn, Silverthorne & mobile plans


Inside Intel

   Image Gallery


At Intel's Spring Analyst Meeting in New York City on May 3, chief executive Paul Otellini and his brain trust provided detailed insight into the chip giant's plans. Amid an intense focus on cost-cutting, following the tough business environment of the past few years, Intel emphasized its intentions to go full-bore with new technologies.

On tap are numerous dual- and quad-core processors, the debut of an advanced 45-nanometer manufacturing process, and an ambitious attempt to jump-start the widespread use of handheld Internet devices. After sifting through the information at the sessions, we present our summary of the most important takeaways from Intel's near-term roadmap, along with an image gallery of the most notable PowerPoints from the executive presentations.


Chips For Handheld Web Browsers

 Code names of the processors on Intel's roadmap. The Core micro-architecture, already here, will spread throughout Intel's processor lineup this year.

(click image for larger view)

Code names of the processors on Intel's roadmap. The Core micro-architecture, already here, will spread throughout Intel's processor lineup this year.

view the image gallery

Intel is calling the downsized device it sees as the next big thing the UMPC, for Ultra-Mobile PC. But it's really a Smart- or iPhone on steroids, with beefed-up Web browsing abilities.

"In the next couple of years, we'll see [the] beginnings of these mobile Internet devices," said Otellini. "They're around now, but not in great volume."

Intel has already launched its first UMPC-class processor, in the form of a chip called McCaslin. Interestingly, it's used not in a handheld browser but in the Apple TV. A more powerful chip—the one which Intel is really pitching for these handheld browsers—is code-named Menlow. It'll be ready in the first half of 2008. Unlike McCaslin, which Intel says runs Windows, Menlow will run Windows and Linux.

After Menlow, smaller, faster, and less power-hungry processors are planned. "This is just the beginning for us," said Anand Chandrasekhar, the general manager of Intel's ultra mobility group. "We get the base technology elements in place with the Menlow platform and then we iterate on that and get the power down and performance up."


More 'Core,' More Quad-Cores

What's in a "Core"? For Intel, the word cuts both ways: Core with a capital "c" is the name of the micro-architecture which is powering Intel's newest processors. Lower-case core is in abundance, too, at Intel, because dual- and quad-core processors constitute its most profitable chips.

Intel already offers several quad-core desktop chips, as part of its Core 2 Quad and Core 2 Extreme families. On the server side, Intel ships no less than at 9 quad Xeons, in its 5300, low-voltage 5300, and 7300 series.

 Harpertown and Yorkfield will be Intel's first 45-nm quads. (65-nm quad cores are already shipping.)

(click image for larger view)

Harpertown and Yorkfield will be Intel's first 45-nm quads. (65-nm quad cores are already shipping.)

view the image gallery

At the Spring Analyst event, Intel emphasized the Core architecture as a whole—which includes both two- and four-way chips--more than it did quad core specifically. Said Intel CEO Paul Otellini: "We will ramp our Core micro-architecture, top to bottom, in all market segments: Single, dual- and quad-core."

Core is characterized by a feature called "wide-dynamic execution." More importantly, it operates at lower power than does the Netburst architecture, which powered the Pentium 4. "We expect to be 100 percent, top to bottom, with [the Core] micro-architecture, by year end," said Otellini. "We are moving very aggressively to replace all the products, even down in the Pentium space and Celeron space with variants of [Core], over the course of this year. This gives us a performance lead, in each of those segments, and it gives us a serious cost advantage."

Nevertheless, it's clear that quads will become increasingly important for Intel, moving forward. On tap for release in the second half of 2007 are two new quads based on Intel's latest 45-nm chip technology: Yorkfield for desktops and Harpertown for servers.


Shift To 45-nm Chip-Manufacturing Technology

In the race to the next advance in chip manufacturing technology, called 45 nanometers, it looks like Intel might have as much as a one-year jump over AMD. (Forty-five relates to the size of the features etched into the chip.) AMD likely won't ship 45-nm processors until late in 2008.

Intel is readying four of its factories to crank out 45-nm chips (GRAPHIC HERE), said Robert Baker, general manager of Intel's technology and manufacturing group, at the Spring Analyst event.

"Penryn and its first-generation products on 45-nm come in the second half of this year," said Otellini. "Nehalem, the next generation micro-architecture, will ship in 2008 on 45-nm. We keep the cadence going in '09, with 32-nm silicon deployment, essentially a shrink of Nehalem. And then in 2010, Sandy Bridge, which is a new micro-architecture."

 Four chip-manufacturing fabs are ramping up to produce 45-nm processors..

(click image for larger view)

Four chip-manufacturing fabs are ramping up to produce 45-nm processors.

view the image gallery

An inherent benefit of the tinier features on a 45-nm chip is lower power operation. As well, shrinking the die size allows Intel to add more features onto the chip, explained Baker. For example, with Penryn, Intel is using that extra space to increase the cache size and to add new instructions, which make the processor more adept at handling video and multimedia.

For Intel's chip designers, going with more features is not a no-brainer. As Baker explained, there's a tension between using the addition space to add more features or to reduce cost. (The latter would be done by not adding new features, and using the unused space put more chips on each wafer, thus upping the yield.)

Moreover, the move to 45-nm doesn't mark the fruition of years of unimpeded advances toward ever-smaller feature sizes. Rather, companies such as Intel and IBM have in recent years grappled with the limitations of silicon—both practical and fundamental physics—as on-chip element more towards sizes where some things are only a few elements big. The biggest problem has been what's called "leakage," where current doesn't stay where it's supposed to.

That's led to a search for new materials. Intel thinks it's achieved a big breakthrough with a new material called High-k metal, which replaces the polysilicon that's been used for the past 30 years. "It's a fundamental change in materials," said Baker.

Moving forward, Intel will use the extra space afforded by 45-nm to begin integrating graphics handling into the processor itself. Interestingly, that's something some chip designers inside Intel urged the company to do in the early 1990s. Reportedly, that tack was rejected at the time because Intel more business advantages in keeping non-CPU features on ancillary chipsets, which could be sold separately.

Forty-five nanometers will also be used to built processors with more than eight cores. One such design, code-named Larrabe, is already on the drawing board. According to Otellini, Larrabe will "address very, very high-performance graphics and high-performance computing needs."


Server Platforms With 8 Cores And More

Intel is working hard to fend off the upcoming challenge from Barcelona, AMD's soon-to-be-released quad-core server chip. AMD is positioning it as the industry's first "native" quad-core, which means it was designed from the ground up to fit four processors onto a single piece of silicon. In contrast, AMD points out that Intel's existing quad chips cram two dual-core devices into a single package. (Intel CEO Otellini has famously argued: " "I think you'd be misreading the market if you think people care about the packaging.")

 Server chips, both dual- and quad-core, will continue to bear the Xeon brand name.

(click image for larger view)

Server chips, both dual- and quad-core, will continue to bear the Xeon brand name.

view the image gallery

Right now, Intel's head start in quad core is reaping benefits. At the Spring Analyst Meeting, Otellini said the quad-core Xeons are being snapped up, particularly in a configuration known as DP, for dual processor. This "DP" refers not to the internal configuration of the chip, but rather to the number of sockets in the server. So, if there are two sockets on a server motherboard, and each socket houses a quad-core processor, the end result is a mightily powerful, eight-processor server.

"In the first quarter of this year, the quad core version of our DP servers out-shipped our competitor's DP line entirely," said Otellini. "And quad continues to grow as a percentage of our DP shipments. Why? Because quad-core DP is the sweet spot for server deployment. Think of it as a very cost effective eight-way machine: Two processors, each of which has four processors on it."


Cutting Back Spending, Canning More Workers

If you can't grown revenues solely by selling more stuff or by commanding higher prices (and, with microprocessors, the latter is difficult for all but the most powerful chips), the way to help boost your bottom line is to cut costs. That's precisely what Intel is doing. The paring will proceed throughout the year, according to the presentations made by Otellini and chief financial officer Andy Bryant at the Spring Analyst Meeting.

 Cost cutting is a significant element of Intel's financial strategy.

(click image for larger view)

Cost cutting is a significant element of Intel's financial strategy.

view the image gallery

As an Otellini PowerPoint summed up the strategy: "Lower cost structure and focus on designing for costs increases our competitiveness in current and new markets;" and "Result: Bottom line growth exceeds top line growth for 2007/2008."

Bryant's presentation encapsulated Intel's management of its finances as "Good progress, more to do." His handouts noted that Intel's capital spending will dropped more than $250 million, or approximately 5 percent, in 2007 compared to 2007. (However, overall capital spending will still be a hefty $5.5 billion.)

Such tight controls means Intel must be very focus on execution as it works to roll out so much new technology, because building processors isn't getting any cheaper. "The cost to bring a leading [edge] product family to market is about $3 billion," Bryant said during his talk.

Intel's cost controls also involve reducing its internal headcount. In April, Intel reported it had reached its goal of reducing its workforce from 103,000 employees to 92,000. "That will continue to drop," Otellini said at the Spring Analyst Meeting.

Otellini didn't provide a precise number. However, recent reports indicate that Intel is trimming 1,000 workers from a chip factory in New Mexico, which is about to undergo renovation.



Big Push Into Emerging Markets Like China And India

For all players in the PC area, chip vendors and systems houses alike, the United States is a mature market, offering nowhere near the kind of percentage growth available a decade ago. But Asia remains a largely untapped market, where most people have yet to purchase their first PC.

An added twist is that, in contrast to the affluent domestic market where gilt-edged processors have been eagerly snapped up, Intel sees cheaper, more stripped-down designs as an important component of its overseas push. Indeed, Intel sees a market of upwards of 130 million people who could be ready to buy soon. "These are all people who are just moving into the space of being able to afford computing," Sean Maloney, the general manager on Intel's sales and marketing group, said at the Spring Analyst event. "We believe the majority of those people will buy notebooks."

To take advantage of the perceived strong overseas demand, Intel is offering a sometimes confusing amalgam of notebook, plain-old PC, and new lighter-weight mobile platforms. There's a low-cost PC platform. Separately, in what seems like a response to MIT's $100 One Laptop Per Child, Intel has designed a $300 "Classmate PC." It's a laptop with a Celeron processor, 2-GB flash drive, and a 7-inch screen.

Also on tap is Silverthorne, chip which seems to have several roles. Intel mostly talks about it as a standalone processor which will power UMPC and handheld Web browsers. However, since Otellini characterized it as having "circa 2003/4 mainstream mobile performance," it could clearly find its way into low-cost notebooks aimed at emerging markets

There are also many traditional style notebook platforms on the way (see next section).


More Notebooks, More Wireless

Intel also sees notebook sales growing in the United States, where it says that current penetration is less than "half a notebook per household." Accordingly, it's planning a bunch of new laptop platforms.

 Santa Rosa is the next-generation Centrino notebook platform.

(click image for larger view)

Santa Rosa is the next-generation Centrino notebook platform.

view the image gallery

"We will introduce our fourth-generation Centrino next week [week of May 6]. It's a platform code-named Santa Rosa," said Otellini. Santa Rosa features the new Merom processor, but in 802.11n wireless, integrated graphics, and Intel's new turbo memory. Otellini said he expects a very steep ramp-up for Santa Rosa, and said it will constitute a majority of Intel's notebook shipments in fairly short order.

The Montevina laptop platform will follow in 2008. Equipped with a 45-nm, dual-core processor, it's be perhaps most notable for its mix-and-match wireless. Montevina will have both standard WiFi and the newer WiMAX, which Intel is continuing to emphasize in hopes it'll soon move into position as the dominant wireless standard of choice.


System-on-Chip (SoC): Tolapai & Silverthorne

Tolapai and Silverthorne are an interesting pair, which mark something of a departure for Intel. The semiconductor giant has heretofore always offered processor products; that is, chips an OEM or retail customer could buy off the shelf.

 Systems-on-a-chip may be the most novel component of Intel's long-term technology strategy.

(click image for larger view)

Systems-on-a-chip may be the most novel component of Intel's long-term technology strategy.

view the image gallery

A System-on-Chip (or system-on-a-chip) is a different slice of silicon entirely. At the 50,000-foot level, one can think of an SoC not so much a product, but rather a collection of intellectual property for building a chip. That is, the semiconductor house has on its shelf all the different blocks of the processor—CPU, logic, RAM, even ancillary stuff like the graphics engine.

When an OEM comes along and wants a specific version of that chip, customized for a particular product it wants to make, the semiconductor vendor springs into action. It pulls the blocks it needs off its virtual shelf, puts them together, burns a mask, and sends the whole thing over to its fab. Voila, customized chip.

This is what Intel envisions for Tolapai and Silverthorne. (The latter will also be offered as an off-the-shelf part.) The designs seem to be aimed high and low, respectively.

Here's how Tolapai was explained in press release issued by Intel during its Developer Forum in Beijing in April:


System on Chip Plans, Enterprise " [Intel senior vice president Pat] Gelsinger unveiled "Tolapai" plans, the first in what will be a family of enterprise-class "system-on-chip" (SoC) products that integrate several key system components into a single Intel architecture-based processor. The 2008 Tolapai product is expected to reduce the chip footprint size by up to 45 percent and power consumption by approximately 20 percent compared to a standard four-chip design, while improving throughput performance and processor efficiency. Tolapai will include the new Intel QuickAssist Integrated Accelerator technology.

In contrast to the enterprise-oriented Tolapai, Silverthorne, as mentioned above, is intended for use in UMPC-class handhelds.

Intel's dipping of its corporate toes into the SoC waters is significant because, traditionally, SoCs have often been aimed at market segments where there wasn't enough volume to justify the launch of a one-size-fits-all product. Intel, which lives on large volumes, has always stuck with the big stuff.

Perhaps, in opting to try SoC, Intel will be engaging in something of a loss-leader strategy. That is, perhaps its signifying that it's content to make less money in the UMPC market (with Silverthorne) to begin with, if it can help nurture that market. (Remember that the ultra-mobile segment doesn't really exist today. Remember, too, that Intel's full-frontal assault on the cellphone chip market failed; perhaps that legacy has inspired an alternate tack.)

How does Tolopai, which is aimed at the high end, dovetail with this assumption? Possibly the high end is where Intel can learn the most about getting its SoC technology up and running. Perhaps, too, Intel envisions a day not too far distant where the leading edge of computer (where 16-, 32-, and 64-core devices are used) is not longer a commodity space, and thus it wants to get a jump on SoC while the getting's good.