Nanotech - The Circuits Blog

The Hot Chips conference in Palo Alto, Calif this week is focusing on high-end chips for servers and scientific computers, with IBM's upcoming Power7 as a standout.

On Tuesday, IBM will give a presentation on its next-generation server chip, the Power7. IBM documentation describes the chip as having up to eight cores. A dual-chip module holds two processors for a total of 16 cores, according to IBM.

Each core has a rated performance of 32 gigaflops, providing 256 gigaflops per processor--one of the fastest chips to date based on this scientific-centric performance benchmark.

Power7 will be used in the National Center for Supercomputing Applications "Blue Waters" supercomputer, the first system of its kind to sustain one petaflop performance on a range of science and engineering applications, according to the NCSA. A petaflop is one quadrillion floating point operations per second.

Power7 "will be the first of a powerful new system design from IBM. The design includes extensive research and development in new chip technology, interconnect technology, operating systems, compiler, and programming environments," according to the NCSA.

Other chips to be described at the conference include the Sparc64 VIIIfx: Fujitsu's new 8-core processor for Peta scale computing. Sun will discuss its "next-generation multi-threaded processor Rainbow Falls" and AMD will spell out its Magny Cours processor, 12-core chip.

Intel will present a paper on its upcoming Nehalem server processor.

Intel will also discuss Moorestown, an upcoming version of the Atom processor targeted at mobile Internet devices and smartphones. Intel will also give a presentation entitled "Understanding the Intel Next Generation Microarchitectures (Nehalem and Westmere) transitioning into the Mainstream."

As chip geometries get infinitesimally small, IBM is looking to DNA to make the manufacture of future chips feasible.

On Monday, IBM researchers and collaborator Paul W.K. Rothemund, of the California Institute of Technology, announced an advancement of a method to arrange DNA origami structures on surfaces compatible with today's semiconductor manufacturing equipment.

Low concentrations of triangular DNA origami bind to wide lines on a lithographically patterned surface.

(Credit: PRNewsFoto/IBM)

"The cost involved in shrinking (chip) features to improve performance is a limiting factor in keeping pace with Moore's Law and a concern across the semiconductor industry," said Spike Narayan, a manager in the Science & Technology division of IBM Research, in a statement.

Moore's Law, named after Intel co-founder Gordon Moore, states that the number of transistors that can be placed on an integrated circuit doubles roughly every two years. For more than four decades, chip manufacturers have been able to consistently shrink chip geometries, allowing Moore's Law to remain on track.

But this may not be sustainable for chips with geometries under 22 nanometers. By 2014, the high cost of semiconductor manufacturing equipment will threaten Moore's Law, "altering the fundamental economics of the industry," according to a report released in June by iSuppli. New chip plants typically cost billions of dollars to build, and the tab goes up as chip circuits get smaller.

Individual triangular DNA origami adhere to a template with properly sized triangular features.

(Credit: PRNewsFoto/IBM)

IBM uses DNA molecules as scaffolding--where millions of carbon nanotubes could be deposited and self-assembled into precise patterns by sticking to the DNA molecules. This approach might provide a way to reach sub-22-nanometer lithography--down to 6 nanometers--more economically, according to a paper to be published in the September issue of Nature Nanotechnology, entitled "Placement and orientation of DNA nanostructures on lithographically patterned surfaces." It was co-authored by IBM and Caltech scientists.

"The utility of this approach lies in the fact that the positioned DNA nanostructures can serve as scaffolds, or miniature circuit boards, for the precise assembly of components, such as carbon nanotubes, nanowires, and nanoparticles," according to IBM. The combination of self-assembly with today's fabrication technology eventually could lead to substantial savings in the most expensive and challenging part of the chipmaking process, IBM said.

The lithographic templates, for chip fabrication, were made by IBM using traditional semiconductor techniques, the same used to make the chips found in today's computers, to etch out patterns.

Globalfoundries, the manufacturing concern spun off from Advanced Micro Devices, plans to announce Wednesday that it has signed up STMicroelectronics--its first true outside customer.

STMicroelectronics supplies the accelerometer for the iPhone 3GS

(Credit: Apple)

Globalfoundries was created last year in order to eliminate the crippling overhead that AMD was incurring to manufacture its processors. Globalfoundries now conducts business as a contract chip manufacturer, commonly referred to as a foundry.

AMD owns 34.2 percent of the company, while Advanced Technology Investment Co. owns the rest. ATIC is an investment company wholly owned by the government of Abu Dhabi, which is part of the United Arab Emirates.

Last week, Globalfoundries broke ground on a $4.2 billion facility in Malta, N.Y., that is expected to put it among the elite chipmakers of the world. Currently, Globalfoundries manufactures chips for AMD at facilities in Dresden, Germany.

STMicroelectronics will commission Globalfoundries to make low-power chips using a 40-nanometer process, which "is ideal for the next generation of wireless applications, handheld devices, and consumer electronics," according to a statement. Production is slated for 2010.

Currently, the most advanced manufacturing processes in the chip industry are at the 34-nanometer level--which Intel uses to manufacture flash memory chips. More typically, chips are made on advanced processes ranging between 40 and 50 nanometers. Generally, the smaller the chip's geometries, the faster and more power efficient it is.

Globalfoundries would not say what specific types of chips it will make for STMicroelectronics, whose product portfolio includes memory, microcontrollers, power management silicon, and MEMS or Micro-Electro-Mechanical Systems.

One of the highest profile products STMicroelectronics supplies today is the accelerometer for the iPhone 3GS. Based on MEMS technology, the accelerometer allows the 3GS to determine device orientation or inclination.

STMicroelectronics was ranked among the top 5 chipmakers in the world based on revenue in 2008, according to market researcher iSuppli.

Advanced Micro Devices' spinoff Globalfoundries will use a groundbreaking ceremony this Friday to proclaim that it's ready to go head to head with the world's largest chipmakers.

Globalfoundries, formerly the chip manufacturing operations of AMD, will break ground next week in Malta, N.Y., on a $4.2 billion facility that may put it among the elite chipmakers of the world. But to get there, it needs big customers. One--or more--of those names will be announced in the next 30 days, according to Jon Carvill, director of corporate communications at Globalfoundries.

Early customers may include companies that design low-power and wireless chips for the consumer electronics market. "There are a lot of benefits to being a significant player in that space," Carvill said.

Globalfoundries also has graphics chip companies in its sights. One large company cited as a potential future customer--though not necessarily among the first customers--is Nvidia, the No.1 graphics chip supplier. Nvidia's CEO, Jen-Hsun Huang, has already stated publicly that he is "seriously evaluating and discussing" the possibility of using Globalfoundries as a contract chip manufacturer, or so-called foundry.

An even more likely customer is ATI Technologies, AMD's graphics chip unit. Carvill said that Globalfoundries has "not run any volume (manufacturing) of AMD graphics processors...not yet," which may imply that there will be volume manufacturing at some point down the road.

Globalfoundries will compete aggressively for graphics chip business at the 28-nanometer manufacturing process level, which is still a few years away. Nvidia and ATI now use a 40-nanometer process in their most advanced graphics chips. Typically, the smaller the chip geometries, the faster and more power efficient the chip is.

It is no coincidence that both Nvidia and ATI currently use the world's largest contract chip manufacturer, Taiwan Semiconductor Manufacturing Company (TSMC) as their foundry. Nvidia and ATI are the kind of elite client that Globalfoundries is looking to wrest away from TSMC. Graphics chip companies are typically among the first to move to a new manufacturing process because they need to pack as many transistors--now up to a billion--as possible into a chip in order to maintain a performance leg-up on the competition.

Globalfoundries may have an advantage here. "They (TSMC) don't have the advanced engineering that AMD does," said Dan Hutcheson, CEO and chairman of VLSI Research, a market researcher that covers the chip manufacturing industry, referring to Globalfoundries in its previous incarnation as AMD. And TSMC has been struggling with its 40-nanometer-class manufacturing process, providing an opportunity for Globalfoundries, according to Hutcheson.

But that doesn't mean it will be a cakewalk. "TSMC is incredibly cost effective and they're very quick at turning things around," Hutcheson said. "That's why they hold 60 to 70 percent of the market."

Why is a quick turnaround important? "If you're late to market, you lose most of your market share in a few weeks," according to Hutcheson. "A chip has to be designed into a socket that's going to go into some electronics product that's going to show up on shelves at Christmas. You can't delay Christmas," he said.

When the facility in Malta is completed and up and running by 2012, it should be among the world's most advanced chip plants. Globalfoundries' current plan is to get started on 28-nanometer technology in the second half of 2012 and then make a quick transition to 22-nanometer. Currently, the most advanced chips are manufactured on 40- and 45-nanometer processes. Intel, whose production is used strictly for its own products and therefore is not a contract chip manufacturer like Globalfoundries, is expected to transition--on a volume basis--to 32-nanometer late this year or early next year.

Joint work with IBM
Another advantage Globalfoundries will have over its competition is close cooperation with IBM, also one of the world's premier chip manufacturers. "That's one of the reasons it's being put there," Hutcheson said, referring to the Malta facility. IBM develops and manufactures chips in New York, with a lot of activity concentrated in East Fishkill.

Last month, Globalfoundries announced at the 2009 Symposium on VLSI Technology in Kyoto, Japan, that it is working with IBM on a technique that allows transistors to be scaled down below 22 nanometers.

And if there's any doubt about Globalfoundries' commitment to work with IBM, the headcount speaks for itself. A total of 70 Globalfoundries engineers work at IBM. "And that number is growing," Carvill said.

Moore's Law may lapse by 2014, according to iSuppli. The high cost of chip manufacturing--not just the impossibly smaller geometries--may be the biggest threat.

Gordon Moore, former chairman and CEO of Intel

(Credit: Intel)

Moore's Law, named after Intel co-founder Gordon Moore, states that the number of transistors that can be placed on an integrated circuit doubles roughly every two years. For more than four decades, chip geometries have gotten smaller and smaller, allowing Moore's Law to remain on track.

By 2014, however, the high cost of semiconductor manufacturing equipment will threaten Moore's Law, "altering the fundamental economics of the industry," according to a report released on Tuesday by iSuppli.

"The usable limit for semiconductor process technology will be reached when chip process geometries shrink to be smaller than 20 nanometers (nm), to 18nm nodes," said Len Jelinek, director and chief analyst, semiconductor manufacturing, for iSuppli. "At those nodes (levels), the industry will start getting to the point where semiconductor manufacturing tools are too expensive to depreciate with volume production, i.e., their costs will be so high, that the value of their lifetime productivity can never justify it."

While further advances in shrinking process geometries can be achieved after the 20-nanometer to 18-nanometer level, Moore's Law will no longer drive volume semiconductor production, iSuppli said.

As a yardstick, Intel is currently in the process of moving to a 32-nanometer manufacturing process. While Taiwan Semiconductor Manufacturing Company (TSMC)--the world's largest contract chip manufacturer--has moved to 40-nanometer for chips it makes for companies such as Nvidia.

There are examples of companies that have already found chipmaking prohibitively expensive. Facing possible bankruptcy, Advanced Micro Devices eventually spun off its chipmaking operations. Some Asia-based memory chipmakers have also faced possible extinction because they couldn't invest the staggering sums of money necessary to update production facilities.

The end of Moore's Law has been prophesied more than a few times in the past but chip equipment cost isn't the only thing conspiring against the law. Exponential growth in every industry eventually has to come to an end, according to an April EE Times report quoting IBM Fellow Carl Anderson. He cited railroads and speed increases in the aircraft industry as examples where exponential growth eventually petered out.

"A generation or two of continued exponential growth will likely continue only for leading-edge chips such as multicore microprocessors, but more designers are finding that everyday applications do not require the latest physical designs," Anderson said in the EE Times' report.

Until 2014, however, the race continues. Globalfoundries, the joint company owned by AMD and Abu Dhabi-based Mubadala Development, said Tuesday that "the semiconductor industry is celebrated for overcoming seemingly insurmountable odds to continue the trend toward smaller, faster, and more energy-efficient products" and, in partnership with IBM, announced research that will enable the continued scaling of semiconductor components to the 22-nanometer level and beyond.

And Intel on Thursday will show off new research that will demonstrate the company's latest advancements with its chip manufacturing technology.

Intel has invested in ASM International, a major semiconductor production equipment maker that has been struggling during the worldwide economic downturn.

Intel Capital's investment amounts to 4 percent of ASMI's total common share capital, based on approximately 54 million common shares outstanding. This follows an investment last week by Tokyo Electron, which took a 4.9 percent stake in the company.

ASMI is currently trading at about $12.40 a share.

"ASMI has been at the forefront of delivering materials and equipment innovation to integrated circuit manufacturers worldwide," Intel said in a statement. ASMI's equipment and materials are used in wafer processing, assembly, and packaging of semiconductor devices.

IBM, Samsung Electronics, STMicroelectronics, and others are teaming up on the development of next-generation chip technology for small, low-power devices with one wary eye on Intel, which is expediting its move to chips with smaller geometries.

(Credit: IBM)

IBM and its semiconductor technology alliance partners are announcing the availability of 28-nanometer (nm) chip technology, a little more than a generation beyond the 45nm technologies currently used by Intel and Advanced Micro Devices in their latest chips.

The first products using chips based on this technology are expected in the second half of 2010, an IBM spokesman said. Devices will include smartphones and consumer electronics products.

The largest, single countervailing force to the IBM-led group is Intel. The Santa Clara, Calif.-based chip giant's chief executive, Paul Otellini, said Tuesday in a first-quarter earnings conference call that Intel is "pulling in" the release of "Westmere" chips based on 32nm technology and will ship silicon later this year.

Generally, the smaller the geometry, the faster and more power efficient the chip is.

The IBM alliance--which also includes the AMD manufacturing spin-off Globalfoundries, Chartered Semiconductor, and Infineon Technologies--are jointly developing the 28nm chipmaking process based on the partners' "high-k metal gate" (which minimizes current leakage), low-power complementary metal oxide semiconductor (CMOS) process technology.

The technology "can provide a 40 percent performance improvement and a more than 20 percent reduction in power, in a chip that is half the size, compared with 45nm technology," IBM said in a statement. "These improvements enable microchip designs with outstanding performance, smaller feature sizes and low standby power, contributing to faster processing speed and longer battery life in next-generation mobile Internet devices and other systems."

IBM said customers can begin their designs now using 32nm technology and then transition to 28nm for density and power advantages without the need for a major redesign.

One prominent customer is U.K.-based ARM, whose basic chip design has been used in billions of devices all over the world. ARM is collaborating with the IBM alliance to develop a design platform for 32nm and 28nm technology and is tuning its Cortex processor family and future processors to exploit the technology's capabilities, IBM said.

Moore's Law is maxing out. This is an oft-made prediction in the computer industry. The latest to chime in is an IBM fellow, according to a report.

Intel co-founder Gordon Moore predicted in 1965 that the number of transistors on a microprocessor would double approximately every two years--a prediction that has proved to be remarkably resilient. But IBM Fellow Carl Anderson, who researches server computer design at IBM, claims the end of the era of Moore's Law is nigh, according to a report in EE Times.

Exponential growth in every industry eventually has to come to an end, according Anderson, who cited railroads and speed increases in the aircraft industry, the report said.

"A generation or two of continued exponential growth will likely continue only for leading-edge chips such as multicore microprocessors, but more designers are finding that everyday applications do not require the latest physical designs," Anderson said in the EE Times' report. Anderson also cited the staggering costs of research and fabs (factories) as a formidable barrier for continued advancement. Few companies can afford chip plants that typically cost billions of dollars to build and maintain.

So, what does the future hold? Anderson cited three technologies: optical interconnects, 3D chips--which have circuits and components stacked on top of each other--and accelerator-based processing as seeing significant advancements, the report said. The latter technology, accelerators, is hot right now.

In addition to IBM, companies such as Nvidia and Advanced Micro Devices' ATI unit supply graphics-processor-based computers to accelerate scientific, engineering, and animation applications. Intel is also expected to bring out its Larrabee chip later this year or early next year that can be used as an accelerator.

In the wake of the latest kerfuffle between Advanced Micro Devices and Intel, AMD's chief counsel seized the moment to sound off on a primal fear at his company: Intel is bent on its destruction. Intel, of course, doesn't quite see it that way.

After Intel accused AMD on Monday of breaching a 2001 patent cross-license agreement with Intel, AMD's top lawyer had some choice words for its bigger rival.

In a phone interview Tuesday, AMD general counsel Harry Wolin refuted Intel's claim that the AMD manufacturing spin-off Globalfoundries is not a subsidiary--and thus cannot legally use Intel intellectual property--and talked more broadly about Intel's tactics.

Intel's ultimate goal, Wolin believes, is to crush rivals into oblivion. "In their perfect world, we wouldn't exist. If they had to deal with the government every now and then, that's fine, and they're still extracting monopoly profits from the industry," he said.

Wolin doesn't buy into the oft-repeated theory that Intel needs AMD to keep the industry honest and to keep the U.S. government at bay. "I don't agree with the premise that they have to have us and they think they have to have us. I think they would absolutely like us dead," Wolin said.

The Dickensian depiction of AMD as the impoverished, distressed victim of Intel's bullying and manipulation is inaccurate and, more importantly, misses the relevant point, according to Intel spokesman Chuck Mulloy. "It's nice of them to try to speak for us. AMD has been a competitor for almost 40 years in one form or another. This is not about AMD going away," he said. "This is about our rights and AMD's rights under the patent cross-license agreement."

Ashok Kumar, an analyst at investment bank Collins Stewart, said the premise of a remorselessly predatory Intel set on killing off its rivals is attention-getting but not that realistic.

"Could Intel put them out of business? Probably. But is it a likely outcome? I don't think so," he said. "Because they'll get a lot of significant push back from the OEMs (PC makers). The OEMs will essentially be making a beeline to Washington, D.C."

Intel contends this is a very localized dispute about whether Globalfoundries is a subsidiary or not, and not a manufactured issue "to distract the world from the global antitrust scrutiny (Intel) faces," as AMD said in a statement Monday. "AMD cannot unilaterally extend Intel's licensing rights to a third party without Intel's consent," said Bruce Sewell, senior vice president and general counsel for Intel, in a statement on Monday. Intel maintains the issue is that Globalfoundries is 34.2 percent owned by AMD and 65.8 percent-plus owned by Advanced Technology Investment Co., an investment company. So, in effect, Globalfoundries is not an AMD subsidiary.

Wrong, AMD says. It is not about ownership. AMD has met the conditions that qualify it as a subsidiary. "It requires that AMD originally contributed at least 50 percent of the assets. If you look at the fact that we've thrown in the German factories, we've thrown in the people, we've thrown in the technology, we've thrown in the intellectual property. I don't think there's any credible argument that says we haven't thrown in more than 50 percent of this. It says nothing about owning. It says you have to originally contribute 50 percent of the assets," Wolin said.

And what happens from here?

"Let's say the parties end up in a lawsuit at the end of 60 days," Wolin said. (Intel says it will terminate AMD's rights and licenses under the cross license in 60 days if the alleged breach has not been corrected.) "Well, you know, that lawsuit doesn't come to court for years and wouldn't come to court until well after the antitrust suit would come to court, which is currently scheduled for February of next year," according to Wolin.

Intel says the next step is mediation, where Globalfoundries is brought to the table. If this doesn't resolve the issue, then they would both be off to the races and the lawsuits would begin.