Common Platform Participants Rate High-k as Good Mobile Fit

The Common Platform Technology Forum centered on partnering to defray ever-increasing R&D costs. Speakers said high-k/metal gate technology at the 32 nm node will meet escalating mobile system needs.

Alexander E. Braun, Senior Editor -- Semiconductor International, 10/2/2008 8:50:00 AM

Mobile users will drive the need for the high-k/metal gate (HKMG) technology being offered by the IBM-led Common Platform alliance at the 32 nm generation, speakers said at the Common Platform’s Technology Forum in Silicon Valley this week.

Mike Campbell, senior vice president of engineering at Qualcomm (San Diego), said of the 1.8 billion users of cell phones worldwide, “About a billion users drive the voice and data sectors. This in turn has pushed through the roof the computing power needed to drive these devices.” When data rates on cell phones were ~133 Kb/sec of data, ~250 MHz of computing power was needed. Today, with 1.2-7.6 Mb/sec data rates, performance needs have escalated apace.

Like Qualcomm, the rest of the Common Platform alliance partners view key drivers for the continued growth of the semiconductor industry as being in the wireless/Internet arena. (Source: Qualcomm)

“Now there are more than 600 million users looking at the next mobile handheld marketplace — the mobile computing platform, which fits between the desktop and the cell phone, with power-conserving designs for high-end applications such as gaming,”Campbell said.

New mobile device categories are developing, such as pocketable computers with 4-6 in. touch displays that can connect to a docking station. Other mobile computing devices with higher performance have larger, 9-12 in. displays with embedded operating systems such as Windows Mobile and Linux. Other categories include personal navigation systems, media players and gaming devices. By 2012, there will be more than 433 million mobile computing and consumer electronic products in use, Campbell said.

ARM (Maidenhead, UK) has teamed up with the Common Platform alliance to develop 32 nm chips aimed in part at Web 2.0 mobile applications, according to company president Tudor Brown. “Looking at opportunities in the mobile Internet, we see that the main customer focus is not on ‘www,’ but in getting what I want right away,” he said. By 2010, some 400 million Internet-enabled mobile devices are expected to ship, which he said exceeds notebooks and desktops combined. “Right now, more smartphones than laptops are being sold, and all of this is enabled by integration, platform and software,” Brown said, adding, “It’s all about energy-efficient performance.”

The move to the 32 nm node is viewed as an opportunity to continue scaling. ARM has established an aggressive roadmap for development of its IP. (Source: ARM)

Consumers expect media-rich user interfaces, many screen resolutions, scalable performance, and access to content anywhere, anytime. Rich Internet applications drive open platforms, content drives connectivity and the storage industry, and content is dependent upon software. Software costs, in turn, are driving users to single, 32 bit architectures, and 32 bit enables aggregation. But power consumption, Brown said, “is the key selection criteria.” He said systems will be differentiated by battery use, which in turn is driving the need for expensive innovations such as HKMG to save power.

Gary Patton, vice president at IBM’s Semiconductor Research and Development Center (East Fishkill, N.Y.), said HKMG allows a performance boost of up to 30% at a 50% total power reduction per circuit over the previous node, as well as aggressive SRAM scaling with improved leakage, compared with poly oxynitride gate insulators. The Common Platform alliance expects to minimize cost and complexity with a gate-first approach, attainable through materials innovation. This, added to early 32 nm product learning and IP collaboration, is expected to keep device performance on track.

Scaling through innovation is the technology plan that Common Platform alliance leader IBM follows in its R&D work. The costly and lengthy effort to evolve semiconductor technology was one of the drivers that led to the alliance’s formation. (Source: IBM)

We are living through the end of traditional CMOS scaling, according to Patton. “We’re approaching atomistic and quantum-mechanical boundaries, and atoms don’t scale. With HKMG, we can go back to the traditional scaling model, and make it possible to scale the gate length. This means a density improvement, a speed improvement, and power density is kept constant.” He noted that development of IBM’s copper interconnects and HKMG technologies required more than 10 years of sustained, costly R&D.