It's taken Intel just three years to shrink its system-on-chip (SoC) process node from 32nm to 14nm (with a pitstop at 22nm along the way), and by 2014, it will be shipping 14nm chips in volume. Can anyone compete? Globalfoundries intends to give it an honest effort by aggressively accelerating its roadmap with a 14nm process of its own. The company's competing 14nm-XM (eXtreme Mobility) process will leverage 3D FinFET transistor technology for higher performance and better power efficiency in mobile devices.
Intel generated a lot of press with the unveiling of their 3D, low-power Tri-Gate transistor technology. Now it's IBM’s turn to hop into the 3D waters. Today, the company announced that it’s entered into a joint partnership with 3M to develop 3D semiconductors. They’re going about things a little bit differently than Intel, though; rather than developing chips with raised elements, IBM and 3M want to create “bricks” out of up to 100 separate silicon chips in a process known as “3D packaging.”
It can be argued that British chip designer ARM is perhaps one of the most self-effacing tech companies out there. Quite a rarity in a world where big tech firms are always on the lookout for opportunities to draw attention to themselves and their products. But with its customers already selling 4 billion chips a year, the Cambridge-based outfit can not possibly evade the limelight.
ARM's latest attempt at modesty is, well, very modest. Although it currently doesn't – and may never – measure up to Intel in terms of vital business statistics, but the fact is its chip designs are virtually everywhere, and they already pose a huge threat to Intel's hegemony by the virtue of their ubiquity alone.
ARM may be comfortably placed in the mobile chip market, but the company is unwilling to rest on its laurels. In recent times, ARM has time and again underlined its interest in the server market. The company hopes to make a dent in the low-power server market with a new chip design that features both virtualization and large physical address support. The next generation of its Cortex-A processor, the Eagle, will be the first to utilize the two key instruction-set extensions that the UK-based chip designer announced at the Hot Chips conference today.
"It's the natural progression of the ARM architecture to move into this domain," said David Brash, ARM's architecture program manager. "We think that that are going to be places for low-power servers, but also new cases." A slide Brash presented at the conference revealed that some of the leading names in the field of server virtualization have already begun developing hypervisor software for the chip design, which is “very close” to being released.
IBM last week said it would begin collaborating with industry leaders and universities scattered throughout the European Union to improve several facets of modern chip design, including the productivity and reliability of semiconductor and electronic systems.
"Designing a microelectronic chip is very expensive and the design costs are the greatest threat to continuation of the semiconductor industry's phenomenal growth," noted Dr. Jaan Raik, senior researcher at Tallinna Tehnikaulikool and coordinator of the DIAMOND project. "The increasing gap between the complexity of new systems and the productivity of system design methods can only be mitigated by developing new and more competent design methods and tools."
The goal of the new integrated approach is to localize and stomp out bugs on all abstraction levels. IBM points out that about 70 percent of today's design efforts are placed on verification and debugging, while soft errors -- like transient errors caused by cosmic radiation -- ranks as a rapidly growing threat.
The DIAMOND consortium will use a holistic approach to develop new tools and methods to help track all of these errors.