Intel, IBM Prepare Dueling Technical Papers on 32-nm Microprocessor Development

Intel, IBM and Taiwan Semiconductor Manufacturing are planning to publish a series of new technical papers at a December conference that will detail each of their efforts to manufacture 32-nanometer microprocessors.

All three of these companies are scheduled to present their papers and analysis at the IEEE’s International Electron Devices conference in San Francisco, which starts Dec. 15. A number of abstracts for these and other technical papers are now available at the IEEE Web site.

In addition to these papers, IBM and its alliance of chip development partners, which include Advanced Micro Devices, Freescale Semiconductor, STMicroelectronics, Toshiba, and the College of Nanoscale Science and Engineering of the University of Albany in New York, will present on paper that will describe their efforts to produce a 22-nm SRAM (static RAM) memory cell.

In the paper, IBM and its partners are calling the SRAM the smallest functioning SRAM cell ever made.

The push toward 32- and then 22-nm processors is considered an essential mission of the semiconductor industry as devices from cell phones to notebooks continue to shrink and customers and IT buyers demand better performance and more battery life from these mobile devices. On the high end, these chips will also allow companies to build new types of servers and high-performance computers that use less power while increasing the performance.

For its part, Intel has already announced that it will offer 32-nm microprocessors—code-named “Westmere”—in 2009. The paper at the IEEE conference, however, is one of the first times Intel engineers will detail their efforts to produce these chips.

When these processors do come to market, they will be based on Intel’s Nehalem microarchitecture. In November, Intel will begin offering the first chips based on Nehalem for high-end PCs and gaming systems. These first chips are built on the company’s current 45-nm manufacturing process.

The 32-nm Intel chip will use the company’s second-generation high-k metal gate technology, which reduces power leakage—the electricity wasted while the transistors are idle—and helps improve performance. Intel’s paper will also detail the company’s efforts to produce these chips using “193-nm immersion lithography” and how it developed low-k interconnect dielectrics or insulators within the chips.

The SRAM test chips that Intel engineers will detail will contain 2 billion transistors, a significant improvement compared with 45-nm chips, which contained at 800 million transistors when those processors came to market in November 2007.

At the same time that Intel is presenting its paper, IBM and TSMC will each present their methods for producing 32-nm chips. In the IBM paper, engineers will discuss how they developed low-power CMOS technology with a high-k metal gate to stop power leakage. CMOS (complementary metal oxide semiconductor) is an integrated circuit that is an important step in developing a fully functioning microprocessor.

In their paper, IBM engineers said these chips will eventually be used in high-data-rate applications such as notebooks.

Finally, TSMC, an independent foundry that manufacturers chips for other IT vendors, will offer a paper that looks at its 32-nm foundry technology. In the paper, TSMC engineers said they will use a high-k metal gate that uses the element hafnium as a dielectric. Intel also uses hafnium in its high-k metal gate manufacturing process.

TSMC will detail some of these technological features in a test SRAM chip.