SESSION TP6

SALON 10

TD: DEEP SUB-MICRON AND DIGITAL DIRECTIONS

Chair: D. Draper, Advanced Micro Devices, Sunnyvale, CA
Associate Chair: T. Baba, Matsushita Semiconductor of America, Palo Alto, CA

6.1 - Beyond Superscalar RISC: What Next? - 1:30 PM

D. Luick

IBM, Rochester, MN

Microprocessor microarchitectures and streamlined instruction sets are required as fundamental physical and logical limits emerge. Improved performance is expected from microprocessors exploiting high in-line programming with advanced compilers but maximizing operating frequency while maintaining short pipeline depth in simple regular designs.

6.2 - A Sub-0.1um Circuit Design with Substrate-Over-Biasing - 2:00 PM

Y. Oowaki, M. Noguchi, S. Takagi, D. Takashima, M. Ono, Y. Matsunaga, K. Sunouchi, Hl Kawaguchiya, M. Kamoshida, T. Fuse, S. Watanabe, A. Toriumi, S. Manabe, A. Hojo

Toshiba Corp., Kawasaki, Japan

Substrate-over-biasing together with gate-substrate-tie circuitry controls standby cycle and active cycle transistor leakage. Substrate-over-biasing applied to a test chip using 80nm gate-length technology results in 40ps gate delay at 0.5V supply.

6.3 - Statistical Circuit Characterization for Deep-Submicron CMOS Designs - 2:30 PM

J. Chen, M. Orshansky, C. Hu, C-P. Wan1, P. Bendix1

University of California at Berkeley, CA

1 Logic Corp., Milpitas, CA

Scaling of CMOS gate lengths to deep-submicron increases device sensitivities to manufacturing variations. Data shows two distinct patterns for this variation that are insufficiently modeled by present methods. A method overcomes this limitation and allows accurate design validation.

BREAK 3:00

6.4 - Multi-Chip Module with Optical Interconnection for Parallel Processor System - 3:15 PM

M. Koyanagi, T. Matsumoto, Y. Kuwana, H. Kurino

Tohoku University, Sendai, Japan

An optical-interconnect parallel-processor system for Monte Carlo simulation is implemented using bare die on MCM. Optical-interconnect performance improvement is evaluated using a test system.

6.5 - A 1M Synapse Self-Learning Digital Neural Network Chip - 3:45 PM

O. Saito, K. Aihara1, O. Fujita2, K. Uchimura

NTT Integrated Information & Energy Systems Labs., Kanagawa

1NTT Network Service Systems Labs., Tokyo

2NTT Electronics Co., Kanagawa, Japan

A 0.25um CMOS gate array contains 64 processing units and 8-port external RAM access control. The chip executes 10G connections per second for forward calculations and 1G connection updates per second for learning calculations and can handle a 1M-synapse network with external SRAM.

6.6 - Bulk Spin Quantum Computation: Towards Large-Scale Quantum Computation - 4:15 PM

I. Chuang, L. Vandersypen1, J. Harris1

Los Alamos National Lab., NM

1 Stanford University, CA

Quantum circuits implement basic logic functions using NMR techniques. Use of micromachining technology to realize an NMR quantum co-processor integrated on a silicon chip can be expected to solve specific problems significantly faster than the fastest classical machines.

CONCLUSION 4:45 PM


Go back to Table of Contents