A High-Speed and High-Capacity Single-Chip Copper Crossbar

Slides:



Advertisements
Similar presentations
COEN 180 SRAM. High-speed Low capacity Expensive Large chip area. Continuous power use to maintain storage Technology used for making MM caches.
Advertisements

BEOL Al & Cu.
2007 MURI Review The Effect of Voltage Fluctuations on the Single Event Transient Response of Deep Submicron Digital Circuits Matthew J. Gadlage 1,2, Ronald.
Wafer Level Packaging: A Foundry Perspective
Minimizing Clock Skew in FPGAs
Noise Model for Multiple Segmented Coupled RC Interconnects Andrew B. Kahng, Sudhakar Muddu †, Niranjan A. Pol ‡ and Devendra Vidhani* UCSD CSE and ECE.
Die-Hard SRAM Design Using Per-Column Timing Tracking
04/09/02EECS 3121 Lecture 25: Interconnect Modeling EECS 312 Reading: 8.3 (text), 4.3.2, (2 nd edition)
Low-Power CMOS SRAM By: Tony Lugo Nhan Tran Adviser: Dr. David Parent.
Institute of Digital and Computer Systems 1 Fabio Garzia / Finding Peak Performance in a Process23/06/2015 Chapter 5 Finding Peak Performance in a Process.
THEORETICAL LIMITS FOR SIGNAL REFLECTIONS DUE TO INDUCTANCE FOR ON-CHIP INTERCONNECTIONS F. Huret, E. Paleczny, P. Kennis F. Huret, E. Paleczny, P. Kennis.
Noise and Delay Uncertainty Studies for Coupled RC Interconnects Andrew B. Kahng, Sudhakar Muddu † and Devendra Vidhani ‡ UCLA Computer Science Department,
Through Silicon Vias EECS713 Daniel Herr.
RF MEMS devices Prof. Dr. Wajiha Shah. OUTLINE  Use of RF MEMS devices in wireless and satellite communication system. 1. MEMS variable capacitor (tuning.
2013 DAC Designer/User Track Presentation Inductor Design for Global Resonant Clock Distribution in a 28-nm CMOS Processor Visvesh Sathe 3, Padelis Papadopoulos.
CSET 4650 Field Programmable Logic Devices
Toshiba Standard Cell Architecture for High Frequency Operation Peter Hsu, Ph.D. Chief Architect Microprocessor Development Toshiba America Electronics.
General Licensing Class Oscillators & Components Your organization and dates here.
Device Physics – Transistor Integrated Circuit
EGRE 427 Advanced Digital Design Figures from Application-Specific Integrated Circuits, Michael John Sebastian Smith, Addison Wesley, 1997 Chapter 4 Programmable.
1 Monolithic Pixel Sensor in SOI Technology - First Test Results H. Niemiec, M. Koziel, T. Klatka, W. Kucewicz, S. Kuta, W. Machowski, M. Sapor University.
1 EE 587 SoC Design & Test Partha Pande School of EECS Washington State University
CMP 4202: VLSI System Design Lecturer: Geofrey Bakkabulindi
Taking evolutionary circuit design from experimentation to implementation: some useful techniques and a silicon demonstration Adrian Stoica Ricardo S.
1 Interconnect/Via. 2 Delay of Devices and Interconnect.
1 Carnegie Mellon University Center for Silicon System Implementation An Architectural Exploration of Via Patterned Gate Arrays Chetan Patel, Anthony Cozzie,
A High-Speed & High-Capacity Single-Chip Copper Crossbar John Damiano, Bruce Duewer, Alan Glaser, Toby Schaffer,John Wilson, and Paul Franzon North Carolina.
Bi-CMOS Prakash B.
An Improved “Soft” eFPGA Design and Implementation Strategy
Chapter 4: Secs ; Chapter 5: pp
A High-Speed & High-Capacity Single-Chip Copper Crossbar John Damiano, Bruce Duewer, Alan Glaser, Toby Schaffer, John Wilson, and Paul Franzon North Carolina.
EE 587 SoC Design & Test Partha Pande School of EECS Washington State University
A High-Speed & High-Capacity Single-Chip Copper Crossbar John Damiano, Bruce Duewer, Alan Glaser, Toby Schaffer,John Wilson, and Paul Franzon North Carolina.
CARBON NANOTUBES (A SOLUTION FOR IC INTERCONNECT) By G. Abhilash 10H61D5720.
High Gain Transimpedance Amplifier with Current Mirror Load By: Mohamed Atef Electrical Engineering Department Assiut University Assiut, Egypt.
Modelling LED Lamps with Thermal Phenomena Taken into Account Krzysztof Górecki and Przemysław Ptak Gdynia Maritime University Department of Marine Electronics.
Piero Belforte, HDT 1999: PRESTO POWER by Alessandro Arnulfo.
High Speed Properties of Digital Gates, Copyright F. Canavero, R. Fantino Licensed to HDT - High Design Technology
Introduction to ASICs ASIC - Application Specific Integrated Circuit
Circuit characterization and Performance Estimation
THE CMOS INVERTER.
The Interconnect Delay Bottleneck.
OVER VOLTAGE OR UNDER VOLTAGE
Different Types of Transistors and Their Functions
LOAD CUTOFF SWITCH UPON OVER VOLTAGE OR UNDER VOLTAGE
Topics SRAM-based FPGA fabrics: Xilinx. Altera..
Digital readout architecture for Velopix
PRESENTED BY SAI KRISHNA.R (2-1) NRIIT TEJASWI.K(2-1)
Low-Power SRAM Using 0.6 um Technology
20-NM CMOS DESIGN.
Written by Whitney J. Wadlow
EE 597G/CSE 578A Final Project
Architecture & Organization 1
Duckhwan Kim and Saibal Mukhopadhyay
VLSI Design MOSFET Scaling and CMOS Latch Up
Reading: Hambley Ch. 7; Rabaey et al. Sec. 5.2
ELEN 468 Advanced Logic Design
3. Advanced Rules & Models
Architecture & Organization 1
Adrian Stoica Ricardo S. Zebulum Xin Guo* Didier Keymeulen
Device Physics – Transistor Integrated Circuit
Programmable Electrically Erasable Logic Devices (PEEL)
Topics Antifuse-based FPGA fabrics: Flash-based FPGAs Actel.
This is the new logo for the XC4000X family
Topics Circuit design for FPGAs: Logic elements. Interconnect.
MCP Electronics Time resolution, costs
Implementation Technology
Device Physics – Transistor Integrated Circuit
EE382M VLSI 1 LAB 1 DEMO FALL 2018.
SAS-3 12G Connector Drive Power Pin Configuration
Presentation transcript:

A High-Speed and High-Capacity Single-Chip Copper Crossbar John Damiano, Bruce Duewer, Alan Glaser, Toby Schaffer, John Wilson, and Paul Franzon Copper Challenge Team #38 North Carolina State University Raleigh, NC Outline Circuit Design and Simulation Advantages of Copper Interconnect Electrical and Physical Characterization

The Copper Crossbar Function Why a crossbar? Crosspoint switch with programmable, non-blocking connections between sets of input and output lines Why a crossbar? The inherently long interconnects can best demonstrate the benefits of advanced interconnect technology

Cell Design Programmed through input lines I/O connection set by writing “1” to latch Latch outputs along output column combined OR-tree Cell Schematic Cell Layout Reset & Pre-Configures provide fast erase & write Cell size (W x L) = 5.68 mm x 19.50 mm (two cells shown at right)

Interconnect Strategies Input lines on M5, Output OR-tree on M3 M4/M6 used as GND planes Global Strategy: Maintain R and Decrease Linewidth to Reduce C Capacitance Values for Crossbar Interconnect Reduced RC load drops => Higher Performance Reliability of Cu not an issue

Simulation Metric #1 - Data Rate Output for 2.0GHz square wave input Metric #1 - Data Rate maximum input signal frequency for the crossbar Reduced RC load using Copper interconnect enables higher data rate vs. Aluminum Copper: 5.3 Gb/s Aluminum: 4.0 Gb/s Aluminum Copper

Simulation Metric #2 - Latency Output for 2.0GHz square wave input Metric #2 - Latency Delay of signal from crossbar input to output Faster edge rate with Copper interconnect enables lower latency vs. Aluminum Copper: 370 ps Aluminum: 425 ps Aluminum Copper

Advantages of Copper Interconnect Performance Copper Interconnect enables 30% higher Data Rate and 15% lower Latency vs. Aluminum Cell Size Tighter metal pitch with same resistivity available with Copper Interconnect Aluminum cell with equivalent performance would be 64% larger due to wider lines, increased pitch, and/or larger drivers Significant for arrayed / SOC applications

Electrical Results Input NOT passed to Output for all I/O configurations Input Output VDD/VSS Diode characteristics NOT observed VDD-VDD & VSS -VSS VDD -VSS

Failure Analysis How could this happen? Die stripped to substrate using HF to investigate VDD-VSS opens Diffusion pattern, created by P20 reticle, discovered to be absent! Only Diffusion pattern visible on die consists of Fill Shapes around original diffusion data Detail of pad cell How could this happen? Detail of crossbar array

Generating Fill Shapes B’ B (A-B’)+B A-B’

Silicon vs. Layout Data pad cell layout pad cell silicon array layout array silicon

Failure Analysis Crossbar pad cells compared to NCSU Team #16 - the diffusion pattern was dropped only for our die Explains unusual electrical data - no active devices present Only solution - new P20 (diffusion) reticle must be generated Good News! UMC has agreed to re-order P20 reticle and start new Copper Challenge lot Team #38 pad cell Team #16 pad cell

Conclusions Copper Crossbar circuit developed to exploit the advantages of copper interconnect technology Crossbar design using copper interconnect achieved a higher data rate and reduced latency, with a smaller cell size vs. equivalent aluminum circuit Puzzling Electrical Results from fabricated chips led to discovery of missing diffusion pattern UMC re-run of Copper Challenge designs promises to yield functional die with advanced interconnect & high performance