Michael Conrad Award Lecture

Slides:



Advertisements
Similar presentations
1 Lecture 20 Sequential Circuits: Latches. 2 Overview °Circuits require memory to store intermediate data °Sequential circuits use a periodic signal to.
Advertisements

B.Macukow 1 Lecture 3 Neural Networks. B.Macukow 2 Principles to which the nervous system works.
Basic Electricity and Electronics Mr. McClean Concepts of Engineering and Technology Copyright © Texas Education Agency, All rights reserved.
Chapter 9 Memory Basics Henry Hexmoor1. 2 Memory Definitions  Memory ─ A collection of storage cells together with the necessary circuits to transfer.
Neural Networks. Background - Neural Networks can be : Biological - Biological models Artificial - Artificial models - Desire to produce artificial systems.
Ovonic Unified Memory.
Neurons, Neural Networks, and Learning 1. Human brain contains a massively interconnected net of (10 billion) neurons (cortical cells) Biological.
Chapter 14: Artificial Intelligence Invitation to Computer Science, C++ Version, Third Edition.
Your Interactive Guide to the Digital World Discovering Computers 2012 Edited by : Noor Alhareqi.
1 Machine Learning The Perceptron. 2 Heuristic Search Knowledge Based Systems (KBS) Genetic Algorithms (GAs)
Memory and Storage Dr. Rebhi S. Baraka
The Memristor.
A Presentation on “OUM “ “(OVONIC UNIFIED MEMORY)” Submitted by: Aakash Singh Chauhan (CS 05101)
Computer Architecture And Organization UNIT-II General System Architecture.
Chapter 3 Digital Logic Structures. 3-2 Combinational vs. Sequential Combinational Circuit always gives the same output for a given set of inputs  ex:
PHY 201 (Blum)1 Microcode Source: Digital Computer Electronics (Malvino and Brown)
CHAPTER-2 Fundamentals of Digital Logic. Digital Logic Digital electronic circuits are used to build computer hardware as well as other products (digital.
Neural Network Basics Anns are analytical systems that address problems whose solutions have not been explicitly formulated Structure in which multiple.
LBSC 690 Module 2 Architecture. Computer Explosion Last week examined explosive growth of computers. What has led to this growth? Reduction in cost. Reduction.
Introduction to Spintronics
OVONIC UNIFIED MEMORY Submitted by Submitted by Kirthi K Raman Kirthi K Raman 4PA06EC044 4PA06EC044 Under the guidance of Under the guidance of Prof. John.
1 KU College of Engineering Elec 204: Digital Systems Design Lecture 22 Memory Definitions Memory ─ A collection of storage cells together with the necessary.
Submitted To: Presented By : Dr R S Meena Shailendra Kumar Singh Mr Pankaj Shukla C.R. No : 07/126 Final B. Tech. (ECE) University College Of Engineering,
Neural Networks. Background - Neural Networks can be : Biological - Biological models Artificial - Artificial models - Desire to produce artificial systems.
Where are we? What’s left? HW 7 due on Wednesday Finish learning this week. Exam #4 next Monday Final Exam is a take-home handed out next Friday in class.
March 31, 2016Introduction to Artificial Intelligence Lecture 16: Neural Network Paradigms I 1 … let us move on to… Artificial Neural Networks.
MEMRISTOR A New Bond graph Element.
نظام المحاضرات الالكترونينظام المحاضرات الالكتروني Introduction :: Computer Organization and Architecture Computer.
Computer Architecture Adapted from CS10051 originally by Professor: Johnnie Baker Computer Science Department Kent State University von Neuman model.
Transistors Student Lecture by: Giangiacomo Groppi Joel Cassell
Computer Organization and Architecture Lecture 1 : Introduction
Memories.
REGISTER TRANSFER LANGUAGE (RTL)
Objectives Overview Differentiate among various styles of system units on desktop computers, notebook computers, and mobile devices Identify chips, adapter.
Artificial Intelligence (CS 370D)
The Central Processing Unit
Welcome Welcome Welcome Welcome Welcome Welcome Welcome Welcome
Modelling & Simulation of Semiconductor Devices
Introduction of microprocessor
Edited by : Noor Alhareqi
Welcome.
Dr. Unnikrishnan P.C. Professor, EEE
New Transformative Possibilities for Ovonic Devices
Machine Learning. Support Vector Machines A Support Vector Machine (SVM) can be imagined as a surface that creates a boundary between points of data.
Edited by : Noor Alhareqi
Satish Pradhan Dnyanasadhana college, Thane
The New Information Age
Conventional Silicon Computers Ovonic Cognitive Computer
COMPUTERS IN CRISIS Computers have not changed their basic approach since the beginning (von Neumann binary) Conventional computers are commodity items.
The OvonicTM Cognitive Computer
Edited by : Noor Alhareqi
BINARY STORAGE AND REGISTERS
Chapter 3 Digital Logic Structures
3.1 Introduction to CPU Central processing unit etched on silicon chip called microprocessor Contain tens of millions of tiny transistors Key components:
European Conference on Phase Change and Ovonic Science
Machine Architecture and Number Systems
Machine Learning. Support Vector Machines A Support Vector Machine (SVM) can be imagined as a surface that creates a boundary between points of data.
Machine Learning. Support Vector Machines A Support Vector Machine (SVM) can be imagined as a surface that creates a boundary between points of data.
Artificial Intelligence Lecture No. 28
Machine Architecture and Number Systems
Machine Architecture and Number Systems
The Ovonic Cognitive Computer A New Paradigm
Edited by : Noor Alhareqi
Notes from Last Class Office Hours: GL Accounts?
Lecture 14: State Tables, Diagrams, Latches, and Flip Flop
Energy Conversion Devices, Inc.
A review of recent phase change memory developments
APRIL 1964, A McGRAW-HILL PUBLICATION
Ovonic Cognitive Computer, LLC formed 9/26/2002
COMPUTER ORGANIZATION
Presentation transcript:

Michael Conrad Award Lecture Ovonic Phase Change Optical and Electrical Memory and Switching Devices That Permit Cognitive Computing Stanford R. Ovshinsky Ovshinsky Innovation LLC Michael Conrad Award Lecture Department of Computer Science Wayne State University, College of Engineering October 11, 2011

Energy and Information are the Twin Pillars of our Global Economy I was seeking to build an analog neuron and its synapses in the late 1940s. It was clear that crystalline materials were not applicable. I had to invent new materials in the 1950s so as to utilize new degrees of freedom for atomic engineering that provided basic new mechanisms – physical, chemical, structural and particularly electronic – in other words, new science and technology I created the field of active amorphous and disordered materials In energy for example, there are as high as 11 different elements in the NiMH batteries that enabled the electric and hybrid vehicle industry In information, I combined neurophysiology with materials to create devices and systems Cognitive Computing has been my driving force. OVSHINSKY INNOVATION LLC

John Bardeen – Inventor of the transistor and Nobel Laureate “Is it rectifying? No? Then it is very, very important.” We couldn’t wait for a time when he was free to visit, so in 1963 he sent Hellmut Fritzsche to validate my new inventions

…the “Ovshinsky effect… is the newest, the biggest, the most exciting discovery in solid-state physics at the moment,” said Sir Nevill Mott*, Director of the Cavendish Laboratory at Cambridge University in England. The discovery of the Ovshinsky effect was “quite unexpected,” said Professor Mott. Mott also said that “the principle of the transistor, discovered in 1947, could originally have been figured out on the basis of old knowledge, but that the Ovshinsky effect represented totally new knowledge.” New York Times, November 11, 1968 * Winner of The Nobel Prize in Physics 1977.

Structure of Nerve Cells and their Interconnections Diagram of a biological nerve cell showing synapses and synaptic contacts Simplified schematic of the communication of information between neurons Nerve Cell input output Nerve Cell axon synapse Afferent nerve terminal Node of Ranvier Myelin sheath Axon process Cell body Synapse Dendrite Synaptic contact S. Dehaene: “The new findings in numerical neuroscience compel us to accept that our mathematics, sometimes heralded as the pinnacle of human activity, is really made possible by conceptual foundations laid down long ago by evolution and rooted in our primate brain. We are clearly not the only species with a knack for numbers.” B. Katz: “Each nerve cell, in a way, is a nervous system in miniature.”

I-V Characteristics Ovonic Threshold Device Ovonic Memory Device Threshold voltage is proportional to chalcogenide thickness and when the thickness is < 200Å the device becomes wingless Switching in chalcogenide materials based on lone-pair excitation: Threshold --- noncrystallizing --- OTS Memory --- phase change --- OMS OVSHINSKY INNOVATION LLC

Causes of Conformation & Bonding Reorganization Lone Pair Orbitals…. Strength of Repulsions The strongest …[Lone Pair Lone Pair] Next…...…... [Lone Pair Bonding Pair] The weakest ….. [Bonding Pair Bonding Pair] Since lone pairs are not tied down into a bonding region by a second nucleus, they can contribute to moderately low energy electronic transitions… Therefore: Light and Electric Fields can couple to Lone Pairs OVSHINSKY INNOVATION LLC

Ovonic phase change devices have been in use for many years Optically in rewriteable DVDs Electrically they preceded Flash and are now considered to replace Flash, DRAM and SRAM etc.

Intel to sample phase change memory in 1H 2007 “Ed Doller, chief technology officer of the flash memory group at Intel Corp., told the meeting that Intel's 128-Mbit had demonstrated 100 million cycles endurance and much greater than 10 years data retention. ‘The phase-change memory gets pretty close to Nirvana,’ Doller said.” Intel to sample phase change memory in 1H 2007 By Peter Clarke (03/06/2007 12:44 PM EST) (Excerpt from the meeting for analysts & press held in California on March 6, 2007.)

A Stackable Cross Point Phase Change Memory In a recent paper, Intel introduced the Ovonic Threshold Switch with Ovonic phase change memory [1], showing the three dimensionality that can be achieved with those devices, as shown here. (from Intel)

Ovonic Phase Change Memory Device The figure shows the relationship between the programming current and the programmed resistance of a standard Ovonic semiconductor memory device. A 50 ns, 1.5 mA pulse will cause the resistance to be high. When 50 ns pulses of increasing amplitude are applied to the device, no change is observed until at about 0.8 mA when the device crystallizes, and low resistance is achieved. Currently devices are made with reset .2 and set .1 mA. amorphous crystalline OVSHINSKY INNOVATION LLC

Ovonic Electrical Multi-State Data Storage When the amplitude of the pulses is changed in small increments, intermediate resistance states can be achieved. This programming mode is direct overwrite. The 16 resistance levels shown in the top figure correspond to storage of four bits in a single device. The data shows repeatability over 1600 programming events. The test was stopped without failure The bottom graph shows binary cycling performance. The test was stopped without failure. Continued testing to 1014 would have taken another year. Device Resistance (W) Multiple-bit storage in each memory cell (Ten pulses per step, repeated ten times) Programming Voltage (V)

Operation of Ovonic Cognitive Device Nerve Cell input output axon synapse Synapses Neuron Synapses

Ovonic Cognitive Device Operation Device resistance remains high after each pre-threshold pulse. Once the threshold is exceeded, the resistance drops dramatically. Different pulse widths can be selected to change the number of pulses used to fire the device. The graph shows consistency over 1,000 “cycles.” The test was stopped without failure. OVSHINSKY INNOVATION LLC

Functional Model of the Ovonic Cognitive Device The Ovonic Cognitive Device can be either optical or electrical OVSHINSKY INNOVATION LLC

I-V Curves of OTS and OQCD Voltage Current The Ovonic Threshold Switch (schematic IV curve at right) has high conductivity when switched on, due to a solid state plasma not seen before. It switches off by plasma recombination. The current capacity is 3x108A/cm2, 30 times higher than CMOS transistors. The Ovonic Quantum Control Device expands the capabilities of Ovonic Threshold switches by adding control terminals and can replace transistors as well as adding new functionalities, and offering new degrees of freedom to the design of computer architecture. Application of a control voltage changes both the threshold voltage and the holding current, providing two forms of control and modulation in an Ovonic Threshold Switch.

The Ovonic Quantum Control Device Threshold Voltage is Modulated using the Control Terminal Input-to-Reference Terminal I(V) curves of the OQC device using differing voltages applied to the control terminal. Both the threshold voltage and the holding current are modulated.

The Ovonic Quantum Control Device Latching / Non-Latching Operation In the non-latching mode the device reverts back to the quiescent state, where the voltage goes from the holding level back up to the applied level. The applied voltage is lower in this case than the latching case to keep the current below the holding level. The latching behavior is demonstrated by the voltage Input-to-Reference staying at the lower holding level after the end of the third terminal pulse. The open symbols show the Control signal and the filled Symbols show Input Terminal signal Small energy signal switches large energy

The Ovonic Cognitive Computer Arrays of computation and storage elements are combined in a conventional computer which: Requires separate storage and processor units or regions Has limited parallel processing capability Is limited to Von Neumann operations An array of Ovonic elements Easily factors large numbers Has attributes of proposed quantum computers without their limitations, such as correlation of states at practical conditions and environments. Performs high level mathematical functions (e.g. vector and array processing) Has high 3-dimensional interconnectivity, huge density, giving rise to high speed, hyper-parallel processing (i.e. billions of interconnected processors) Interconnectivity is simply and inherently reconfigurable Has adaptive learning capability The Ovonic Devices are: Mass produced in exceptionally dense, all thin film, uniquely interconnected arrays Mass manufactured as a thin film, flexible device using proven technologies Ovonic Quantum Control Device - unique, ultra high speed, low cost 3 or more terminal device. Nanostructure capable of carrying large current density huge current density which is controlled by a new quantum mechanical effect invented by Stan Ovshinsky. Capable of order of magnitude more modulation than a transistor. Can be a single device circuit with multi-functions Can replace present day transistors creating a new post transistor electronic industry We proved that these devices can also be used as memristors

The Ovonic Cognitive Computer Conventional Silicon Computers Each Element: Computes based on single bit (binary) manipulation Manipulates data sequentially, bit by bit Ovonic Cognitive Computer Manipulates, processes and stores information in a non-volatile radiation hard and multilevel manner Hardware and software are unified Low voltage and low current operation Performs arithmetic operations (+,-,x,) on multi-bit numbers (0,1,2,3…n) Performs modular arithmetic Combines logic and memory in a single device Executes multi-valued logic Stores the result in a non-volatile manner Simple, powerful encryption – the device is absolutely forensically secure Acts as a neurosynaptic cell; i.e. possesses intelligence capability Scales down to angstrom dimensions; huge density Device speed so fast they have never been measured Capable of massive parallelism; also addressing different values

Ovonic Cognitive Device Quantum computing gained its reputation by factoring 15 at liquid helium temperatures and very short lifetimes. Quantum computing is many years away. Ovonic factoring is a trivial room temperature operation for us and is stable and reproducible and can do much more than factor 15.

Plasticity in Ovonic Chalcogenides Plasticity is the ability of a biological material or its non-biological analog to be able to adapt or change in response to incoming energy signals so as to enable learning, decision making and other characteristics of intelligence as well as storage of learned information. The resulting changes are structural in nature very much like electrical or optical signals can make for conformal and configurational alterations in the amorphous phase of the Ovonic Phase Change Devices. A single Ovonic device is capable of synaptic function receiving and weighting multiple inputs threshold activation accumulates input energy signals without responding until the total accumulated energy reaches a threshold level and then the device changes from a high to a low resistance state in a process that mimics the firing of a biological neuron. Individual Ovonic devices can be readily interconnected Highly dense two-dimensional arrays Three-dimensional, vertically integrated networks The threshold level of individual Ovonic devices can be controlled by various means The Ovonic Quantum Control Device The plasticity of the Ovonic neurosynaptic arrays opens up possibilities of unifying software and hardware. OVSHINSKY INNOVATION LLC

Origins – and the Future! The foundations of my work in multi-element atomic engineering of amorphous and disordered materials started with my investigations of brain function in the early 1950’s, moving on to inventing semiconducting devices that have similar and important kinds of functionality including plasticity. I described this in more than 25 papers and presentations over more than the last 40 years, including EPCOS, and 21 patents. I reasoned that analogs of biologic neurons could be an inherent component of Group VI semiconductors. Not only could threshold firing be realized, but also variously weighted synaptic interconnections. Accumulating Inputs Biomimetic Circuit Variable Resistance Devices Device Resistance (W) Taken from MRS Symp. Proc. 554, 339 (1999) Programming Voltage (V) OVSHINSKY INNOVATION LLC Jap. J Appl. Phys. 43, 4695 (2004)

Microsoft, Amid Dwindling Interest, Talks Up Computing as a Career Energy Conversion Devices, Inc. “If you invent a breakthrough in artificial intelligence, so machines can learn,” Mr. Gates responded, “that is worth 10 Microsofts.” “If you invent a breakthrough in artificial intelligence, so machines can learn,’ Mr. Gates responded, “that is worth 10 Microsofts.” End of Article Excerpt from Article from The New York Times – March 1, 2004