Transcript Future Computers

Future Computers
CSCI 107, Spring 2010
When Moore’s law runs out of room
• When transistors become only tens of atoms thick
– Quantum mechanics applies
– Defects are harder to control
– Heat is extreme
• “Dual-core” chips avoid these issues
What’s next?
• Alternative architectures and
nanomaterials
• Perfecting new ways to process
information
– E.g., quantum computing and biological
computing
New Architectures-Memristor
• Smallest transistors are 32 nanometers
wide—about 96 silicon atoms across
• crossbar approach has parallel
nanowires in one plane crossing over a
set of wires at right angles
• A 1 molecule thick buffer layer is
between them
• The intersections between the two sets
of wires act like switches, called
memristors
• They represent 1s and 0s as transistors
do, but also store more information.
• 1 memristor can do the work of 10 or 15
transistors.
Multiple Cores
• When clock cycles reached 3 to 4
GHz chips reached the heat ceiling
• For greater performance, designers
placed two processors on 1 chip
• Personal computers now have
quadruple cores
– Intel i7
– AMD Phenom X4
• Need to create languages and tools
for software developers of
consumer applications
– Microsoft’s F# programming language
– More needed
Faster Transistors
• researchers hope to make graphene
transistors
– 10 nm across and one atom high
– Faster than field-effect transistors.
– Lose very little energy from scattering or colliding
with atoms in the lattice, so less heat is generated
Different Computing Schemes
• Current Efforts
– Optical
– Biological
– Quantum
• Criteria for being a
computer
– Represent information
– Operate on that data
• Turing machine
Optical Computing
• Representing information
– photons carry information, not electrons, and
they do so at the speed of light
• Computation
– Controlling light is much more difficult
– Current work: optical switches and optical
interconnect between traditional processors
DNA Computing
• Representing data and instructions
– DNA molecules
– Theses molecules store the “programming” that
directs the lives of our cells
DNA Computing
• Computing Tools
– Watson-Crick pairing
• every strand of DNA has its Watson-Crick complement
– Polymerases
• copy information from one molecule into another
– Ligases
• binds molecules together
– Nucleases
• cut nucleic acids
– Gel electrophoresis
• A solution of heterogeneous DNA molecules is placed in one
end of a slab of gel, and a current is applied
– DNA synthesis
• write a DNA sequence on a piece of paper, send it to a
commercial synthesis
• Massively parallel, energy efficient, clean
Quantum Computing
• Representing Data
– The energy state of
a hydrogen atom
• An atom in its
ground state, with its
electron in its lowest
possible energy level
can represent a 0
• The atom in an
excited state, with its
electron at a higher
energy level can
represent a 1
Representing Information
• Quantum computers aren't limited to two
states
• Quantum bits, or qubits, can exist in
superposition
– when checked, the qubit will read 1 half of the
time and 0 half of the time
• Quantum Physics
Quantum Computing
• Qubits can be set and read using lasers to pulse
energy
• Operations:
– AND, NOT, COPY
• Big Problem: How to isolate atoms:
– Ion traps use optical and/or magnetic fields
– Optical traps use light waves to trap and control
particles.
– Quantum dots are made of semiconductor material
and are used to contain and manipulate electrons.
Quantum Parallelism
• Quantum entanglement
– if you apply a force to 2 atoms in superposition, they can become
entangled
– In entanglement the original information no longer resides in a
single quantum bit but is stored instead in the correlations
between qubits
– Measuring one bit, thereby putting it in a definite state, causes
the other bit to also enter a definite state
• “Quantum Parallelism”---massively parallel, nondeterministic computing
– Put all the input bits in equal superposition of 0 and 1---an equal
superposition of all possible inputs.
– Run this input through a logic circuit that carries out a particular
computation.
– The result is a superposition of all the possible outputs of that
computation.
Benefits
• Clean, fast, and can
solve a new class of
problems