Transcript Optical Interconnects

Optical Interconnects
Speeding Up
Computing
PICTURE
HERE
Matt Webb
In the near future
All logic
operations
solved using
with optics
 Have to start
somewhere

Picture of something
to signify amazing
computing speeds
Why do we care?

Problems with scaling of electrical
interconnects (EI)

Ex: Telecommunications already
moved away from electrical lines
Performance between chips is
already affected by EI
 Near future will be a problem on
chips

Solutions

Several possibilities

New architectures
• Minimize interconnections

New design approaches
• Emphasize interconnection layout

New medium for interconnection
• Optics
Background
Research on optical interconnections
has been going on for >20 years
 With optoelectronic digital
computing
 Development of SEED’s and VCSEL’s
 Practical to implement

SEED/VCSEL picture?
Aspect ratio limitations
Bandwidth limitations
Delay limitations
Typical model
Laser diode driver
 Transimpedance Amplifier

Latency vs. Length

3Gb/s, 0.25μm CMOS
Power consumption vs.
Length
Channels vs. Length
Scaling of OI
transmitter/receiver

Only viable if
technology for
the design of
TX/RX can
keep up with
future
generations of
silicon
technology
Timing for optical signals
Virtually independent of temperature
 Virtually no degradation of signal on
the scale of meters
 Slower propagation, but very
reliable and predictable
 Could eliminate high power clock
circuits

Benefits of OI
Optoelectronic devices can be used
as impedance transformers
 No inductance on an optical line
 Do not generate or detect radiofrequency signals or interference
 Long or short does not matter

More benefits
Larger synchronous zones, even on
multiple chips
 Allows “fire-hose” architectures
 Lower power dissipation after
“break-even length”

• 100μm – 10’s of cm
Benefits cont.
Voltage isolation
 Increasingly important due to
smaller power supplies
 Larger density for long distance onchip and off-chip interconnects
 No need for hierarchy of
interconnects

Main Challenges for OI
Young and expensive
 Systems that could take advantage
of optics will most likely have
different architectures than today's
 Problems and benefits are
misperceived by those not involved
in recent research

Receivers
Power dissipation
 Small capacitance
 Larger noise immunity
 “Receiverless”

Transmitters
Quantum-well modulators
 VCSEL’s
 LED’s
 Si based optoelectronic devices

Technology
Absence of low-cost and practical
 How to integrate


III-V devices
Compatibility
 Hybrid integration


Solder-bonding
Misperceptions

Wavelength is too large


Not true for longer interconnect lengths
Conversion of optics to electronics is
inefficient
Power, area, and time
 Current generation of technology

Conclusions
Use of OI could solve many of the
problems faced in today’s electrical
systems
 Much work remains to make the
technolgy feasible
 When it’s all said and done OI’s are
the future
