Importance of the Technology
Optical methods of communication (network interconnection between voice and data communications devices) have fueled more than two decades of expansion in telecommunications, driving down the cost of communication bandwidth through a succession of technological advances and improvements in manufacturing costs. As the cost of optical-interconnection componentslasers, detectors, driver circuitry, and optical fiberhas decreased and end-user demand for bandwidth has increased, optical interconnections have replaced conventional copper interconnects at shorter and shorter distances. As optical interconnects edge closer to the computer itself, far more processing could take place in the optical domain. Systems will not use the all-optical logic or computation devices that generated hype in the 1970s and 1980s but will adopt optical data processing, in which optics help transfer and switch data between functional (electronic) processing and memory blocks for computing applications.
The increasing demand for bandwidth and dropping cost per function that are inherent in Moore's law have driven the computing and semiconductor industries for almost two decades. But the bottleneck to computing functionality lies less in raw processor speeds than in the ability to transmit information around the motherboard and to the world outside. For some time, processor speed has doubled every 18 months in line with Moore's law, but main-memory transfer speeds have been increasing only 10% to 20% per year in the same period.
From a macro view of the world of interconnections, the solution may appear obvious: Optics should play the same role in computing that it has in communications. Because optical interconnects have successfully replaced electrical interconnects in the wide area over hundreds of kilometers and down to storage-area networks that connect distances of 100 meters or less, many people assume that they will inevitably make their way into the computer to help process data. For example, optics might transfer and switch data between electronic processors and the hierarchy of memory modules (the two main solid-state memory caches L1 and L2 and the main memory) to which they interconnect, or they might even transfer high-speed clock signals around the chip as intrachip optical interconnects. The beginnings of such a revolution are in sight, with parallel optical-interconnect companies claiming that optical backplane interconnects will soon be commercially viable for distances of meters. But the real challenge and opportunities lie lower down in the system. At stake are the future of processor design and future demand for components, equipment, and services that enable the interconnection of devices in high-speed electronic equipment. These elements have substantial implications for integrated-circuit makers and communications companies. But optics may not be able to scale down all the way to the chip level: Opportunities are certain to exist in optical data processing, but their scale and nature are still uncertain. |
|
| Optical Data Processing Contents |
 |
 |
 |
| |
Importance of the Technology |
| |
Recent Developments |
| |
|
VCSEL Developments |
| |
|
Improved Light Emission from Silicon |
| |
|
High-Speed Silicon Photodetectors |
| |
|
Broadband Multiwavelength Lasers |
| |
|
Fiber-to-the-Processor Architectures from Primarion |
| |
|
Focuts from Peregrine Semiconductor |
| |
The Technology in Brief |
| |
|
VCSELs |
| |
|
Photonic Band-Gap Materials |
| |
|
Hybrid Integration |
| |
|
Monolithic Integration |
| |
|
Silicon-Based Light Sources |
| |
|
Fast Optical Switching |
| |
|
DWDM, Switching, and Tunable Lasers |
| |
|
Board- and Chip-Level Waveguides |
| |
Commercial Development Parameters |
| |
Areas to Monitor |
| |
|
Latency |
| |
|
Packaging |
| |
|
Silicon-Based Light Sources |
| |
|
Processor/Wiring Architecture |
| |
|
Manufacturing Costs |
| |
|
Bandwidth Growth |
| |
|
Low-Power Devices |
| |
|
Photonics-on-Chip Integration |
| |
|
Competing Interconnects |
| |
|
Optical Switching |
| |
|
Electrical Limits |
| |
Implications of Commercialization |
| |
|
Expansion beyond Niche Applications |
| |
|
Monolithic Optoelectronic ICs: Silicon as King |
| |
|
Growing Demand for III-V VCSELs and Smart Pixels |
| |
|
The Critical Role of Optoelectronic IC Assembly |
| |
|
Collision Course in Consumer Computing Devices |
| |
Applications |
| |
|
Shelf-to-Shelf Optical Backplanes |
| |
|
Board-to-Board Free-Space ODP |
| |
|
MCM-to-MCM ODP |
| |
|
Chip-to-Chip ODP |
| |
|
On-Chip Optical Clock Distribution |
| |
Players |
| |
Updates |
|
|