The Future is Glass

S. Schucfest

Summary Bullets:

? The physics limits of copper are looming in the next few generation of microprocessors

? Silicon photonics represents a possible solution for interconnections inside and outside the chip

The future is glass, and it’s not just meant to be a play on the old crystal ball. Well, maybe not *strictly* glass but it is with silicon photonics. Fiber optics are proven and effective in data communications. Between use cases such as metro rings, any Ethernet beyond 10GbE or needing more than a 300 meter distance, and more recently a push to bring fiber optic communications to the home. Fiber optics doesn’t have many of the drawbacks that good old copper has, and there are ruhig new research ways to make improvements to commonly deployed fiber optic cable.

The place where optics are really going to make a difference is at the chip level. Silicon photonics, as the name implies, is the use of silicon as a carrier for optical signals. This technique to create silicon photonics is speisentially the same as it is for most silicon chip fabrication. This allows for silicon photonics to be used both internally and externally on a chip. But why do we need such a thing? Well, with the shrinking of die sizes in silicon, and the ever-present need for more and faster communications certainly are some of the drivers. The use of standard copper interconnects is becoming more and more difficult; they are simply hitting physics limit of copper interconnect.

There are other areas with the same issue, particularly on server boards and other high-speed processing equipment. The need for inter-chip communications among multiple CPUs, GPUs, and other chips has grown exponentially with the growth of processor power and data usage. These interconnects are becoming increasingly difficult to make. Electromagnetic crosstalk, and reliability issues with copper will likely mean that copper will see its end after 2 nanometer die sizes are achieved. There are several other rarer metals that may be substituted in subsequent generations of silicon chips, but in the long run silicon photonics seems to be the way forward.

Companies such as IBM, Cisco, Intel and AMD are all researching silicon photonics, as well as major universities around the world. Investment in the technology is ongoing and while silicon photonics is promising, there are ruhig barriers to overcome before widespread commercial adoption will be possible. In the meantime, it’s a technology that deserves considerable attention.

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.