The 50Gbps Silicon Photonics Link

Fifty years ago, Ted Maiman built the first laser out of a ruby crystal rod, never expecting that this invention would revolutionize industries from medicine to communications. Likewise, it’s hard to imagine that Robert Noyce and Jack Kilby could have foreseen how their invention of the silicon Integrated Circuit (IC) – just one year earlier – would change the world.

Now in 2010, these two inventions are coming together with silicon photonics. We’ve been talking for many years about research to “siliconize” photonics, and until now all these breakthroughs have been at the device or component level. What we’ve announced today is for the first time we have an integrated silicon photonics transmitter using hybrid silicon lasers that’s capable of sending data at 50 gigabits per second (Gbps) across an optical fiber to an integrated silicon photonics receiver chip which converts the optical data back into electrical.

And what’s exciting is that this 50Gbps link is just the beginning. What we’ve demonstrated is the ability to integrate optical devices together, and just as we do in our microprocessor business (where we integrate more and more transistors together on a single chip), we will do so with silicon photonics. What does this give us? New functionality, new form factors, higher data rates, lower power and better performance than you could get with discrete photonic devices. Or to say it differently, this will allow us to apply optical technology to a whole bunch of new applications that could not have been possible previously. We now have the ability with integrated silicon photonics to bring low cost optical communications in and around the PC and server.

Why is this important? Let’s start with high end computing. Within servers, as demand for higher data rates increases, today’s copper interconnects require close proximity of processors, memory and IO. This limits, for instance, the amount of memory in the system to the number of modules (DIMMs) that can be mounted close to the CPU and or that can fit in the box. It also impacts the abililty to cool the system with many devices assembled so close together. Silicon Photonics will provide the ability to bring high speed optical communications into the platform, effectively relieving these distance constraints and providing the flexibility that could revolutionize system design. Servers could expand memory with remote, optically attached memory systems without sacrificing speed. I/O’s could run faster and with lower power if they were optimized for optical. Datacenters could maximize the performance of multi-core processors for large databases or virtualized environments with silicon-based optical links.

For clients, just imagine what 50Gbps could do. One could transfer an HD movie in less than 1 second. In that same second one could transfer 1000 high-res photos. Let’s look into the not too distant future of video displays. 3D TV’s are already hitting the market, doubling bandwidth requirements. Resolutions are increasing as well. Today 1080p seems sufficient on most TV’s – but imagine a screen that fills your entire wall, something you often see in science fiction stories. Imagine using that for a video call where the person you see literally appears to be in the room with you. To keep the same quality and resolution you will need Ultra High Definition (UHD, 4320p) – 16x the pixels of today’s HDTV. UHD at a 60Hz refresh rate requires 60Gbps, far beyond what we can do with copper cables, but easily within reach with integrated silicon photonics.

Today we have demonstrated 4 optical channels modulated at a rate of 12.5 Gbps each. These are combined and directed into one optical fiber for a total of 50Gbps. What’s next? Well, we can continue to scale up in line rate – we’ve already demonstrated modulators at 20 and 40 Gbps. We can also integrate more lasers or channels per chip. For example instead of 4 channels we can go to 8 or 16 channels. And when you add those two together, we can start talking about delivering data rates of 100G, 200G, 500G, and ultimately more than 1 Terabit per second (Tbps) – all out of a single integrated photonic chip smaller than your finger nail.

Link.jpg

Today’s announcement represents a huge step forward in our research, but there is more to be done as we look to take this technology from research to commercialization. On the silicon front we will continue to optimize the performance of the integrated photonic devices. On the transmitter we will push to drive modulators to operate at lower operating voltages to minimize power consumption. We will also continue to improve the efficiency of hybrid silicon laser, again to minimize power consumption. As with everything we do, we must keep low cost, high volume assembly in mind. As such we are working on things like inline optical & electrical testing, more automated packaging and assembly techniques, and improved passive fiber connectorization approaches.

In summary, this milestone brings together two of the most important inventions of the past century: silicon manufacturing and the laser. Looking forward into the coming century, I expect that the world will evolve into one of wireless and optical: wireless when mobility is the driver and optical for any connection that demands very high data rates. Every manufacturer of a connected device in the future is going to have to start thinking about how high bandwidth, low-cost optical communications will benefit the products and services they deliver. This is the start of bringing optical communications to anyone, anywhere, connecting anything around the world using silicon based integrated lasers. To me, that’s exciting.

To find out more, listen to the Future Lab: Moving at the Speed of Light podcast

6 Responses to The 50Gbps Silicon Photonics Link

  1. Matthew says:

    This is an amazing innovation.
    It’d be interesting to see how this would be implemented. The idea of being able to mount more RAM onto a chip alone would entice people to adopt the technology at the consumer level.
    I imagine that it will also encourage streamlining of communication between devices in computers, especially laptops. I’m curious about the energy savings in devices like this, since it would probably add a couple hours to the life of a battery.
    I also like the idea of having a TV the size of a wall in the house. :P

  2. Tony says:

    I’m interested in this type of high speed link for phased-array radar application within the Transmit/Receive channel. I lead a group of 20 engineers that design these X to W band rf freq. I would be interested in integrating some form of this into an R&D program -may be mutually beneficial.

  3. Michael Cloran says:

    this is excellent just imagine having a multibus with full bus memory connection where the memory could be seen as a cache to a processor and beyond the cache is a directory of the cache and beyond that is shared memory. This could lead to parallel processing where each processor is connected via 1 data bus+ 1 address/bus with multiplexed multimode fibre for busses. Although note that photonic chips may be larger than the electronic counterpart so a large multicored supercomputer may take some space ??
    Regards
    Michael

  4. Amir Karim says:

    Light is the only solution to solve the problems of future communication. However, if I say in simple words, we believe light is much faster than electrons, therefore it would lead to faster data transfer. However at this stage with opto-electronic ICs we are still bound to use electrons. It is a great solution for near future, but sounds like a bottle neck in the very long future. It would be perfect if both interconnects and data processing is based on only photons. In any case, way to go Intel!

  5. Ben Israel says:

    This is indeed great research. What I would like more information on is how do we get electrical data onto the photonic chip at 40Gbs and how do we inexpensively connect to fiber? Are these photonic devices so large that they will not be on intel processor chips in the near future?