Intel sees a bright future for silicon photonics, moving information at the speed of light in data centers and beyond
This article is part of the Technology Insight series, made possible with funding from Intel.
It is increasingly difficult to improve the performance of electrical connections between compute, storage, and networking resources, so server interconnects quickly move to optical I / O to handle growing bandwidth requirements. pass through the data center. Intel, for example, recently revealed that more than 5 million of its 100-gigabit optical transceivers already provide rack-to-rack connectivity.
The company remains focused on integrating low cost, low power consumption optical I / O technologies into servers and processor packages. Success will require massive scale-up, from millions of devices discussed during Labs Day 2020 to billions of dollars – making faster connections between disaggregated computing, storage and networking.
Intel’s vision is based on silicon photonics, which combines integrated circuits and semiconductor lasers. Compared to traditional optical systems comprising hundreds of discrete parts, silicon photonics brings together all of this complexity on a single silicon chip. This results in easier assembly, lower power consumption, smaller form factors, and lower cost.
- Intel is currently launching several new silicon photonic transceivers, including 200 Gb / s FR4 and 400 Gb / s DR4, with product sampling at 800 Gb / s.
- Ethernet switches with integrated photonics are required to meet the bandwidth demands of network switches. Expect a 51.2 Tbps solution with 100 Gbps electrical channels in 2023.
- Beyond the data center, high-volume silicon photonics will open up new markets, such as next-generation LiDAR for autonomous vehicles.
But before light becomes the basis of connectivity inside servers or accelerators, the silicon photonics industry needs to pass a number of milestones. Faster transceivers are just around the corner. The optics “co-packaged” with Ethernet switches have already been demonstrated, with retail availability expected in 2023. By around 2025, Intel says its autonomous driving subsidiary, Mobileye, will use photonic integrated circuits to power its next-generation light sensing and telemetry (LiDAR) suite of sensor technologies.
Make photonics a billion dollar business
Intel traces its silicon photonics research back to 2004, when scientists created the first transistor-type device capable of encoding data on a light beam. In 2011, the company announced a 50 Gb / s silicon photonic link created by multiplexing four hybrid silicon lasers. The technology went into prime time in 2016 with the launch of a 100G PSM4 optical transceiver. The company claims that the overall revenue of the 100G product line has exceeded $ 1 billion.
The demand for even more speed is exploding. According to the Equinix Global Interconnection Index, interconnection bandwidth is expected to increase at a CAGR of 45% between 2019 and 2023, totaling 16,300 Tbps worldwide. As data rates increase, optical links become more prevalent, even over short distances.
“When you walk into a data center today, you’ll see 100 Gbps copper cables running from the servers to the top of the rack switch,” says Robert Blum, senior director of new business and marketing at Intel. “These cables are good for about four meters. But everything beyond the rack already uses optics. As we increase the data rates to 200 or 400 gigabits, the range of copper becomes much shorter and we’re starting to see this trend where optics goes all the way to the server. “
Intel currently uses 200G FR4 (single mode fiber with a range of 2 km) and 400G DR4 (single mode fiber with a range of 500 m) plug-in transceivers. The company recently started sampling 800 Gbps hardware, which uses eight lasers.
Beyond transceivers: Ethernet switches with co-packaged optics coming soon
Plug-in optics aren’t the end of Intel’s game, however.
“It was never just about these transceivers,” says Intel’s Blum. “This was the learning curve because we knew that later on the optics had to be co-packaged with Ethernet switches. And it will eventually also be integrated into the CPU or XPU. “
Intel took its first steps towards this bolder vision in March 2020, when it introduced a 12.8 Tbps Barefoot Tofino 2 switch co-packaged with built-in 1.6 Tbps photon engines passing traffic. 400 Gb / s Ethernet compliant with DR4 standards.
Bringing photonics to the case, right next to the ASIC switch, provides several benefits, including energy savings. Instead of inefficient traces of copper running from the switch bundle to the large pluggable interfaces (often with re-timers in between), fiber goes from photonics modules to a faceplate connector. These smaller connectors also help increase density. And wafer-scale manufacturing means reliable on-chip photonic engines can be integrated at a lower cost.
As switching throughput increases from 12.8 Tbps to 25.6 and 51.2 Tbps over the next two generations, the number of transceivers per switch and the data rates of these transmitters -receptors will also increase. Density scaling will promote smaller connectors, lower wattage, and efficient cooling.
To get a feel for what this will look like, Intel says its own 51.2 Tb / s solution should be ready for commercial deployment at the end of 2023 using 100 Gb / s lanes to expose up to with 64 plug-in interfaces for 800 Gb / s transceivers.
Silicon photonics paves the way for next-generation LiDAR
The strengths of silicon photonics extend well beyond the data center. Think about autonomous vehicles (VA). Sensors under development for Level 4 (fully automated) and Level 5 (optional steering wheel) VAs model the surrounding environment using a radar cocoon and forward-facing lasers. Unfortunately, existing LiDAR systems are susceptible to interference and limited in their ability to measure the speed of other objects.
The next generation LiDAR overcomes these problems by investigating the differences between the emitted light and the reflected signal. The increased complexity makes it prohibitive to produce this type of coherent LiDAR technology with discrete optics. However, the construction of lasers and optical amplifiers in a photonic integrated circuit reduces costs, while reliability and performance improve.
At this year’s CES, Mobileye announced an on-chip LiDAR system that will combine active and passive elements to create 184 vertical lines for scanning, moved through optics. Intel plans to manufacture the SoC at the same manufacturing facility in New Mexico receiving a $ 3.5 billion investment for advanced packaging technologies. Comprised of over 6,000 individual components, this SoC is sophisticated equipment. And yet, the application of silicon photonics is expected to cut costs for Mobileye when it starts shipping in 2025.
“What’s unique about our platform is that we are able to integrate these lasers, or gain, at the wafer level,” says Blum of Intel. “No one else has this in production. And this is a great example where we have a huge value proposition by switching to silicon photonics. “
This is just the beginning
Mature, high-volume silicon photonics opens the door to additional opportunities. Blum continues, “We can use integrated optics in the biomedical space, personal health, sensing – any application where you can use lasers to detect something becomes more attractive. Some of the things that might have been prohibitively expensive or too complex for discrete optics suddenly become possible. “
There is a strong value proposition in doing certain types of calculations in the optical field. Imagine a future with optical chips alongside processors used for certain machine learning workloads. This is only possible with silicon photonics on a scalable manufacturing platform.
Technological advancements announced at Intel Labs Day 2020 foreshadow silicon photonics inside servers, although optical interconnects are not yet on the way to product implementation. But what we’ve seen in the company’s roadmaps – faster transceivers, co-packaged optics with next-gen Ethernet and LiDAR switches for autonomous vehicles – demonstrates that the future of moving data with light is really brilliant.