Avatar

It’s not every day you get to work on products that continue to evolve so quickly that before you finish talking about one device the next generation comes out. However, with Cisco Silicon One™ that’s exactly what we’ve done.

We recently announced our 19.2 Tbps P100 piece of routing silicon and I wanted to spend some time going over what it means from a system perspective.

Cisco Silicon One P100—In a class of its own

There are several types of routing silicon available on the market today from third-party silicon providers to full system vendors. This silicon covers a variety of market needs but broadly speaking routing silicon breaks down into silicon that is built for core, peering, and aggregation markets versus those focusing on broadband aggregation. If we take a look at all the silicon currently available on the market, it’s clear that Cisco Silicon One is in a class of its own, and still today there’s no routing silicon that achieves higher bandwidth, even than our original Q100 announced back in 2019. Since then, we rolled out the Q200 with higher bandwidth and lower power consumption than the Q100. Today, less than 24 months from our initial Q100 launch and just 12 months from our Q200 announcement, we released the P100, further increasing our lead over the competition. We believe the Cisco Silicon One P100 is 2.6 times higher ethernet bandwidth than other routing silicon in the market.

The Cisco Silicon One P100 builds upon the great foundation that we created with Cisco Silicon One but continues to add features, buffering, scale, and programmability enhancements while growing the bandwidth.

Cisco Silicon One
Figure 1. Routing silicon available in 2021

Because Cisco Silicon One devices can flexibly assign ports to be generic ethernet or a fully scheduled fabric, the Cisco Silicon One architecture enables optimized fixed boxes and modular systems.

Figure 2. Cisco Silicon One – Common architecture across form factors

This capability is unique in the industry, and it allows us to offer products from a fixed box measuring just one rack-unit with a single piece of silicon, to a massive modular chassis with many pieces of silicon, to a fully disaggregated chassis with even more pieces of silicon. This is all accomplished with a common architecture, SDK, and P4 forwarding code.

This means that when we come out with a new piece of routing silicon, we enable our customers to enjoy benefits across their portfolio and network.

If I try to oversimplify the impacts of the Cisco Silicon One P100 in the market, we see that:

  • P100 enables high bandwidth fixed boxes with 6–12 times less silicon. This leads to massive power, space, and cost savings while simultaneously reducing latency.
  • P100 enables modular systems that are more than 2 times higher bandwidth than anything else available on the market, with the industry’s first 36×800, 28.8 Tbps modular line cards, while significantly improving power efficiencies of modular systems.
Figure 3. Cisco Silicon One P100 vs. competition

Fixed box advantages

Because we can use a single piece of silicon to create a 19.2 Tbps, 24x800G, or 48x400G router we can drive a level of efficiency unheard of in the industry today. The next best piece of routing silicon on the market tops out at 7.2 Tbps.

This means that you need to use between six and 12 devices to build an equivalent system. This difference means Cisco Silicon One enables a smaller, lower latency, more power-efficient router at the system level.

Figure 4. Cisco Silicon One P100 fixed box advantage

Modular line card advantages

Modular systems are built with multiple pieces of silicon. Based on limitations on the faceplate, Printed Circuit Board (PCB), optics form factors, silicon, and orthogonal connector densities, most system vendors produce 36x400G line cards with customer-facing 14.4 Tbps of bandwidth.

Using our advanced architecture, coupled with 100G PAM4 Serializer/Deserializer (SerDes) technology, Cisco Silicon One P100 enables line cards to push through previous limitations in front panel optics, PCB technologies, and orthogonal connectors, resulting in line card hardware that’s two times higher bandwidth than other modular line cards. When combined with our 25.6 Tbps fabric element, the Cisco Silicon One G100 is 2.6 times higher bandwidth than other fabric elements, so customers can build modular routers that are up to three times higher bandwidth.

Table 1. Cisco Silicon One P100 modular system bandwidth

The bandwidth increase coupled with the power efficiency of Cisco Silicon One drives significant power efficiency gains compared to all other routing silicon in the market.

Figure 5. Cisco Silicon One P100 modular line card advantage

Building the future of the internet

Cisco Silicon One P100 enables the highest bandwidth and most efficient fixed box routers based on a single piece of silicon. It also enables the highest bandwidth modular line cards that are two times higher bandwidth than what can be built with our competitors’ silicon.

Cisco continues to validate the claims we made in December 2019 by proving the benefits of a fully converged architecture built from the ground up for the future of the internet.

For more information, read my full analysis on

the benefits of Cisco Silicon One P100 (pdf)

 



Authors

Rakesh Chopra

Cisco SVP & Fellow

Common Hardware Group Architecture and Platforming