Five Considerations for Comparing Optical Power Consumption

Recent articles and TV documentaries have highlighted the Internet’s “dirty secret” – it currently accounts for around 10% of the world’s energy consumption, a number that could grow to 20% by 2030, equal to the current energy consumption of the entire United States.

Streaming Netflix for one hour consumes enough electricity to drive a Tesla Model S more than 30 km, power an LED lightbulb constantly for a month, or boil a kettle once a day for nearly three months.

The 5 billion Youtube streams of Despacito in April 2018 burned as much energy as 40,000 U.S. homes use in a year. Bitcoin annual energy consumption is equivalent to Chile’s, with a single transaction consuming the same amount of energy a British household consumes in two months.

5G will only make matters worse, with power consumption forecast to be three and a half times that of 4G. The bottom line is, power consumption and the CO² emissions that result are a big concern for the industry and the planet.

Figure 1: Coherent power consumption vs. time

Optical equipment vendors have been trying to do their bit by reducing the power consumption of optical transport equipment, and in particular of coherent transceivers. More efficient DSPs based on the latest CMOS process nodes, photonic integration, advanced packaging, and holistic design have all contributed greatly.

Coherent transponders went from just under 2 watts per Gb/s circa 2013 to less than 0.2 W per Gb/s today, with coherent pluggables based on the 400ZR MSA expected to deliver close to 0.04 W per Gb/s based on 400 Gb/s and 16 W.  However, comparing power consumption metrics between different technologies, products, and vendors can be challenging. Here are five things to consider for an “apples to apples” comparison:

1. Scope: DSP vs. Optical Engine vs. Module vs. System

Figure 2: Power consumption scope

When some vendors provide impressive power consumption figures, they may be referring only to the DSP itself. Other vendors may define the power consumption for the complete optical engine, which includes the DSP, photonics, and analog electronics. The DSP typically accounts for around 70% of the power consumption in a high-end optical engine, though this can drop to around 50% in a coherent pluggable.

Other vendors may provide a power consumption figure for the transponder/muxponder module that includes the optical engine and other components. Yet another definition might be for a fully loaded system that includes the transponder/muxponder modules, controllers, fans, and power supplies, with the total system power consumption divided by the total capacity. Even here an apples to apples comparison should consider whether redundant controllers, fans, and power are included or not.

2. Typical Power vs. Maximum Power

Like other types of equipment, optical hardware has typical and maximum power consumption. Individual copies of the same component will have a degree of variability in terms of their power consumption. Other factors that can influence power consumption include temperature, processing load, traffic load, and configuration. For example, the power consumption of fans can vary dramatically depending on how hard they have to work to keep the temperature within the required operating range.

Maximum power consumption is a calculated figure based on the worst case for each component. Typical power is usually based on measurements with multiple samples. Some vendors might quote typical power consumption, while others might quote maximum power consumption.

Both need to be considered: the former when comparing likely energy costs and the latter when dimensioning the power supply.

3. Client Pluggables: Included or Not?

Figure 3: Client pluggables

Another factor to consider when comparing the power consumption of modules and systems is whether client pluggables are included or not. Some vendors will quote the power consumption with no pluggables on the basis that the same types of pluggables will have the same impact on power consumption for all products.

There can also be variation in the power consumption of client pluggables depending on the type used. For example, OSFP and QSFP56-DD 400G pluggables vary from 7 W to 15 W, while 100G QSFP28 pluggables can have a maximum power consumption of 3.5 W (SR4) or 4.5 W (LR4/ER4).

However, some vendors will quote a maximum power consumption figure that include pluggables, with the maximum power consumption for that pluggable form factor, for example 5 W for QSFP28 and 20 W for QSFP56-DD.

4. Actual Bandwidth Requirements

Watts per Gb/s can provide an important tool for comparing the power consumption of devices with equivalent reach or that are all capable of meeting the reach requirement. This is normally calculated by taking the typical power consumption for the fully loaded device then dividing by the maximum capacity.

However, this can be somewhat misleading if your actual required bandwidth is well below the maximum of one or more of the devices under evaluation. In this scenario, a better strategy is to determine the required capacity and use this as the basis for a power consumption comparison.

5. Distance: Watts per Gb/s per Km

While watts per Gb/s is a useful tool for comparing power consumption, a vital dimension is missing – distance. For example, a 400ZR pluggable with power consumption in the range of 16 to 20 watts and giving ~0.04-0.05 W per Gb/s might look much more power-efficient than an embedded optical engine with power consumption of less than 0.2 W per Gb/s.

However, when we factor in distance, the picture changes significantly. In the best case, a 16W 400ZR pluggable at a 120 km reach delivers 333 µW/G/km, while in the worst case, a 20W 400ZR at 80 km would be 625 µW/G/km. By contrast, an optical engine with less than 0.2 W per Gb/s delivering 800G wavelengths at 950 km has a power consumption per kilometer of around 200 µW/G/km.

Plus, the much longer reaches enabled with lower wavelength speeds will deliver even more impressive results for the embedded optical engine.

So to conclude, when we the state power consumption of ICE6 in the newly announced Groove platforms as less than 0.2 W per Gb/s, we are using typical power consumption for the complete Groove platform fully loaded with CHM6 modules running with a line rate of 800G per wavelength. In addition, we would recommend that you use µW/G/km for the actual capacity and reach you require to get the most relevant comparison metric.