Up to x3 more energy required for 5G vs 4G
How much energy will 5G consume?
5G networks have larger antennas, larger bandwidths, and higher base station density. How can the industry expect x20 more energy efficiency by 2030?
5G is the next generation of mobile communication technology which started rolling out in 2019 and is planned to replace 4G over the coming decade. You may have heard about the idiotic conspiracy theories about 5G and burning cell towers (BBC, 2020), but prior to CVOID the hype focused on the main benefit of speed: download speeds are expected to be up to 10Gbit/s (Hoffman, 2019) compared to less than 1Gbit/s for 4G (ITU, 2008). I have measured 4G speeds of over 100MB/s in central Tokyo, but otherwise I expect that type of performance is rare! It will be more likely with 5G.
The improved 5G speeds are achieved through higher-frequency radio waves, but they have shorter ranges. 3 different types of antenna will support 3 different frequencies to allow operators to trade-off range with speed, with devices selecting the highest speed option in range. The fastest speeds are available from the highest frequency (mmWave). These will be deployed in dense urban areas but the lower range means more cells are required.
Although some networks have started advertising 5G connectivity, this tends to be the lower-speed frequency, few devices offer 5G support, and those that do tend to be awful. This will change by 2025 when it is suggested that 2.8 billion 5G subscriptions will exist, accounting for 30% of all mobile subscriptions. 4G will peak in 2022 at 5.1 billion subscriptions (Ericsson, 2020). We will know when 5G is ready when Apple include it in the iPhone.
All sounds good, until you consider energy consumption. In a study of Finland, which has some of the highest mobile data consumption in the world, mobile networks were reported as being responsible for 0.7% of annual electricity consumption, 0.6 TWh in 2017 (Pihkola et al, 2018). The study also showed that energy efficiency improved from 12.34 kWh/GB in 2010 to 0.30 kWH/GB in 2017. Although efficiency is improving, more data is being transmitted.
Indeed, this is what has happened with the rollout of all mobile technologies — from 2G through to 5G — resulting resulting in an overall increase in total energy consumption (Joshi, 2019; Han & Bian, 2020). An industry survey supports this view, suggesting that 94% of respondents expect 5G to increase energy costs (451 Research, 2019). This is a manifestation of Jevon’s Paradox.
The industry is aiming for 5G to be x20 more energy efficient than 4G by 2030 (Orange, 2020). There are several features in 5G that are hoped to reduce energy consumption, such as more efficient software (Ericsson, 2020), but the main focus is on new sleep mode functionality. 4G networks must send signals every 1ms whereas 5G can do so every 20ms. During idle periods (such as in the early morning), this can significantly reduce unnecessary transmissions (Frenger & Tano, 2019; Dahlman & Parkvall, 2018; Lähetkangas et al, 2014).
The German Federal Environment Ministry and the German Environment Agency released a joint report on the climate footprint of video streaming (Umwelt Bundesamt, 2020) which also examined 5G. The report is in German so it is difficult for me to analyse e.g. whether the carbon footprint included the manufacturing of devices, but the English press release shows HD video streaming over fibre connectivity generated 2g CO2/hr, 5g CO2/hr for 5G and 90g CO2/hr for 5G. Wired connectivity is always going to be the least energy intensive method of connectivity because it doesn’t involve radio transmission, but 5G performs very well in comparison.
However, the end-user device is excluded from the methodology. Historical studies have reported that the majority of energy is consumed by the end-user device and transmission network (Shehabi, Walker & Masanet, 2014). The impact on the device battery is being discussed as a reason why 5G may consume x3 more energy than 4G (Koziol, 2019).
Past studies show that base stations are responsible for 57% of power consumption (Han et al, 2011). It is assumed this means the radio transmission because compute requirements were historically smaller and fixed as a constant (Samarakoon, 2016), but compute overhead increases with larger traffic volumes over more base stations. Moving processing to the edge will also increase the compute requirements and subsequent energy allocation. An analysis of the compute and transmission overheads suggest this will shift the balance so that more than 50% of energy will be consumed by compute resources in 5G base stations (Ge et al, 2017).
Improving the energy consumption of 5G networks is going to be a combination of options, with tradeoffs around speed and range. Wu et al, 2017 proposes several options ranging from making beam forming antennas more efficient and centralising signal processing, through to statistical models that manage packet queuing based on the availability of renewable energy. Smarter orchestration of workloads could consider the energy (and bandwidth) cost of backhauling traffic to centralised data centres vs performing processing closer to (or at) the edge (Zhang et al, 2016; Yang et al, 2018). This is interesting when combined with serverless event driven architectures.
The challenge with 5G energy consumption is a function of the design: larger antennas, larger bandwidths, and higher base station density (Han & Bian, 2020). However, compared to when 4G was rolling out, energy consumption and the associated environmental/climate impacts now attract much more public interest and media attention. This will be a positive forcing function on the industry.
Originally published at https://davidmytton.blog on September 18, 2020.