Your browser is not supported. Please update it.

7 December 2020

Nokia and Telefónica find 5G 90% more energy efficient than 4G

There is no great surprise in the finding by Nokia and Telefónica that 5G networks are up to 90% more energy efficient than legacy 4G networks per unit of traffic. That is not enough to offset the rise in traffic anticipated by even the most conservative forecasts, which is why both parties admit that more must be done to cut energy consumption over cellular networks.

The research, conducted over three months, evaluated power consumption of the RAN for 4G and 5G traffic in Telefónica’s own network. In total, 11 predefined traffic load scenarios were analyzed, measuring energy consumed per-Mbps on the basis of how the load was distributed.

The study utilized Nokia’s AirScale portfolio, including base stations and massive MIMO adaptive antenna systems. It combined on-site base station energy consumption readings under these different traffic loads, with remote monitoring of actual power consumption via the network management systems.

The announcement was accompanied by some virtue signaling about meeting commitments to climate change and limiting global warming to 1.5 Celsius. This is somewhat fatuous, especially as a mere doubling in efficiency would fail to offset prospective increase in data traffic by a large margin. Typical of projections is Ericsson’s in its latest Mobility Report, which estimates that global mobile-only data traffic will rise over fourfold from 51 exabytes (EB) per month in 2020 to 226EB in 2026, with 5G networks carrying over half of that.

Still there is no doubt that 5G energy efficiency will improve a lot more. For the first time energy saving has been taken into account at the design stage of 5G, which has the effect of giving more scope for further improvement with experience and tuning, on top of giving a boost from the start.

Such built-in measures operate at various levels, from network sharing down to processor optimization, as well as more advanced transmitter technologies. The latter includes MIMO, which improves energy efficiency by increasing the throughput of each antenna, since multiple beams can reuse the cell’s frequencies. But configuration in the field is complex, so with experience there will be optimization that improves energy efficiency progressively.

Similarly, over product lifecycles processor optimization will improve in line with architectural evolution, with virtualization of the RAN and more deployment of fronthaul.

There is also scope for cutting energy consumption by exploiting sleep modes, involving selective turning off components at periods of inactivity or very low traffic. 4G was limited in scope here because of its chatty radio interface, transmitting reference signals constantly at high rates.

That has been addressed in 5G and there is scope for saving very high proportions of energy in base stations when traffic is light. This can be seen by considering that power consumption at the base station has two components, a fixed one required to keep power supply and air conditioning systems going, and a variable part that increases with traffic as processors have to work harder.

By putting the base station into sleep mode, the fixed power consumption can be drastically reduced, and this accounts for up to 90% or even a bit more of total power at low load of around 10%, which can be quite a lot of the time for some cells. This is a promising approach then, but still the subject of research rather than implementation, the challenge being to determine which components to turn down and when, without imposing undue overhead or wake-up latency.

It is hard then at this stage to predict accurately how much better energy efficiency can become, but signs are quite encouraging. We have seen some research from Orange Labs that while a 5G antenna consumes three times more energy on average today than a 4G antenna, while delivering about five times as much bandwidth, that penalty should drop to 50% by 2021 and 25% by 2022. If that were the only factor then, 5G would end up consuming almost five times less energy per bit by 2022. That will be associated with equal improvements elsewhere, but to achieve even parity with 4G for overall energy consumption, 5G will probably have to cut energy consumption per bit at least 10 times eventually.