A 100G transceiver draws about 3.5 W, a 400G about 12 W. That is 3.4 times more, so more energy-hungry. Well, not exactly.
I prefer to count watts per gigabit transported.
A 400G port is 4 times the capacity for 3 times the power. And not 4 times, as one might think.
All thanks to a technical feat: PAM4 modulation, which doubles the bit rate per channel without doubling the frequency or the power consumption.
The result: it becomes possible to move from 100G to 400G while improving energy efficiency per bit.
Many players talk about 400G. But behind these promises, there is often an aggregation of 4 × 100G links.
4 transceivers = 4 signal lanes = 4 times the management complexity and more power consumption than an equivalent native 400G port.
Same throughput, but more watts, more cables and more surface to cool. And cooling accounts for 30 to 40% of a data centre’s consumption.
Reducing the physical footprint is not a detail: it has an immediate knock-on effect on the bill.
So does 400G solve digital technology’s energy problem? Absolutely not: global IP traffic keeps growing. 400G does not reverse the trend; it makes it slower than it would otherwise have been.
Today, migrating to 400G is no longer just a capacity decision; it is an energy architecture decision.
A robust architecture is also one that minimises its exposure to fluctuations in electricity prices, and that is the very “essence” of what we do at France-IX.
Peering Plus
This news was first published on LinkedIn. To comment or share it, head to the original post. → See the post