Controlling the amount of electricity being used in computing use has become increasingly more important over the years. Wither it is in the home lab, and the consumption directly affect one's own pocket book or the datacenter/server room has heating or power limitations.
In this case I thought it might be interesting to look at what impact of networking speed has on electrical use. One could conclude from simply looking at number of and size of the heatsink on the cards that there must be a difference. For this test my trusty Kill-A-Watt was pulled out of storage. A normal desktop was used, a Dell 3rd Gen i3, the specifics are irrelevant. Each time the machine was booted off of USB thumb drive into "Parted Magic" (a Linux Rescue Utility); I chose that OS over using Windows so other variables such as drive indexing or Windows Updates wouldn't skew the results. The machine was left to idle for 5 minutes to let things settle down before measurements where taken. Extra GBICs where removed, and the same DAC cable was used. The machine was plugged into a 10gb switch; no 25gb switch was available for testing, donations are welcome! :)
Bare machine: (no additional hardware) = 32watts
Machine w/ 10g Intel i520 NIC no cable =35watts
Machine w/ 10g Intel i520 NIC w/ cable =37watts
Machine w/ 25g Mellanox CX4 NIC no cable =40watts
Machine w/ 25g Mellanox CX4 NIC w/ cable =42watts
Machine w/ 25g Mellanox CX5 NIC w/ cable =41watts
Machine w/ 25g Broadcomm BCM51414 w/ cable =41watts
Machine w/ 25g Chelsio w/ cable =49watt (very hot to the touch!)
