I spent seven months instrumenting a Ryzen 9 5950X build server with a smart plug to track actual power consumption during package builds. Here are the numbers.
The setup
The build server runs FreeBSD and handles about 5,000 package builds across various ports. I wanted hard data, not estimates, so I used a WiFi-enabled smart plug that logs power draw at regular intervals and uploaded the data to a local InfluxDB instance.
The server itself pulls 55W at idle with nothing running.
What the data showed
Over the seven-month period, the server consumed 98.4 kWh total. At my electricity rate, that worked out to about 32 AUD in power costs.
Chromium was the worst offender by far. A single Chromium build takes about 4 hours and consumes 0.77 kWh. If you're building packages regularly and Chromium is in your queue, that adds up.
Why this matters
Most people running build infrastructure don't think about power costs because they seem negligible. But when you're looking at scaling decisions - whether to add another build node, whether to upgrade to faster hardware, whether to run builds during off-peak hours - having actual numbers helps.
The measurement itself was straightforward. Any smart plug that can log power data will work. The key is capturing it over a long enough period to see real patterns rather than outliers.
The practical takeaway
If you're running build infrastructure, instrument it. The numbers are usually smaller than you'd expect, but they're concrete. When you're making capacity decisions or optimizing build times, power consumption is one more data point that's easy to collect if you're already measuring.
Top comments (0)