Closed geerlingguy closed 9 years ago
I'll be monitoring the power consumption over a few days' worth of testing, but this is much better than I originally hoped. Running Pis headless seems to save a TON of energy. Most other benchmarks must be run with a display attached, a keyboard and mouse, and maybe a WiFi dongle.
So this is extremely odd...
I was using a Kill A Watt for the original tests (and still had it running today when doing new tests), and I bought a Mini USB Charger Doctor to do inline load tests so I could see how the power is distributed to each Pi.
Today's measurement of one of the Pi 2s:
Pi State | Power Consumption |
---|---|
Idle | 420 mA (2.1W) |
ab -n 100 -c 10 (uncached) |
900-1200 mA (~4.5W) |
400% CPU load | 800-1100 mA (~4W) |
Interesting...
Measurements on the database Pi:
Pi State | Power Consumption |
---|---|
Powering on, 1x USB 64GB SSD | 900-1400 mA (~4.5W) |
Idle, 1x USB 64GB SSD | 960 mA (~4.8W) |
ab -n 100 -c 10 (uncached), 1x USB 64GB SSD |
1100 mA (~5.5W) |
400% CPU load, 1x USB 64GB SSD | 1250 mA (~6.25W) |
Continuing further research on the wiki: https://github.com/geerlingguy/raspberry-pi-dramble/wiki/Power-Consumption
Also of note: apparently the Kill A Watt is pretty inaccurate at lower wattages/amperages... see:
Apparently it does a lot better in its sweet spot, likely 50-1200W or so. The USB current doctor, or a decent multimeter, is probably the best option here... though not as simple as the Kill A Watt.
This will be helpful information when it comes time to work on a potential inline UPS.
Some numbers:
These readings were taken with the Pis in the following configuration:
stress -c 4
to max out all 4 CPU cores.