Rationale
A few days ago I was building rack 19″ server based on AMD Athlon 5350 CPU. The main goal was to create a hardware set that will operate on wide supply voltage ranges (48V, 24V, 19V) and will be power efficient.
I decided to take a look at quite cheap AMD AM1 Athlon APU with TDP said to be 25W.
Testing platform
The hardware:
- AMD Athlon 5350 CPU (4 cores, 2.05Ghz, 2MB L2 Cache)
- MSI AM1I Motherboard
- Crucial Ballistix 8GB (2x4GB) DDR3 1600MHz
- Kingston SSDNow! V300 120GB
- PicoBox Z3-ATX-200 Power supply unit.
- Gembird rack 19″ 2U chasis with 2 fans connected to 12V
Some facts about software:
- Gentoo linux with kernel 4.0.5
- Semi-tickless kernel with 300MHz ticks
- Conservative CPU Speed governor with 800MHz to 2.05Ghz frequency
Testing hardware:
- Standard DC power supply with voltage and current monitoring, set to 24.1V
The „idle” state: Operating system running, one ssh client connected, no tasks
The „full load”: Linux kernel make -j5
In idle state, CPU Speed governor set minimal frequency on all cores. For testing purposes I also forced maximal frequency to check for differences.
In full load state CPU Speed governor set maximal frequency of course, I also forced minimal for checking.
The results
The table below show total power consumption under different conditions.
Test condition | Frequency (GHz) | Current (A) | Power (W) |
---|---|---|---|
Idle | 0.8 | 0.67 | 16.147 |
Idle | 2.05 | 0.67 | 16.147 |
Full load | 0.8 | 0.90 – 0.93 | 21.69 – 22.413 |
Full load | 2.05 | 1.35 – 1.40 | 32.535 – 33.74 |