Power consumption test for a "first draft" server build

[Bullwinkle]

 

I've been building my first server and have some results on power consumption that might help others... so this note.

 

Several people and posts helped me progress to this point, but I was directed to this post by @Miska which influenced my decisions about the hardware guts:

 

 

I used the same processor (Intel i5-7600T), a mobo of the same flavor (Supermicro c7z270-cg-l), four 4Gb DDR4 memory sticks (Crucial CT4G4DFS8213), a 512Gb M.2 SSD (Toshiba OCZ RD400), and a 3Tb SATA3 hard drive (Hitachi HDS723030ALA640) where for now the Win 10 Creator OS lives. No video other than that onboard.

 

The case is a spacious Antec 2U rackmount with all fans yanked. The cpu is cooled by a (nearly one pound) aluminum heatsink with no fan (Arctic Alpine 11 Passive). The mobo receives its power from the HDPlex 300W DC to ATX converter, which in turn was fed from an Acopian lps with plenty of headroom (24 volt, 15 amp). I had communicated with HDPlex support and was told that, over the usable input voltage range, the HDPlex would have peak efficiency at the upper end of that range (which is 24 volts).

 

I wanted to know how many watts this thing actually eats when I push it. (The Acopian is a monster and hums like the 18-minute gap on Nixon's tapes; I ultimately want something silent and less obnoxious-looking.) I pushed it with the Sisoft Sandra Lite benchmarking software, running several cpu-intensive benchmarks.

 

As I expected, it was the image-processing benchmarker that pushed the cpu to its limits: All four core temperatures quickly rose to the 80 C junction temperature of the i5-7600T and stayed there for a good five minutes or so as each of the cores quickly went flat out at 100% utilization. I was very happy they didn't melt. I may need to revisit cooling. However the multimedia benchmarker produced the highest peak current (though only very briefly and the time integral of current was clearly greater for image-processing).

 

The reason for posting this is the power consumption numbers, however--the rest above is just background. I put a meter in the line between the Acopian and the HDPlex and monitored the current on that line pretty continuously (only glancing every now and then at the HWMonitor display to check on temperature). The maximum current I read was just under 3.1 amps, for a peak system power consumption of 3.1 amps x 24 volts = 74.4 watts.

 

This is good news. I have been told that many of those online calculators of power requirements give very high estimates; I had used one for guessing at this thing's appetite and that guess was 188 watts under duress and 133 watts normal.

 

It also suggests that one can pair an industrial lps (such as say an Acopian A24H1200 or A24H850, which you will find used on ebay for fifty bucks) with something like the HDPlex 300W converter and get plenty-enough clean power on the cheap -- or at least that's my guess as of today -- for many applications folks want to pull off. Or does it? Waiting to be schooled...