Power Consumption

H

hotzen

Limp Gawd
Joined
Jan 29, 2011
Messages
349
Hey guys,

I'm trying to get a realistic picture of energy consumption, using the soon-to-arive Ivy Bridge Xeons.

Is the following calculation somewhat correct? Any other things to factor in?
autudv.jpg



ODS-Sheet Download
 
Those are reasonable numbers. Except you are missing loss by power supply. Also remember that the CPU numbers will be closer to the about the same than your numbers support. The reason is the slower "energy efficient" processor will run longer its high power state than the faster processor to do the same amount of work so the difference in the real world will not be as much.
 
I actually have done power consumption testing recently with several Sandy Bridge setup. CPU+MOBO+16GB ECC+2.5"HDD consumed about 22W-26W at AC running Server 2008 R2 under my testing. There was a little variation between CPUs and Motherboards. I've tested various combinations of 1240,1260l.i3 with s1200BTL,X9SCM-F, S5512WGM2NR, and P8B WS, and the power consumption variations were only few watts. The idle power number of 1230 in your chart seems high, perhaps the figure was measure when not in its lowest c-state. One thing to note though, I've also tested the systems with Server 2008 R2, Solaris 11 Express, and ESXi 5, and I've found that ESXi had higher idle power consumption by about 8-10W over both Server 2008 R2 and Solaris. Surprisingly, there was only 1-2W difference between Solaris and Server 2008.

HDD power consumption might be a bit off depending on the model. I've tested lower rpm drives before, and they run about 3W idle and 5W load while performance drives usually run 6W idle and 10W load. On the other hand, I have Hitachi 7k1000.C 1TB which are probably most energy efficient 7200rpm drives, and they run 4W idle and 6W load at the lowest power setting in their power management utility.
M1015 does run about 5W at idle, so that seems in line.

PSU will change the idle power consumption obviously. I tested the systems with a 80plus 250W PSU and the power consumption rose 3-5W when paired with 500W 80plus PSU.

Please do some testing once you get Ivy Bridge Xeons and have them running. I'm curious if they would do any better in idle power consumption. I can't imagine it getting any lower, but it'd be neat if a basic server has lower power consumption than a typical CFL bulb. One of my servers with 6 disks run with same power as a 60W bulb, and I am continually astounded whenever I think about it. It was just a 4-5 years ago when I tested some 775 socket Xeon setup, and they ran 100W+ with just CPU+mobo+hdd.
 
drescherjm said:
Those are reasonable numbers. Except you are missing loss by power supply. Also remember that the CPU numbers will be closer to the about the same than your numbers support. The reason is the slower "energy efficient" processor will run longer its high power state than the faster processor to do the same amount of work so the difference in the real world will not be as much.
Click to expand...

That's actually a really good point. I saw some testing done between an Atom and an i3 system. The i3 system won out substantially for exactly that reason--the less time it took to complete a job, the faster it could return to a lower power state and overall save energy.

Interesting stuff, and worth considering "power over time".
 
Thank you very much, somename and others.
I am aware that a minimal TDP only affects the maximum heat dissipation and does not necessarily imply energy-efficiency. I just wanted to have rough figures comparing the 1230 with the 1260L, indeed I will choose the 1230.

I am very suprised by your idle-values of under 30W, that is awesome and way less than I hoped for. However I will take the +10W ESXi route... Thanks.
 
Best bet is get a clamp on meter or other measuring device and test at the wall socket, that way you get the actual numbers.
 
You must log in or register to reply here.
Back
Top