Power consumption of freeBSD vs other OS

G

ghostdunks

Just wanted to find out other people's impressions of the power consumption/power efficiency of computers running FreeBSD vs other OSes. Based on my limited testing(possibly flawed), my impressions is that FreeBSD has ended up not faring so well vs Windows7 and Ubuntu, and wanted to see if other people have experience the same or suggest any tweaks I can make.

Built a rig with following components:
  • Intel G620 CPU (Sandybridge)
  • Intel DH67GD motherboard
  • 2 x 2gb Kingston RAM
  • Seasonic 330W power supply

The above setup is running headless, so no keyboards/mice/display devices are connected. For testing purposes, no HDDs are connected internally(to be used for RAID later)

I tested with the following OSes:
  1. FreeNAS 8 (based off FreeBSD 8), with powerd running - 31W
  2. FreeNAS 7 (based off FreeBSD 7), with powerd running - 31W
  3. Windows 7 Ultimate - 23W
  4. Ubuntu Desktop 10.04 - 28W(with display/keyboard/mouse connected). Believe this is <25W with the peripherals disconnected.

I realise that FreeNAS is only a derivative of FreeBSD and haven't tested with the latest version of vanilla FreeBSD yet,but I reckon it'll still be similar to the figures I'm getting with FreeNAS.

Is FreeBSD just known as not so power efficient? I would have assumed that it would be better than a bloated beast like Win7....

Would appreciate any thoughts.
 
After enabling powerd, there is still question about graphics chip frequency downscalling which is probably not supported under FreeBSD.
 
Yup, powerd was enabled on both versions I tested, with the default options. Could see that the cpu frequency was lowered, from 2600Mhz to around 350-600 or something like that. Lowered the temp on the cores from around 29-28 degrees to about 26. Even with the cpu frequency throttled, the idle power usage only dipped maybe 1 watt. The Sandybridge cpus are supposed to idle very well, even without the cpu freq being throttled, so I'm thinking in terms of power usage for SB chips, powerd is not needed or has very little effect.

Hmm, good point on the graphics chip frequency downscaling..Maybe I just disable the IGP altogether for my tests, don't need it anyway since its headless.
 
There are also sleep states which should lower power consumption even more, check
# sysctl dev.cpu.0.cx_supported

For example on my box:
Code: 
% sysctl dev.cpu.0.cx_supported 
dev.cpu.0.cx_supported: C1/1 C2/1 C3/162

To set them: (for each core)
Code: 
sysctl dev.cpu.0.cx_lowest=C3
sysctl dev.cpu.1.cx_lowest=C3
 
Thanks for the tips. my CPU supports C1 C2 and C3, so I tried setting it to C3, didn't do anything though.

I'll have a look at that link and see what else I can play with.
 
How did you measure power consumption?

If there was no hard drive you did a netboot?

28W is too low for such config. An moatherboard with atom will eat 35W.
 
My personal experience has been that the BSD family of operating systems are more power efficient on the varying CPU architectures.
 
I measured the power consumption by wattometer, the Intel Core 2 Duo T8100 with 965GM Mini-ITX board and 2 * 2TB Seagate Barracuda LP drives uses about 39W on my box (with 80PLUS PSU 200W if I recall correctly).
 
You must log in or register to reply here.
Back
Top