Power usage of a HP MicroServer N40L

[twoj]

I just ran some power consumption tests and i thought i would share the results if anyone is interested.

The unit is a HP Microserver N40L, and i think the older N36L would be very similar.

The configuration is 5 Seagate Barracuda 7200.14 ST3000DM001 (7200rpm) plus 2 USB drives

When the unit first starts it seems it can get close to 150W, however this is just on startup and last less than 2 minutes.
Normal running usuage for my unit is 61-63W (no activity, ie no cpu or disk activity)
and in idle mode (hard drives powered down) is 36W

There are probably a bit more efficient NAS units (like the synology products) however i find this quite efficient for what the unit can offer.
Hope this helps.

To give a rough estimate, running this for 1 year if it was at 63W for a cost of $0.07/kWh (I live in an area that has relativly low cost of electricity) would cost you $38.63, equivalently at 36W the price drops to $22.08.
Once Nas4Free gets the WoL working for these units you could additionally shut the unit down and the power consumption would be negligible.

[simonlebon]

Thank you for these figures.
I'm currently using a home built system based on a AMD Dualcore X5050e which performance is excellent. I've read the performance of the Microserver is very good, too.

However, over here (in the Netherlands) electricity costs € 0,23/kWh (that's about $ 0.28/kWh), so every Watt usage means € 2 (= about $ 2.50) on a yearly base.

I used a efficient Power supply (Green Power) and I still have to measure the systems total power usage.
Your measurements helps me to decide wether to purchase a Microserver or not.

THANKS!

[ku-gew]

CPU frequency scaling active?
I have a HP ProLiant ML110 G7 (Xeon E3-1220) 18 GB DDR3 ECC with 4x 3TB drives (2 WD Red, 2 WD Green) and one 250 GB drive with OpenIndiana (I think power management is handled by BIOS) and it uses 62 W when idle.
It seems HPs are more or less the same.

As info: an Athlon 64 X2 3800+ with 2 GB RAM DDR2 667 and 2x 2TB WD Red and an old IBM 40 GB and idle it is about 55W (N4F with power management enabled).

[simonlebon]

Yes, the CPU frequency scaling is active. The level varies with the load from 1000MHz (10817mW) - 2600MHz (50000mW)
Power management is enabled.

The set contains:
-Gigabyte GA-MA74GM-S2H - AMD 690V chipset
-AMD ATHLON 64 X2 5050E 2.6GHz AM2 45W
-2*Corsair 2048MB XMS2-6400 DDR2 (4GB total)
-PSU: Antec EarthWatts EA-380D EC Green
-4*WD WD20EURS (in ZFS-r)

The disks are put in stand by after 30 minutes and the advanced power management is set to level 127 (Intermediate power usage with stand by). The Antec PSU has a good efficiency on 25-50W load, see: Tom's Hardware and AnandTech

So i guess the power usage will approx. be the same as the Athlon X2 3800+

[ku-gew]

Yes, the CPU frequency scaling is active. The level varies with the load from 1000MHz (10817mW) - 2600MHz (50000mW)
Power management is enabled.

I wonder where are those power vales coming from. Are they written inside the processor, or are they from a database somewhere else?

[simonlebon]

I think the latter, because the TDP of the 5050e is 45Watt (=45000mW).
When reading and writing several files from different machines, the load tops out at 80% max.
I do know the CPU stays cool, even under load.

[ku-gew]

Is it possible to get an official answer about the source of the power usage data corresponding to the different frequencies?

[b0ssman]

the psu in the 40L is very bad in terms of power efficiency.

there are users who have dropped those figures by using a pico psu or a 1u fsp psu.

according to this
http://www.hardwareluxx.de/community/f1 ... st20007596

the idle usage with a pico psu is 15w

[ku-gew]

Only 4W less than the standard PSU, not worth the effort.

[b0ssman]

he had 36W in idle

[ku-gew]

Where have you read that? 36W is the value measured by twoj in this forum, 15W is the value measured by totti1001 in the other forum. Unless they used the same instrumentation and hardware, you cannot compare.

What we know for sure is what totti1001 measured before and after:

Im Idle mit WHS2011 ist auch der Verbrauch etwas runter gegangen, 19W mit dem orig. Netzteil 15W mit der PicoPSU.

19W with the original PSU, 15W with PicoPSU. Same person, same instrumentation, same hardware, same OS. This is a comparison you can trust (more)! The values themselves not really, but the ratio before/after is (probably) reliable.
Of course, if the cos fi of the two power supplies is very different and the instrumentation is very bad (the cos fi affects reading of bad meters), then you can say nothing, but 36W against 15W is (improbable and) not reliable.

[ku-gew]

I will try next week the N40L with only an internal usb stick and no HDDs, my meter should be able to give good readings.

[ku-gew]

Concerning the MHz and mW values, they are retrieved (apparently) from the cpu itself:

http://forums.freebsd.org/showthread.php?t=16384

Code: Select all

$ sysctl -o dev.cpu
dev.cpu.0.%desc: ACPI CPU
dev.cpu.0.%driver: cpu
dev.cpu.0.%location: handle=\_PR_.P001
dev.cpu.0.%pnpinfo: _HID=none _UID=0
dev.cpu.0.%parent: acpi0
dev.cpu.0.temperature: 33.0C
dev.cpu.0.freq: 2000
dev.cpu.0.freq_levels: 2000/25000 1875/23437 1800/20736 1687/19440 1575/18144 1462/16848 1350/15552 1237/14256 1125/12960 1000/9680 937/9075 875/8470 812/7865 750/7260 687/6655 625/6050 562/5445 500/4840 437/4235 375/3630 312/3025 250/2420 187/1815 125/1210 62/605
dev.cpu.0.cx_supported: C1/0
dev.cpu.0.cx_lowest: C1
dev.cpu.0.cx_usage: 100.00% last 4965us
dev.cpu.1.%desc: ACPI CPU
dev.cpu.1.%driver: cpu
dev.cpu.1.%location: handle=\_PR_.P002
dev.cpu.1.%pnpinfo: _HID=none _UID=0
dev.cpu.1.%parent: acpi0
dev.cpu.1.temperature: 38.0C
dev.cpu.1.cx_supported: C1/0
dev.cpu.1.cx_lowest: C1
dev.cpu.1.cx_usage: 100.00% last 12017us

[ku-gew]

An interesting post about reducing power usage:
http://lists.freebsd.org/pipermail/free ... 06436.html

[ku-gew]

My new HP N40L eith 2x WD red 2TB still spinning and processor in low power (400 MHz) uses 21W. It makes sense conpared to the 19W mentioned in the other forum: 21W — 2x4.5W = 12W and without cpu power save it could consume some watts more and get to 19W already mentioned.

36W in idle without disks spinning is crazy for a N40L processor and mobo.

As info: I have 2 GB ECC.

[twoj]

When I did the test it was without touching any CPU scaling so presumably its using 100% full power which could explain the difference. The results I got are just from quick measurements - to do proper tests I would need to pull out the drives and test with different cpu scaling to verify the results. I have another n40l on the way that if I get time I will try to do some more detailed examination of the power consumption. I wasn't too worried about this one since it is offline most of the time.
I find it hard to believe that the n40l runs at 12W but i'm open to results if anyone else can post / confirm their consumption.
Also I tested the system stock - so as I said, no changing the cpu scaling, also that includes the stock fan which could be 5W right there.
I agree that the cost and time of switching the psu probably will not equate to a big difference in cost savings over the lifetime of the unit.

[V.I.P.E.R]

Hello Guys,

could anyone say anything about the power usage of a n40l/n54l with newer versions of nas4free ? Im interested in some values to compare them with my debian installation.

best
V.I.P.E.R

[crowi]

There is a german thread about this piece of hardware:
http://www.hardwareluxx.de/community/f1 ... 52079.html
And there is a collection of power consumption on differentr setups:
17W @ Idle -> 1x2,5" SSD, 2GB, SUSE 11.3 (Quelle)
18W @ Idle -> 2x3,5", 1x2,5" SSD, 4GB, 2008R2 (Quelle)
19W @ Idle -> 1x2,5", 8GB, 7/2008R2 (Quelle)
20W @ Idle -> 2x3,5", 1x2,5" SSD, 4GB, WHS2011 (Quelle)
21W @ Idle -> 2x3,5", 1x2,5", 8GB 2008R2 (Quelle)
22W @ Idle -> 2x3,5", 1x2,5", 5GB, TV-Karte, WHS (Quelle)
26W @ Idle -> 1x 2,5" SSD, 8GB ECC (1 Riegel), RAC, 8GB USB Stick, XPEnology DSM 4.2 (Quelle)
28W @ Idle -> 1x 2,5" SSD, 3x 3,5" WD RED, 2GB ECC (1 Riegel), OMV (Quelle)
30W @ Idle -> 1x3,5", Win2008 (Quelle)
31W @ Idle -> 2x3,5", Fedora 14 (c’t Ausgabe 2/2011, S. 54)
35W @ Idle -> 2x3,5", 2GB, PicoPSU, XPEnology DSM 4.2 (Quelle)
35W @ Idle -> 1x3,5", DVD, 4GB, Win2008R2 (Quelle)
36W @ Idle -> 5x3,5", 1x2,5", 8GB, USB3.0, Remote Karte, 2008R2 (Quelle)
36W @ Idle -> 3x3,5", Fedora 14 (c’t Ausgabe 2/2011, S. 54)
40W @ Idle -> 3x3,5", 1x2,5", 16GB, DVD-LW, picoXT, FreeNAS 9.2.1 (Quelle)
41W @ Idle -> 4x3,5", Fedora 14 (c’t Ausgabe 2/2011, S. 54)
47W @ Idle, 54W @ Last -> 4x3,5", Ubuntu 10.04 (Quelle)
56W @ Idle -> 4x3,5", 1x2,5", 8GB, 3ware 9650SE, KVM, DVD (Quelle)
40W @ Last -> 3x3,5", 1x2,5", 4GB, Win2008R2 (Quelle)
...

[V.I.P.E.R]

yeah - sure. I know it very well. But as you can see - no free bsd ..

[b0ssman]

doch
40W @ Idle -> 3x3,5", 1x2,5", 16GB, DVD-LW, picoXT, FreeNAS 9.2.1 (Quelle)