random picopsu/snb/mini-itx idle power consumption numbers

PSUs: The source of DC power for all components in the PC & often a big noise source.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee, Devonavar

Post Reply
matt_garman
*Lifetime Patron*
Posts: 541
Joined: Sun Jan 04, 2004 11:35 am
Location: Chicago, Ill., USA
Contact:

random picopsu/snb/mini-itx idle power consumption numbers

Post by matt_garman » Sat Jan 07, 2012 11:06 am

A few months ago, I posted a writeup and picture of my HTPC. One of the things that always left me curious was that the following configuration used about 24 watts of AC power when idle: i5-2500k, asrock h67m-itx, picopsu-160-xt + edac 150w brick, 2x4 GB RAM, SSD. Other people with similar configurations seemed to usually get 20 Watts or below.

So I played a little bit. Due to time constraints, I couldn't test as thoroughly as I'd have liked, but I got a few readings that I thought I'd share.

First, I rebuilt my HTPC. I moved it into a new case (Apex MI-008), and replaced the two-pci-slot-wide Zotac GT430 with a single-slot-width passive Asus GT430. I removed the WD Green storage drive, kept the SSD system drive, and have only one fan (Thermalright TY-140, on top of TR AXP-140). This is a super tight fit (had to hack the case a bit to fit it all in there), and I'll try to post some pics eventually.

Anyway, this configuration, without the GT430 had the following idle AC power consumption: 21-22 Watts. Wait, that should be 24--25 right? That's what I said above and in my previous post. Well, two things: since taking that first reading, I down-volted the RAM from 1.4V to 1.5V. Also, the other config had two fans, the TY-140 and a Noiseblocker PK-3 fan. When I pull the TY-140 fan's plug, power consumption goes down another watt. So I think it's reasonable to assume that 2--3 Watt AC difference comes from the fan and RAM undervolt.

That's with the Edac 150 Watt AC-DC brick, model ea11803a-120 (fanless).

I bought a picopsu-90-xlp kit with an 80 Watt brick, model ea10953a (also fanless). If I used this brick with the picopsu-160, power consumption went down another watt, to 20--21. In other words, same exact config described above, changing only the AC-DC brick (i.e. still haven't tested the picopsu-90 yet). That to me clears the efficiency concerns of the Edac 150W unit; it's practically the same as the 80W unit.

Now, I went back to the 150W brick, and put the Asus GT430 in. Idle power consumption reads 35 Watts! I wanted to see if I could shave a few more watts, so I underclocked the RAM (it's rated 1333, but I set it to 1066), and also undervolted it further, down to 1.325 V. It tested clean for 8 hours of memtest86+, and brought idle power consumption down to 30--31 Watts! I consider that a pretty massive savings.

One more note: for all of the above testing, two cores were disabled in my i5-2500k. But when I was doing similar testing several months ago, disabling cores didn't seem to make any difference in idle power consumption. And from what I've read online, it seems generally true that all the Sandy Bridge chips have very similar idle power consumption.

The RAM model I used in the above testing is: Corsair XMS3 cmx8gx3m2a1333c9. Two 4GB DIMMs (8 GB total). Yes, total overkill for a HTPC. :)

Another takeaway from all this is it suggests that the GT430 adds an additional 13--14 Watts of AC power consumption. (I know you can't directly extrapolate power like that, since the efficiency curve of the PSU masks the real value, but it's good enough for the kind of rough estimations I'm doing here.) In other words, I believe that if I were to remove the GT430 after further tweaking the memory, my idle power consumption with this configuration would be 19--20 Watts. (You might think that would be easy enough to test, but it's a bear getting that GT430 in and out of the Apex MI-008 case!)

Now, for another project I'm working on, I acquired a second Asrock H67M-ITX board, an i3-2100, and a Habey EMC-800B case. The Habey case comes with a built-in DC-DC power supply: E.Mini LR1005-120W12VDC-11G20, and the included AC-DC adapter is an 80W model: E.Mini wt1205000. I used one stick of the following RAM for this testing: Crucial 2rx8 pc3-10600e-9-10-e0 ct25672ba1339.18ff 2gb (yes, that is ECC memory because that's what I happened to have laying around). I used a Kingston 16GB SSD for the system drive, and the stock Intel heatsink/fan.

With the E.Mini power supply, idle power consumption is about 20 Watts. If I swap the E.Mini AC-DC brick with the Edac 80W brick, it's the same.

If I use the PicoPSU-90-XLP and Edac brick, I get 18 Watts idle. Now that's more like it---the sub-20 Watt idle power consumption is what I've seen a lot of people get with these SNB mini-itx picoPSU configurations. Also note, based on my previous experience (see last sentence of first post here), the Intel DH67CF board is another two or three watts more efficient at idle than the Asrock board. Presumably, using that board would take me down to 15 or 16 watts idle power consumption. And I consider that close enough (given my imprecise methods) to the Missing Remote 14.5 Watt idle power consumption that I've considered a "benchmark" of sorts.

So... what's the conclusion? Ugh, umm... I guess my takeaways are as follows: (1) 8 GB of 1333 RAM at 1.5V clearly adds a few measurable watts of power consumption. If you don't need the performance, drop the speed and voltage down for some easy power savings. (2) The picopsu-160 + Edac 150W brick might add a watt or two versus the picopso-90 + Edac 80W brick. (3) The GT-430 is a power hog relative to the rest of my system, accounting for 1/3 the power consumption alone!

Here's my raw notes... after you've read the above, the below might be a useful summary (or just make you more confused ;)).

Code: Select all

crucial mem: 2rx8 pc3-10600e-9-10-e0 ct25672ba1339.18ff 2gb

habey emc-800b dc-dc: E.Mini LR1005-120W12VDC-11G20
               ac-dc: E.Mini model number: wt1205000 12v / 5.0 A
               output

20111229:
    media pc:
        asrock h67m-itx
        i5-2500k (2 cores disabled)
        2 x corsair xms3 cmx8gx3m2a1333c9 (undervolted 1.4v) (8gb total)
        picoPSU-160-XT
        kingson 16gb ssd
        only fan: thermalright ty-140

        ac/dc brick:
            edac ea10953a (12v/6.6A/80w) ............ 20--21 Watts
            edac ea11803a-120 (12v/12.5a) ........... 21--22 Watts

        rest using ea11803a:
            replace mem w/1 stick 2gb ecc crucial ... 20 Watts
            stop cpu fan ............................ 19 Watts
            add asus engt430 ........................ 35 Watts
            reduce dimm voltage 1.325,
                underclock ram 1066 ................. 30--31 Watts

    config #2:
        asrock h67m-itx (different board than above)
        i3-2100
        picoPSU-90-XLP
        1 stick 2gb crucial ecc mem
        kingston 16gb ssd

        picoPSU-90-XLP + edac ea10953a .............. 18 Watts

        habey emc-800b psu + 60w acdc brick ......... 20 Watts
                            edac ea10953a ........... 20 Watts
Anyway, I hope others find this useful, or at interesting enough to be worth the time you spent reading it.

p.s.: all measurements were taken on a Kill-A-Watt. I'm the USA, so have 110V mains. I have two Kill-A-Watts actually, but both have always taken consistent measurements from 15 watts and up. And I did a "poor man's calibration" on them using 25W and 50W incandescent light bulbs. I wouldn't bank on this precision, but I think it's good enough for an Internet forum posting. :)

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: random picopsu/snb/mini-itx idle power consumption numbe

Post by Abula » Sat Jan 07, 2012 11:25 am

Nice write up, very interesting stuff, im my experience the lowest i have been able to take my mini mi was around 13.7W, this is without mouse, keyboard and mouse, but with 3 fans, probably with one fan i could go around 12 or 11W, with the memory i wished the intel mobo could tweak it... my memory is 1.25V, but i never been able to set it even though the bios has a place for changing it doesnt stick.

The only brick that i do want to test at some point is the 60W level V, but wasnt out when i bought my picoPSU, but that with 90XLP is what i think would net the lowest consumption... maybe with ivy bridge we could break down the 10W barrier (not cosidering atoms or mobile cpus).

Post Reply