Intel DQ45EK Mini-ITX LGA775, 4 Sata, 1 E-Sata, 1 PCI-E x1, 26w idle - Works!


Recommended Posts

I just picked up this board to replace my ancient laptop domain server with Server 2008 64bit running a Hyper-V WHS. (Edit: Hyper-V network performance was horrible for me so this didn't work out.)Of course I had to test it with unRAID first.

 

There isn't much to report really. The Intel gigabit nic is supported by the infamous e1000 driver and my five drives attached to the ICH10DO ports were detected. The parity creation and check had no issues, and I've attached the syslog. Multi-gigabyte files copied over and matched their checksums too. Everything seems to work as it should.

 

This is the most efficient socket 775 motherboard available. With an E5200 and one 2gb DIMM, I measured 27w at the wall when idle in unRAID. That's without drives and with a 80% bronze 425w psu. Plugging in the gigabit network added 2w to that, and a second 2gb module added another watt.

 

A WHS user with the same cpu/mobo has reported 39-40w when idle with a picoPSU-120, 3 WD10EACS drives, and a 140mm fan. Not bad for a desktop cpu, chipset, drives, ram, and fan. http://www.xiirus.net/articles/article-custom-windows-home-server-build-putting-it-all-together-e1g2y.aspx Since he finished his build last month, a newer 150w Pico Psu has come out. It now has 24 pins and includes a 12v aux power cable for the cpu. It's a much better fit for recent mini-itx offerings.

 

This board also features Intel's AMT remote administration technology. This doesn't seem to be terribly useful with Unraid, but it does enable remote bios configuration. The remote power control of the system from a web gui, even if the systems frozen is a nice trick too. In any case, its more useful than graphic hd decoding, 7.1 audio, HDMI on the similar HTPC oriented DG45FC.

 

All in all, I'm still leaning towards a Gigabyte GA-MA780G-UD3H build for a lower power Unraid server. With an undervolted processor, power consumption should also be below 40w and the board has massive expansion possibilities with 5 pci-e slots. The micro-atx GA-MA74GM-S2 would also be great with the upcoming Supermicro 8 port pci-e x4 card. Add a pci-e x1 Syba card with the cache drive on it and there would be 15 full speed data ports for $200. That's a great start compared to $100 per port NAS setups. (Edit: I later purchased and tested that board. It was disappointing overall. In the end I settled on a G33 ATX board that only used a few watts more than this Intel one.)

Link to comment

moredrives, Nope, none of these basic Intel boards have undervolting. An undervolted 45w AMD on a 690g/740g/760g/780g board will use less energy, have more expansion slots, and be cheaper. It makes more sense for uNRAID as the Intel mini-itx form factor is wasted on multiple 3.5" drives. Newegg got the GA-MA74GM-S2 back in stock today and I've been debating purchasing it all day.

 

smino, HashCalc in Windows. I had it installed for MSDN isos.

Link to comment

The  GA-MA74GM-S2 looks like a nice board. It's more than half the price of the other 740G's I've been eyeing. The only problem is it's not mini-ITX which is a must for the Chenbro ES34069 I've fallen in love with. I agree that undervolting/underclocking is the way to go. Difficult, if not impossible, to find in these mini-ITX platforms though.

Link to comment

The  GA-MA74GM-S2 looks like a nice board. It's more than half the price of the other 740G's I've been eyeing. The only problem is it's not mini-ITX which is a must for the Chenbro ES34069 I've fallen in love with. I agree that undervolting/underclocking is the way to go. Difficult, if not impossible, to find in these mini-ITX platforms though.

 

if you go with a mobile CPU there will be a difference in voltage to start.

if you add CPU speed or something similar there will be measurable results in power savings.

The only issue is initial cost. Mobile CPU's are not cheap, however I've purchased each of mine used on eBay at a fair price.

 

 

Link to comment

I swapped this board into my unRAID array and ended up with a couple small surprises. The 8gb flash drive wouldn't boot without usb drives being set as  "Removable Disks". That seems to be like "Forced FDD" mode on the P5B boards. It gave the flash drive it's own boot category which automatically booted first with the boost usb first option.

 

Also unRAID 4.4.2 didn't detect the Intel network adapter. Going to 4.5 beta 2, like my second test key, fixed that.

 

The idle at the wall watt consumption was good, as expected. 48w with 5 1.5tb Seagates, 1 750gb Samsung, 620w Seasonic S12, an e7400, 4gb, gigabit connected, 92mm cpu fan, and 2x 120mm fans for the drives. With all the drives spun up and idle, I think it was 98w. Another failing 1.5tb drive got my attention at that point.

Link to comment

The four WD10EADS drives for this build arrived today. With then, an e8400 c0, 4gb, gig-e, and a seasonic s12 430w the system is pulling 38w when idle, 51w when spun up, and 64w writing. Power usage was 36w without any drives.

 

I'll check these numbers with a Chenbro ES34069-180w when it arrives on Thursday. Tomorrows going to be more interesting though. Gigabyte GA-MA780G-UD3H and 5050e play time.

 

 

Link to comment

The Chenbro ES34069-180w arrived and I was able to assemble it during lunch.

 

With an e8400 c0, 4gb, Chenbro HSF, and the two 60mm case fans:

30.5w idle with no drives, network unplugged

38.5w with 4 drives spun down, gig-e linked (seems like too big of a jump, I need to recheck the 30.5w number)

50w with 4 spun up but idle (heads parked) drives, gig-e connected

64w during parity creation, gig-e

103w spike when turning on

 

Basically the same as my Seasonic.

 

The Foxcon 180w brick surprised me by being being fanless, but the case fans were louder than I expected. They do keep the WD green drives extremely cool though. Depending on the drive temperatures this summer I might swap them out or wire them to a fan controller. The Antec 200mm fan I bought to mod into the side panel might make a similar amount of noise on medium. On low it's nearly silent.

 

Edit: This build has now ran for a week with Server 2008 on a OCZ Vertex 30gb. With 3 drives spun down and the hyper-v whs drive up it uses 42w. When all the drives are spun up and there's light system background activity it uses 60w. Over 154 hours it has averaged 53w.

Link to comment

Thanks for the update.

 

BTW, you can program the WD drives to power on in standby mode.

This will prevent spn up until unRAID starts thereby diminishing the startup spike.

There have been people who said this board with a high powered CPU stresses the power supply when you do heavy duty computation.

 

I suppose they are also using higher powered disks.

 

I would try the power on in standby mode to see how it works.

 

I have not had any issues, however I am using a 2.4ghz mobile CPU.

 

 

Link to comment

...

48w with 4 idle drives, network unplugged

38.5w with 4 drives spun down, gig-e network

...

Are one of these a typo? It uses almost 10 watts less with gig-e on?

 

Great setup. I'm definately going with that case but still leaning toward a 740g mini-itx that I hope delivers the same or lower power usage. Hope to lower it even more if I ever determine whether these Jetways have underclock/undervolt BIOS options. I'm amazed how low the power usage is with an e8400 that's not downclocked.

Link to comment
I'm amazed how low the power usage is with an e8400 that's not downclocked.

 

UnderCLOCKing alone will have practically no effect on power usage.

 

UnderVOLTing is what will reduce power usage.... but you generally can not undervolt very much unless you also underclock.... so you have to do both.

Link to comment

@ WeeboTech, I wasn't aware of the PUIS jumper, that might be handy in a future build. Thanks. Have you used the WD Idle tool to change the head parking time from the default 8s delay? I need to increase it on the WHS drive, so I'm considering setting it to the 25.5s max on all 4 drives and seeing how that works out while I have the case open near me.

 

@ moredrives, the "idle" drives were spun up but had their heads parked. The gig-e added 2w in earlier testing. Even older Intel adapters like the Pro 1000 PT only use 5w.

 

I've decided not to keep the Gigabyte 780g setup I purchased. A 1ghz .8v AM2 doesn't hold up to the 45nm Wolfdale's similar idle numbers and working throttling. I'll have a cheap Foxconn G31 board later this week that I'll pair with the e5200.

Link to comment

@ WeeboTech, I wasn't aware of the PUIS jumper, that might be handy in a future build. Thanks.

 

You can also set it with hdparm using the -s1 --yes-i-know-what-i-am-doing parameters.

 

 

Have you used the WD Idle tool to change the head parking time from the default 8s delay? I need to increase it on the WHS drive, so I'm considering setting it to the 25.5s max on all 4 drives and seeing how that works out while I have the case open near me.

 

I have not had this need. So I left that at the default.

 

 

 

Here is my scriptlet to set this from the go script (actually /boot/custom/etc/rc.d/S01-hdparm

You'll need to change your disk assignments.

Use either the /dev/sd? or the serial number symlinks as below.

 

#!/bin/bash

PUIS="-s1 --yes-i-know-what-i-am-doing"
HDPARM_TEST="-Tt "
UDMA6="-X udma6"

TMPFILE=/tmp/S01-hdparm.$$
trap "rm -f ${TMPFILE}" EXIT HUP INT QUIT TERM

while read DISK PARM
do
    [ -z "${DISK%\#*}" ] && continue
    echo
    ls -l ${DISK} > ${TMPFILE}
    read -a LS    < ${TMPFILE}
    # hddtemp -f /dev/null ${DISK} 2>/dev/null
    hdparm -I ${DISK} | egrep -i 'Number:' | tr -s " " " " | tr -d '\t\n'
    echo " -> ${LS[10]}"
    if [ "${SET:=YES}" = "YES" ]
       then hdparm ${PARM} ${DISK}
    fi
    if [ "${TEST:=NO}" = "YES" ]
       then hdparm ${HDPARM_TEST} ${DISK}
    fi
    # hdparm -I ${DISK} | egrep -i standby
done <<-EOF
# Symlinked devies by serial number
# include list with a similar command like this.
# ls -1 /dev/disk/by-id | grep scsi-SATA* | grep -v part >> S01-hdparm 
/dev/disk/by-id/scsi-SATA_ST31000340AS_3QJ08RZW -S242
/dev/disk/by-id/scsi-SATA_ST31000340AS_5QJ0BH1D -S242
/dev/disk/by-id/scsi-SATA_ST31000340AS_9QJ1Q30B -S242
/dev/disk/by-id/scsi-SATA_WDC_WD10EACS-00_WD-WCASJ0353226 -S242 ${PUIS}
/dev/disk/by-id/scsi-SATA_WDC_WD10EACS-00_WD-WCASJ0437718 -S242 ${PUIS}
/dev/disk/by-id/scsi-SATA_WDC_WD10EACS-00_WD-WCASJ0808397 -S242 ${PUIS}
/dev/disk/by-id/scsi-SATA_WDC_WD10EACS-00_WD-WCASJ1352706 -S242 ${PUIS}
/dev/disk/by-id/scsi-SATA_WDC_WD10EACS-00_WD-WCASJ1353170 -S242 ${PUIS}
/dev/disk/by-id/scsi-SATA_WDC_WD10EACS-32_WD-WCASJ0126857 -S242 ${PUIS}
#
# Hard coded devices 
# /dev/sda -S242 -s1 --yes-i-know-what-i-am-doing
# /dev/sdb -S242 -s1 --yes-i-know-what-i-am-doing
# /dev/sdd -S242 -s1 --yes-i-know-what-i-am-doing
# /dev/sdc -S242 -s1 --yes-i-know-what-i-am-doing
# /dev/sde -S242 -s1 --yes-i-know-what-i-am-doing
# /dev/sdf -S242 -s1 --yes-i-know-what-i-am-doing
# /dev/sdg -S242 -s1 --yes-i-know-what-i-am-doing
EOF

 

Link to comment
  • 2 weeks later...

WeeboTech,

Can you dumb down your post a little bit for me?

 

I have 3 x 1TB WD Caviar Black drives and want to stagger spin-up. With the PUIS jumpers attached, I get

 

...kernel: ata3.00: failed to IDENTIFY (SPINUP failed, err_mask=0x4)...

 

for each of the 3 drives during boot up. It boots fine and by the time I get to a client and open the web interface, all the drives are spun up and working properly / shares work fine too.

Can you give more details about your script and how to set it up?

Do I create a shell script and put it in the /config directory of the flash drive and then call that in the go script? What will the script's path be during boot up? If I want to just use the PUIS jumpers (instead of using hdparm -s1) how would your script be different (or can I even do this / maybe I don't understand - won't be the first time)?

 

Bottom line: how to get rid of the errors on boot (or do I just ignore them)?

 

Thanks from a linux greenhorn

Link to comment

I have 3 x 1TB WD Caviar Black drives and want to stagger spin-up. With the PUIS jumpers attached, I get

 

...kernel: ata3.00: failed to IDENTIFY (SPINUP failed, err_mask=0x4)...

 

for each of the 3 drives during boot up. It boots fine and by the time I get to a client and open the web interface, all the drives are spun up and working properly / shares work fine too.

 

Bottom line: how to get rid of the errors on boot (or do I just ignore them)?

 

Ignore them, I think it's part of the startup because the bios has not recognized them.

If they spin up one by one from the kernel and are recognized after spin up, you;re good.

If you are using the PUIS jumper, you do not need my script.

 

Link to comment

WeeboTech,

Thanks for the reply - sorry to be so dense.

 

I rebooted a few times and paid more attention to the messages on the console (they fly by pretty fast).

With the PUIS jumpers attached, the BIOS does recognize the drives - I see them identified prior to booting.

During boot, I hear the drives spinning up one at a time. Each time I get the following messages (one set of these for each drive in turn):

 

Apr 12 21:12:47 HunRAID01 kernel: ata1.00: qc timeout (cmd 0xef)

Apr 12 21:12:47 HunRAID01 kernel: ata1.00: failed to IDENTIFY (SPINUP failed, err_mask=0x4)

Apr 12 21:12:47 HunRAID01 kernel: ata1: link is slow to respond, please be patient (ready=0)

Apr 12 21:12:47 HunRAID01 kernel: ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 300)

Apr 12 21:12:47 HunRAID01 kernel: ata1.00: ATA-8: WDC WD1001FALS-00J7B0, 05.00K05, max UDMA/133

Apr 12 21:12:47 HunRAID01 kernel: ata1.00: 1953525168 sectors, multi 0: LBA48 NCQ (depth 0/32)

 

Note the lines following the failure notice. It's clearly recovering from that just fine (I just have to read ALL the lines - duh!).

 

By the way, on my configuration this adds a 15-second delay per drive. During each of those 15-second intervals I can hear the individual drive spin up and then I get the "SATA link" message. That's plenty of time for the power spike to settle down and, in my opinion, a small price to pay to avoid the power spike of spinning them all up [nearly] at one time.

 

Thanks again.

Link to comment

By the way, on my configuration this adds a 15-second delay per drive. During each of those 15-second intervals I can hear the individual drive spin up and then I get the "SATA link" message. That's plenty of time for the power spike to settle down and, in my opinion, a small price to pay to avoid the power spike of spinning them all up [nearly] at one time.

 

I've noticed the same delay.

What I like about this is during debugging of hardware I can power the machine up without all the drives spinning up.

If I'm having bios or hardware issues it relieves the strain on the drives spinning up and powering off.

 

My bios and SIL3132's handle the power up in standby fine.

Link to comment

When I briefly had 4 1tb greens connected the parity check speed seemed to fluctuate and didn't seem to be full speed. Every other Intel ICH* controller has worked great for me so it was a bit odd to see. I've been intending to retest it with fast 1.5tb drives, check out the cheap pci-e x1 riser I found on ebay, and play around with the side mounted 200mm fan on the Chenbro. Maybe tomorrow.

 

For an economical file server, a mATX G31 board that costs almost $100 less and only uses 2-3w more power makes a lot more sense. I shouldn't have gone so far with this build because it didn't work out as intended and cost too much. The e8400, 30gb Vertex ssd, and server 08 sure are fast though!

Link to comment

SSD are not all they are cracked up to be. Their write speeds are slower than disk, and read speeds are also a little slower.

The 1.5GB Seagate drives are faster on both read and writes, and cheaper. The bottleneck will most likely be the GigE Network.

I want to test network teaming soon.

 

Also overtime, the SSD tend to lose their speed.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.