Intel G1610 vs AMD E350 .. idle power usage ?

[cyan]

I'm trying to build small (6 hdd) server

Since AMD E350 + 2 sata pcie card cost almost the same with Intel G1610 + mobo .

I wonder which one have better idle power usage ?

 

Thank you

[b0ssman]

People have gotten the e350 down to 10 watt idle. That's by using ulv ram, pico psu and undervolting the CPU.

No one has achieved that with a ivy bridge yet. Maybe haswell when the celerons come out.

[garycase]

I have 6 3TB WD Reds on a SuperMicro X7SPA D525 Atom board, and it idles at 20w with the drives spun down.  Excellent performance, and very low power consumption.

 

I've thought about trying the Pico PSU, but I don't think it can handle the spin-up power for 6 hard drives.    The problem with any of these very-low power draw systems is that the PSU is going to be operating well below its maximum efficiency, so there's at least 2-3 watts "wasted" just due to PSU efficiency loss.

 

[dirtysanchez]

Just to throw in an Ivy Bridge comparison, I have a Core i3-3220T on an Asus P8H77-I mobo with 5x 7200 RPM drives, and it idles at 33W with all drives spun down.

[Pauven]

AMD E350 + 2 sata pcie card

 

Since you have to add-in a pcie card to get enough SATA ports with the AMD E Series, the add-in card might push your power consumption quite a bit higher, eliminating any efficiency advantage the E350 probably has. 

 

For maximum efficiency on a 6-drive server, you will want a MB that has all necessary SATA ports on-board... stay away from any add-in cards.  Unfortunately, that removes the AMD E Series from consideration (unless someone has found an E Series MB with 6+ SATA ports...).  To get an AMD solution with enough on-board SATA ports, I think you have to go with the A Series, which seem to idle in the 20's - I have an A10-5700 that idles in the mid twenties, and I've read of a A4-3300 that idles in the high twenties - so I don't think the A Series can win on efficiency.

 

The G1610 has a surprising amount of horsepower, tons more than you would ever need for unRAID (it doesn't even break a sweat on my 16-drive server), and it is certainly faster than the E350.  While my particular setup idles around 17W (without drives), that is with a barebones MB that only has 2 SATA ports.  You'll have to get a more fully featured MB that has at least 6 SATA ports, so you should expect the power consumption to be a few watts higher, may 20W as a guess.

 

If you really want to save electricity, you would want to go with Atom, as garycase suggested.  I think that is the only max efficiency option.  The Atom should beat the G1610 by 4 to 7 watts.

 

We really need someone with one of these Atom based servers to try out a Pico PSU.  Assuming it has the power to spin up 6 eco drives (unknown), you could shave 2-3 watts off of garycase's number.  Keep in mind that unRAID does not, and probably never will, support staggered spin-up.

 

The i3-3220T is a 'Throttled' processor.  It idles the same as a non-T processor, only limiting max frequency and voltage under load.

 

-Paul

[garycase]

In addition to the idle power comparisons, don't forget that you're comparing two VERY different CPU's in terms of "horsepower" !!  The G1610 is more than 3 times as powerful as an E350 [PassMark 2584 vs. 775 for the E350].    If you want to run any add-ons that may need some of this processing power, there's no comparison.    An E350 is more appropriately compared to an Atom-based Intel system, both in terms of power consumption and CPU processing power [An Atom D525 scores 694 on PassMark, the D2700 scores 841].

 

[b0ssman]

Unfortunately, that removes the AMD E Series from consideration (unless someone has found an E Series MB with 6+ SATA ports...).

 

what about this

 

http://www.asus.com/Motherboards/C60M1I/

[ZeroK]

I have both the asus e350 mobo and the asus c60 mobo and I will tell you I prefer my e350 with unraid than the c60.  The c60 has a slower processor and it really shows transferring files to unraid.  Streaming is not bad, I think I had 3 streams running at one time, but streaming and copying at the same time was horrible.  The e350 never had issues with this.  I have 6 drives running off the c60.  On the e350 I had a supermicro add-in running 8 drives and my parity drive was a 3tb WD off the mobo.  The e350 ran for two years flawlessly until recently I replaced it, and no it wasn't with the c60.

[Pauven]

Interesting, I wouldn't have expected the C-60 to be so bad - good to know.  The biggest difference between C and E series, besides frequency and graphics capability, is that C is Ontario core, and E is Zacate core - otherwise the C-60 and E-350 have very similar specs (dual core, single channel DDR3).  I wonder what made the C-60 experience so bad for you.

 

Since both the C series and E series are BGA413, so I wonder if ASUS makes another version of that motherboard with an E-350/E-450 soldered on instead of the C-60.  That motherboard has enough SATA ports, so I am intrigued.

 

Edit:  Answered my own question, ASUS makes the E35M1-I with 6 SATA:  http://www.asus.com/Motherboards/E35M1I/#specifications

 

This is not the one listed on NewEgg (my goto site), the one on NewEgg has 5 internal SATA and 1 external eSATA.  So be careful, pay attention, and you might find one of these available somewhere.

 

From what I've read, the idle power consumption of an E-350 board is about half that of the D525 (X-bit shows 7.3W versus 16.8W, so less than half)!!!  If true, this could be a 12W server, versus a 20W server with a Atom.  Simply amazing!!!

[ZeroK]

I will say that my c60 was open box from Newegg.  I don't think there is anything wrong with it as it's now currently running the free version of unraid versus running the pro version before, but it hardly ever gets used as I'm using it as a "store unsorted crap until I get time to figure out where to put it" type of system.

[garycase]

a "store unsorted crap until I get time to figure out where to put it" type of system.

 

...That's exactly why I keep buying more (and larger) hard drives 

[dmacias]

I had a e35m1-i (E350) and currently a C60m1-i.  Both have 6 sata ports but the e35 only has 2 sata 6Gb/s & 4 sata 3Gb/s ports and the c60 has 6 sata 6Gb/s. I know with some older 2Tb sata 3Gb/s green drives it would take 30-36 hrs to preclear. I now have new  2Tb WD sata 6Gb/s green drives with 1Tb platters and it took less than 24 hrs to preclear them. 

 

All my movies and tv shows are 720p or 1080p. I have no problems with several different computers watching them and sabnzbd downloading and unraring to ssd cache, mysql, xbmc plugin, apache, sickbeard and even had tvheadend streaming hdtv.

 

I don't write a lot to the array from other devices.  Everything is mostly downloaded to the cache drive and moved with the mover script later.

[Pauven]

with some older 2Tb sata 3Gb/s green drives it would take 30-36 hrs to preclear. I now have new  2Tb WD sata 6Gb/s green drives with 1Tb platters and it took less than 24 hrs to preclear them. 

 

I think the performance difference is related to the 1TB platters in the newer 2TB drives, and I personally don't think the 3Gb/s connector speed was any impact for a spinning 5400rpm hard drive.  You need a fast SSD before 3Gb/s becomes a constraint.  I don't think a couple older 3Gb/s ports on the E35M1 would be a detractor for unRAID use.

 

That said, I looked and ASUS has newer E45M1 edition motherboards that have all 6 SATA ports as 6Gb/s... unfortunately one of those six is an external eSATA port.  I suppose you could always route the external cable back into the case, and use an adaptor to make it plug into an internal SATA drive - not elegant but might be worth it. 

 

So if I'm understanding you correctly dmacias, you have good results with the C60M1-i, unlike ZeroK?  Can either of your report your idle power consumption?  I'm expecting it to be slightly better than the E350.  I found one reviewer that said it idled at 26W with 6 WD Green drives, but this was under Windows (which never really idles) and the reviewer didn't clarify if the drives were spinning or in standby, so the actual idle with drives in standy could have been in the mid teens.

 

Another thought is that ASUS might not be the only game in town making E350/E450 motherboards with 6 SATA ports.

[b0ssman]

good news

 

according to german c't magazin there is this board.

http://www.tragant.de/produkt/33079/merkmale/1.html

which is managing 10,3W idle.

 

Unfortunatly they dont say which haswell cpu they used.

[Ford Prefect]

good news

 

according to german c't magazin there is this board.

http://www.tragant.de/produkt/33079/merkmale/1.html

which is managing 10,3W idle.

 

Unfortunatly they dont say which haswell cpu they used.

 

...as there are not that many CPU models launched for socket 1150 and as this is a desktop board (no ECC and no onboard GPU - like for IPMI), I gather all currenbt Haswell i5/i7 will idle the same.

Any hints if they used LV RAM, underclocking or a pico-PSU to achieve these numbers?

[b0ssman]

no unfortunatly not

[dmacias]

I think the performance difference is related to the 1TB platters in the newer 2TB drives, and I personally don't think the 3Gb/s connector speed was any impact for a spinning 5400rpm hard drive.  You need a fast SSD before 3Gb/s becomes a constraint.  I don't think a couple older 3Gb/s ports on the E35M1 would be a detractor for unRAID use.

 

That said, I looked and ASUS has newer E45M1 edition motherboards that have all 6 SATA ports as 6Gb/s... unfortunately one of those six is an external eSATA port.  I suppose you could always route the external cable back into the case, and use an adaptor to make it plug into an internal SATA drive - not elegant but might be worth it. 

 

So if I'm understanding you correctly dmacias, you have good results with the C60M1-i, unlike ZeroK?  Can either of your report your idle power consumption?  I'm expecting it to be slightly better than the E350.  I found one reviewer that said it idled at 26W with 6 WD Green drives, but this was under Windows (which never really idles) and the reviewer didn't clarify if the drives were spinning or in standby, so the actual idle with drives in standy could have been in the mid teens.

 

Another thought is that ASUS might not be the only game in town making E350/E450 motherboards with 6 SATA ports.

I agree the 5400 rpms couldn't even saturate the sata 3Gbps and the platter size made the difference. Also I don't really know the cache size on all the drives either. But it definately makes a difference with ssd's.  Iv'e been meaning to check the idle wattage for awhile.  I'll see what I can do.  I think, given the exact same system the c60 mb would be slightly lower watts than the e35. I think it would depend on power supply.  I have a Seasonic SS400-FL fanless with 5 2TB green drives, a ssd cache and 2-4GB Samsung green ddr3s.

 

I have been very happy with the c60. I don't write to the array much from other computers and use the mover script to move all downloaded content from cache drive. But even then I am able to steam content to a several computers. The web page can become slower when writing to the array and streaming and downloading. I have never felt like I needed more cpu power or had problems due to lower cpu power, except when I ran virtual box with a mythbuntu vm. But for a low power nas with sab, sick, mysql, xbmc, vpn, local webserver it does great.  All that said I would have used the e35 over the c60 if I had them at the same time.  Asus really seems to be the only one to have 6 sata in a cpu combo itx board and esp for under $100.  the e45 and e35 deluxe are closer to $200 and same for other 6 sata itx.

[cantharides999]

For your info I'm testing a brazos chipset right now:

 

I'm testing an ASUS E35M-I-deluxe right now and have the following measurements (no drives installed), combined with standard ATX 300 W power supply and one 4GB ram dimm.

 

Standard (no tweaking) cpu running at 1.6 GHz :

 

Powered off: 1W

Boot: 39W

Usage (memtest): 33W

idle: 29W

 

Underclocking (Cpu running at 1.4 Ghz):

idle: 28W (only one watt less).

 

Power consumption in idle standard is somewhat disapointing (had expected less then 25W at least), maybe due to a mobo with many bells and whistles), but under load never goes higher then the 39W at boot (that's not bad).

 

Undervolting haven't tried yet.

[b0ssman]

it depends heavily on the psu.

 

even if your psu is 80+ certified most of them are not efficient in the 10% load range.

[cantharides999]

Maybe for everyone who's interested in using brazos platform:

 

Also after a day or two testing with the E350 asus board there appear to be some minor issues in syslog, so I don't know if the chipset is reliable enough to use as an unraid server.

 

The reported minor problems in syslog are the following:

 

Jul 26 09:50:17 Tower kernel: ACPI BIOS Bug: Warning: Optional FADT field Pm2ControlBlock has zero address or length: 0x0000000000000000/0x1 (20130117/tbfadt-599) (Minor Issues)

Jul 26 09:50:17 Tower kernel: xor: measuring software checksum speed (Minor Issues)

Jul 26 09:50:17 Tower kernel: acpi PNP0A08:00: ignoring host bridge window [mem 0x000c8000-0x000dffff] (conflicts with Video ROM [mem 0x000c0000-0x000ce1ff]) (Minor Issues)

Jul 26 09:50:17 Tower kernel: acpi PNP0A08:00: ACPI _OSC support notification failed, disabling PCIe ASPM (Minor Issues)

Jul 26 09:50:17 Tower avahi-daemon[1343]: WARNING: No NSS support for mDNS detected, consider installing nss-mdns! (Minor Issues)

 

I don't know if this is related to the chipset or specific to the asus chipset and bios implementation.

 

Didn't cause any instability until now though.

 

 

[BobPhoenix]

Maybe for everyone who's interested in using brazos platform:

 

Also after a day or two testing with the E350 asus board there appear to be some minor issues in syslog, so I don't know if the chipset is reliable enough to use as an unraid server.

 

The reported minor problems in syslog are the following:

 

Jul 26 09:50:17 Tower kernel: ACPI BIOS Bug: Warning: Optional FADT field Pm2ControlBlock has zero address or length: 0x0000000000000000/0x1 (20130117/tbfadt-599) (Minor Issues)

Jul 26 09:50:17 Tower kernel: xor: measuring software checksum speed (Minor Issues)

Jul 26 09:50:17 Tower kernel: acpi PNP0A08:00: ignoring host bridge window [mem 0x000c8000-0x000dffff] (conflicts with Video ROM [mem 0x000c0000-0x000ce1ff]) (Minor Issues)

Jul 26 09:50:17 Tower kernel: acpi PNP0A08:00: ACPI _OSC support notification failed, disabling PCIe ASPM (Minor Issues)

Jul 26 09:50:17 Tower avahi-daemon[1343]: WARNING: No NSS support for mDNS detected, consider installing nss-mdns! (Minor Issues)

 

I don't know if this is related to the chipset or specific to the asus chipset and bios implementation.

 

Didn't cause any instability until now though.

The ACPI error could cause your server to not be able to sleep if you wanted to install a plugin to provide that capability but it may not affect anything else.  If you upgrade/downgrade your bios you may find that the ACPI error goes away.  The NSS warning is a new one that I've noticed on my unRAID server logs as well but I haven't noticed any problems associated with it.  Hopefully someone else can provide more complete answers for you.

[dmacias]

So if I'm understanding you correctly dmacias, you have good results with the C60M1-i, unlike ZeroK?  Can either of your report your idle power consumption?  I'm expecting it to be slightly better than the E350.  I found one reviewer that said it idled at 26W with 6 WD Green drives, but this was under Windows (which never really idles) and the reviewer didn't clarify if the drives were spinning or in standby, so the actual idle with drives in standy could have been in the mid teens.

I checked the power consumption today.

 

C60M1-i, 2 Green Samasung 4GB DDR3, 5 WD 2TB 6Gbps, 1 Patriot 60GB 6 Gbps SSD, 140mm and 120mm Silenx fans and Seasonic SS-400FL2 400W 80 Plus Platinum Fanless power supply.

 

Boot:  39W

Load:  31W  (serving a couple 1080p mkvs)

Idle:    26W

 

Plugins running but not doing anything Sabnzbd, Sickbeard, mySQL, utorrent, dropbox, openssh

I think the power supply is the limiting factor even though it gets good reviews at low loads.  Maybe a pico and separate power for the drives would do better.

[Pauven]

Thanks for the info dmacias!  That was very kind of you to provide.  Since you only have 5 spinners, you might be able to use a Pico for your whole system, including hard drives if they are eco units.  That should save you a good number of watts.

 

I would like to know the spinning state of the two fans.  Were these fans running full speed, idled to a low speed, or stopped completely?  A quick search on Silenx fans shows that both sizes consume about 1W at full speed.

 

I'm a little baffled by the total power consumption numbers.  They are good numbers, and they align with several end user reports I've read, but they do not align with X-Bit's numbers for the E-350, and I would expect the C-60 to beat the E-350.

 

Perhaps X-Bit's numbers are the outliers and should be ignored, or perhaps the E-350 miraculously idles lower than the C-60 based systems.  For now, it looks like the D525 based Atom systems are still the top dog in idle power efficiency.

[dmacias]

I have the bios set to silent and I might have tweaked them lower. I think I had to turn the alarm down because the rpms are so low.  So the fans were spinning but at low rpms. I plan on expanding so I don't think the pico will work.  I've been meaning to  recheck the bios and the power consumption again and I'll check the fan rpms from the command line when I do but someone's always streaming something.  I think it just comes down to power supply efficiency.

[dmacias]

I checked again.  It was still 26w.  However I remember the silenx fans had ceased and I had replaced them with Enermax silence.

http://www.frozencpu.com/products/11426/fan-734/Enermax_TB_Silence_120mm_x_25mm_Twister_Bearing_Fan_UCTB12_.html  The front is 140mm running @ 500rpms and top is 120mm running @ 600rpms. Full speed the 120 is 1.8w.