Low Power Server G2120 on SM X9SCM vs Atom D510 SM X7SPA

[tr0910]

Someone suggested that the G2120 would have a lower base idle power use on a X9SCM than an X7SPA would.

 

http://lime-technology.com/forum/index.php?topic=5133.msg245475#msg245475

 

Since the X7SCA of GaryCase is able to hit 20W with all drives spun down, this sounds very ambitious.

 

Has anyone tried this yet?

[garycase]

My system uses a X7SPA-H-D525-O

 

... not an X7SCA  [Actually I don't think there's a motherboard with that designation, but there is an X7SBA, which is a Socket 775 board - not the Atom board I'm using]

 

 

But I doubt seriously that any board with a G2120 will idle with anywhere near as low a power draw as my Atom does 

[tr0910]

But I doubt seriously that any board with a G1220 will idle with anywhere near as low a power draw as my Atom does 

 

Yes, perhaps I wasn't understanding the post correctly...

 

Another typo.  Meant G2120 not 1220.  Sorry...

[Ford Prefect]

This comparison was based on my personal measurements with the X7SPA-HF, as I can only get it to 23-25W.

The difference is marginal, I think and might be related to meter-model and PSU quality.

Gary and myself discussed this in another thread.

 

The G *will* idle lower as the Atom, as the Atom has de-facto no power management at all...it will run low...and idle low.

But in combination with mobo, chipset, RAM, PSU and meter-model YMMV.

Here is a post in a german forum, where the OP managed to tweak the combo of the ASRock Pro3-M with an i7-3770 down to 18.5W. http://www.meisterkuehler.de/forum/vorstellung-stromspar-pcs/33379-vorstellung-mini-power-cruncher.html#post502258

 

Edit:  ...based on this http://www.servethehome.com/intel-pentium-g2120-review-benchmarks/  it runs at 26W idle....but this was with 32GB and inside a Norco (where I can believe that the fans were all off for this measurement  ).

[garycase]

I think the difference in the 20w I measure, and the 23-25 Ford Prefect measures are probably due to 4 potential factors:

 

(1)  Probably the most significant factor is that I use the "H" board, he uses the "HF" board, with integrated IPMI.    While IPMI is nice, and doesn't use MUCH power, it likely does add 2-3 watts

 

(2)  Our memory modules could be slightly different -- perhaps accounting for a watt or so

 

(3)  We have different power supplies, which can easily account for a watt or two of difference due to differences in PSU efficiency at these very low draws.

 

(4)  Obviously the meters may be slightly different.  I use a Kill-a-Watt model rated at  0.2% Accuracy, so it SHOULD be a very accurate measurement [e.g. 0.2% of 20watts = 0.04 watts] ... but there's always a chance it's a bit out-of-spec [although I've got 3 Kill-a-Watts, and they all measure the same, so I'm actually fairly confident in this measurement.

 

Bottom line:  I suspect the majority of difference in our measurements is simply the power needed for the IPMI chip.

 

[garycase]

Edit:  ...based on this http://www.servethehome.com/intel-pentium-g2120-review-benchmarks/  it runs at 26W idle....but this was with 32GB and inside a Norco (where I can believe that the fans were all off for this measurement  ).

 

That is indeed a very impressive idle consumption.  AND it supports ECC memory 

This chip, in an Asus mini-ITX board and a Lian-Li PC-Q25B case would make a REALLY nice low-power SMALL server with ECC RAM and up to 20TB of storage (with 4TB drives).

 

And while the Atom D525 could do the same at a bit less power draw, it wouldn't have ECC RAM, and wouldn't have the "horsepower" for additional add-ons that the G2120 does.    I may have a new "favorite" low-power build !!

 

Edit:  Never mind;  the Asus board doesn't support ECC RAM ... you'd need a Supermicro board for that; and I don't think they have a mini-ITX board that provides that support.