Looking for advice for low-power consumption build

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

maxermaxer

Active Member
Oct 28, 2016
289
48
28
49
I've got a Supermicro X10DRI-T server motherboard and I would like to build a low-power consumption server with this board. I know some would say that go with ATOM or Xeon-D but before I do that I would like to know what is the best I can do to build a low-power-consumption power with this MB.

The lowest TDP E5 V4 Xeon I can find is Xeon E5-2630L-V4. If I just build a basic server with just adding 8-16GB RAM and one 128GB SSD drive to the board without anything else what is the best I can expect for power consumption?
 

acquacow

Well-Known Member
Feb 15, 2017
787
439
63
42
Make sure you enable C-states, speed step, and max power savings in the bios for extra savings. Enabling C6/7 states can drop idle core voltage, which is very helpful.
 

Evan

Well-Known Member
Jan 6, 2016
3,346
598
113
If the server is idle the latest e5's don't use too much power, at 100% CPU of course a lot more than say Xeon-D but they also do more work for the power consumed :)
 

Rand__

Well-Known Member
Mar 6, 2014
6,634
1,767
113
My Xeon-D 1528 idled around 50W (which I found a lot for a 35W TDP box tbh, my E3 & E5 boxes went up to 70/90 Watts.

-A Dual board is not really focused on low power consumption;)
-You have the -T variant - 10GBe uses more power than 1GBe and you can't turn it off as you only have the two 10g nics); you could add a 1g nic
-Use as few ram modules as possible (even if DDR4 uses less power it adds up)
-Make sure to get an adequately sized PSU (either small/pico or Platinum so it will have high efficiency at low load)
-Turn off ipmi if you dont need it.
-Usage of L CPUs will not decrease power utilization, this primarily hard limits the heat generated for thermal dissipation considerations (1U box). Power utilization depends on workload and efficiency - in this chart the 2650L is the most efficient cpu (but 30L is not bad either, but as efficient as the 2699;)) Note: This is based on TDP which is not a really good indication of actual power consumption.


=
https://www.servethehome.com/wp-con...Xeon-E5-2600-V4-Family-Comparison-800x570.png


Also see a Performance per Watt benchmarks here, they are comparing 2609v4 (85W TDP) against 2687W v3 (160W TDP) and the 2687 is way better (10,2 MFLOPs/Watt vs 15,8 MFLOPs/W) (and not as good as a Core i7;) )
 
Last edited:

Patrick

Administrator
Staff member
Dec 21, 2010
12,513
5,805
113
Also see a Performance per Watt benchmarks here, they are comparing 2609v4 (85W TDP) against 2687W v3 (160W TDP) and the 2687 is way better (10,2 MFLOPs/Watt vs 15,8 MFLOPs/W) (and not as good as a Core i7;) )
That is a fairly bad comparison set. The E5-2609 is basically a SKU to allow a LGA2011-3 system to be turned on. The E5-2687W V3 is a processor to do work.

Also, remember that when it comes to system power consumption, there are many other components that are going to mean higher power CPUs will be a higher percentage of overall power consumption. That increases efficiency.
 
  • Like
Reactions: Rocknrolla

PigLover

Moderator
Jan 26, 2011
3,186
1,545
113
Regarding the 'L' (and 'T') cpu - the 'L' version will idle using almost exactly the same power as the 'normal' CPU. The 'Low Power' part of these CPUs is about limiting their max power consumption under load by running at lower clocks and lower voltage.

If your concern is about idle don't waste extra money getting the special version.

Sent from my SM-G950U using Tapatalk
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
@PigLover the "L" seem to idle a bit less due to decreased frequency from the get go, so I guess if 5-14w matter it could be slightly beneficial depending on the exact chip and if it's 1 or 2 (L56xx generation for example).
 

maxermaxer

Active Member
Oct 28, 2016
289
48
28
49
Thanks all! Guess it would be better to sell the board out and then buy Xeon-D ones. :)
 

maxermaxer

Active Member
Oct 28, 2016
289
48
28
49
My Xeon-D 1528 idled around 50W (which I found a lot for a 35W TDP box tbh, my E3 & E5 boxes went up to 70/90 Watts.

-A Dual board is not really focused on low power consumption;)
-You have the -T variant - 10GBe uses more power than 1GBe and you can't turn it off as you only have the two 10g nics); you could add a 1g nic
-Use as few ram modules as possible (even if DDR4 uses less power it adds up)
-Make sure to get an adequately sized PSU (either small/pico or Platinum so it will have high efficiency at low load)
-Turn off ipmi if you dont need it.
-Usage of L CPUs will not decrease power utilization, this primarily hard limits the heat generated for thermal dissipation considerations (1U box). Power utilization depends on workload and efficiency - in this chart the 2650L is the most efficient cpu (but 30L is not bad either, but as efficient as the 2699;)) Note: This is based on TDP which is not a really good indication of actual power consumption.


=
https://www.servethehome.com/wp-con...Xeon-E5-2600-V4-Family-Comparison-800x570.png


Also see a Performance per Watt benchmarks here, they are comparing 2609v4 (85W TDP) against 2687W v3 (160W TDP) and the 2687 is way better (10,2 MFLOPs/Watt vs 15,8 MFLOPs/W) (and not as good as a Core i7;) )
The chart is very helpful! Saved to my iPhone for future use. ;-)
 

Rand__

Well-Known Member
Mar 6, 2014
6,634
1,767
113
That is a fairly bad comparison set. The E5-2609 is basically a SKU to allow a LGA2011-3 system to be turned on. The E5-2687W V3 is a processor to do work.
I know its apples to oranges, but it was the only one I could find.
Would be a niece piece to find out what really is most efficient per watt/dollar cross product line wise (d,e3,e5,core,ryzen) at various relevant performance levels ;)

Thanks all! Guess it would be better to sell the board out and then buy Xeon-D ones. :)
Not necessarily :) Totally depends on what you want to do with it.
Xeon-D is expensive and it will take a while to recuperate with power savings alone (depending on power costs). Additionally its limited in terms of expansion capabilities (usually 1 pci slot only).
 
Last edited:

britinpdx

Active Member
Feb 8, 2013
367
184
43
Portland OR
I've just transferred my X10DRi motherboard into an Intel P4216 chassis. The power supply is an Intel 750W (IPMI View reports it as DPS-750XB A)

The motherboard is the "non -T" variant, and has 2x i350 1GBe NIC rather than 2x X540 10GBe NIC.

The rest of the configuration is:
2x E5-2658v3 ES
2x Supermicro SNK-P0050AP4 CPU Coolers
8x 8GB Micron DDR4-2133 memory
1x Sandisk 120GB SSD
1x 120mm chassis fan is connected

With system idle in Windows 8.1:
CPU Core Voltage 0.685v
Kill-A-Watt reports 65W
IPMIView PSU Input power reports 63W

Running AIDA64 Stress Test:
Kill-A-Watt reports 298W
IPMIView PSU Input power reports 304W

I only ran the stress test for a few mins, but CPU temps were starting to normalize at ~50C max.



Here's the reported power draw from IPMI Web page showing idle and "stress test" power draws ..

Cinebench R15 reports a score of 2832 ..



Best of all, this build is very quiet. CPU fans run at under 2000 rpm and the chassis fan runs at ~800rpm.

I consider this a very reasonable blend of idle power and available performance (application specific of course)
 
  • Like
Reactions: Reflexiony and Evan

zer0sum

Well-Known Member
Mar 8, 2013
849
474
63
I've just transferred my X10DRi motherboard into an Intel P4216 chassis. The power supply is an Intel 750W (IPMI View reports it as DPS-750XB A)

The motherboard is the "non -T" variant, and has 2x i350 1GBe NIC rather than 2x X540 10GBe NIC.

The rest of the configuration is:
2x E5-2658v3 ES
2x Supermicro SNK-P0050AP4 CPU Coolers
8x 8GB Micron DDR4-2133 memory
1x Sandisk 120GB SSD
1x 120mm chassis fan is connected

With system idle in Windows 8.1:
CPU Core Voltage 0.685v
Kill-A-Watt reports 65W
IPMIView PSU Input power reports 63W
That is a great data point and it's interesting how low the idle power consumption is on that dual CPU setup :)

For reference...I have a 1230v2 system running ESXi 6.5 with 1 x Virtual machine - XPENology with plex/sabnzbd/radarr/sickrage/nextcloud/etc
It is "idling" at around 60-70W with all of the following:
  • SuperMicro X9SCM-F
  • Corsair HX850 PSU
  • 2 x LAN and 1 x IPMI all active and plugged in
  • E3-1230v2 @ stock clocks
  • 4 x 8Gb ram
  • 2 x SSD's
  • 4 x 3TB Hitachi HDD's
  • Dell Perc H310
  • 2 x 140mm case fans
  • 1 x Hyper212+ fan
  • 1 x small 40mm fan on the H310
 

Evan

Well-Known Member
Jan 6, 2016
3,346
598
113
@britinpdx thats an impressively low idle usage for a dual e5 for sure but does show it can be done and a single should be a good deal less.
 

acquacow

Well-Known Member
Feb 15, 2017
787
439
63
42
You guys ruined me... I just bought an E5 2648L (70W) to replace my 2670 (115W)

I already had the 2670 complete system idling at 100W with 5 HDDs, 4 ioDrives and 8 DIMMs installed, now I'm gonna shave some more =P

I repeat: =P