I currently have a Xeon e3 1230v2 file server, and was wondering if there was a lower power replacement for it? I need at least 4 PCI-E slots for raid controllers, HCAs and quad port nics. Right now it pulls 170 watts with 10 HDs, 8 SSDs, LSI 9271, dual port nic, a CX-3 card and 32 Gbs of ram (it's also a hyper-v server also) I know that the xeon is pretty power efficient, but is there a more power efficent platform to put it on that is reasonably priced?
Low power File server replacement?
- Thread starter Darkytoo
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
I would be interested in any insight on this as well. Or possible power optimization options for a x9scm and e3-1230 v2 at idle. Seems like 170watts isn't all that much with 18 HDD/ssds + 32gb ram + 4 pcie cards. Idle on the chip is like 32 watts, each hard drive is like 3-5 watts, each pcie card is +/- 15 watts, and fully populated ram requires some amount of power. So I am wondering if messing with c state power,pcm or fan speed, pcie power management, and spinning down drives could get this similar set up idling at 100 or so watts?. My interest in this is for a three node clustered storage spaces set up that I would like to keep up and running without having to get a second job to pay for the power bill 
.
Beyond optimization if you are looking at new gear possibly look at asrock e3c224d4m-16re and the e3-1230v3 slightly less power proc, sas3 hba with expander for up to 22 hard drives 18 at data 3. 2x PCI 3.0 with a mezzanine slot for 10gbe. Or if you can get away with it possibly a c2750/2758 from supermicro or asrock with LSI hba built in and a SAS expander will run lower power but may not be as performant as the e3 Xeon platform.
. Beyond optimization if you are looking at new gear possibly look at asrock e3c224d4m-16re and the e3-1230v3 slightly less power proc, sas3 hba with expander for up to 22 hard drives 18 at data 3. 2x PCI 3.0 with a mezzanine slot for 10gbe. Or if you can get away with it possibly a c2750/2758 from supermicro or asrock with LSI hba built in and a SAS expander will run lower power but may not be as performant as the e3 Xeon platform.
I'm with Patrick here. There isn't much more you are going to do to get cup power down. By your own summary here, the system is running 170 watts idle, 32 of which is the cpu...actually, I'd guess about 5 of which is the cpu and 27 is the support chips around it, memory, etc.
If you got the actual cpu idle to 0 watts I doubt you'd be much happier.
Most of your power is in the drives. The real trick is to get the drives to stop spinning when idle. Unfortunately this is currently not possible in most self-configured file servers (e.g., LSI controllers won't pass spindown, zfs configs for it are tricky, Storage Spaces won't do spindown, etc.).
If you got the actual cpu idle to 0 watts I doubt you'd be much happier.
Most of your power is in the drives. The real trick is to get the drives to stop spinning when idle. Unfortunately this is currently not possible in most self-configured file servers (e.g., LSI controllers won't pass spindown, zfs configs for it are tricky, Storage Spaces won't do spindown, etc.).
Running Hyper v on Windows 8.1 here. Host would go to sleep when not home, and spin down 6x4 TB disk drives just like a desktop machine. Wake on Lan when needed.
Best of both world ( server and desktop ).
My electric bill is around $50 for 3 people, I lived in the city as PigLover and serve by the same power company (PG&E) as Patrick.
Forgot to mention, there is no idle power consumption difference between running Windows 8.1 vs Windows 8.1 with hyper-v role ( without any running VM ).
Last edited:
That was sort of my suspicion. I looked and saw an e3 1220Lv2 that has a max TDP of 25 watt, but really the CPU i have now isn't using much more than that. I can't go atom because I need at least PCI-E slots. I looked that amd e350s but they only seem to have one PCIe slot, and at the very minimum i need 2.
As other have noted, The cpu is already very efficient here. Perhaps you could look around at your components to see where savings can be had. Fans chew up a lot of power. Can they be replaced to save some watts? Sometimes there are features you can disable on your MB to save power. Are there any features you could disable? Do your hard drives spin down ever? That will save power.
A little here and a little there...
A little here and a little there...
Thats sort of what i've been doing. I recently figured out my HP dual port Nics were using 12+ watts each when I can get quad port i350 nics that use 4.4 watt. I'm stuck with the CX3 card, stuck with the LSI 9271, stuck with the SAS expander. I'm checking the fans next, but those darn LSI cards make some serious heat.
The LSI card probably uses 10W alone, for what you have I think your power consumption is pretty good. Changing the CPU isn't going to make any difference, or if it does the minimal difference will take a long time to repay the price of the new chip through power savings.
Perhaps splitting the file serving roles and Hyper-V role could save? A NAS box with enough drives from the likes of Synology or Asustor could use less power, especially if you use big drives. Then you could try and run Hyper-V on a more power efficient machine (perhaps the old server minus the HBA/expanders?).
I used to run an all-in-one server, but to save power I split the roles, the Asustor does a lot of the 24/7 stuff like downloading and obviously the NAS duties. My workstation does Hyper-V stuff, but when it's not being used it sleeps/hibernates. I brought my power draw from 80W 24/7 to 20W for 18 hours a day, 25W for 3 hours and approx 100W for 3 hours.
Perhaps splitting the file serving roles and Hyper-V role could save? A NAS box with enough drives from the likes of Synology or Asustor could use less power, especially if you use big drives. Then you could try and run Hyper-V on a more power efficient machine (perhaps the old server minus the HBA/expanders?).
I used to run an all-in-one server, but to save power I split the roles, the Asustor does a lot of the 24/7 stuff like downloading and obviously the NAS duties. My workstation does Hyper-V stuff, but when it's not being used it sleeps/hibernates. I brought my power draw from 80W 24/7 to 20W for 18 hours a day, 25W for 3 hours and approx 100W for 3 hours.
Last edited:
I'm stuck with using it because I spent alot of money on the raid controller and all the cables, plus before it was an iSCSI target, but now it's storing all my VHDXs for my 2 node Hyper-V cluster. It also has a ConnectX-3 card for SAN the other cluster members use, so I think i'm pretty well stuck with it for now. It was fine being a combined Hyper-V & iSCSI target until now, when apparently when you use SMB storage instead of iSCSI, Virtual Machine Manager has issues managing everything.
I would love to split it off, but right now i've got my power draw for my lab down to approx. 420 watts, and for me to add another server, i'd really want to be able to balance the new server power draw with savings from the old server. I actually looked at just replacing the CPU, but that darn xeon is sooo efficient, a pentium or celeron has the same power draw or the power difference is so small it's not even worth the price.
Does anybody have experience with using an atom board for a file server? would something like this work?
Supermicro X7SLA L Motherboard 2GB RAM Atom CPU I O Shield Low Power Server 0672042047529 | eBay
is it even worth it?
A couple of data points based on my experiences. I think that the discussion so far has captured the main points, it's not necessarily the CPU that's puling the power, it"s likely to be all of the other components resulting in "death by 1000 cuts".
I'm fortunate enough to be able to house my gear in the basement, but my main workstation desk is also there so I've got to make it a quiet and cool environment to the best of my ability. I've tinkered with a lot of the Supermicro range of chassis and motherboards, and they are very modular and can be modded easily, and they seem to have become my "go to" grab bag. I've spent some time (too much really) swapping power supplies and fans into various chassis to find the best operational point focused on thermals and noise for my specific use cases.
My main 24/7 system is similar to your build and essentially is deployed as a file server with very light VM duty (no more than 2 VM at any time). For reference, build components are as follows ..
Just for grins, a couple of other data points ..
A SM836 chassis configured with the same SM PWS-563-1H psu as above, SM X9SCM-F, Xeon E3-1220 with active cooler, 2x 2GB Samsung PC3-10600E, Sandisk 240GB OS SSD after boot at login screen draws ~31W. This is the older 32nm Sandy Bridge E3, 4C4T but this shows that idle draw is minimal.
And, just arrived, a SM SYS-5018D-MTF 1U chassis, based on a 350W gold psu and x10SLM-F motherboard. This is for a basic low power 24/7 file server use case.
I dropped in a G3220 2C2T with 2x 2G Samsung PC3-1060oE and Sandisk 240GB OS SSD after boot at login screen draws ~17W. No doubt about it, the latest generation of Intel Silicon provides a very "green" base on which to build !
The next project will be to add 4x drives and give either FreeNAS or Xpenology a spin (both new to me), with the goal of a low power NAS.
I'm fortunate enough to be able to house my gear in the basement, but my main workstation desk is also there so I've got to make it a quiet and cool environment to the best of my ability. I've tinkered with a lot of the Supermicro range of chassis and motherboards, and they are very modular and can be modded easily, and they seem to have become my "go to" grab bag. I've spent some time (too much really) swapping power supplies and fans into various chassis to find the best operational point focused on thermals and noise for my specific use cases.
My main 24/7 system is similar to your build and essentially is deployed as a file server with very light VM duty (no more than 2 VM at any time). For reference, build components are as follows ..
- Chassis: SuperChassis 825TQ-563LPB.
That's a 2U chassis with a 560W Gold rated PSU, and the psu is whisper quiet. The 3x standard chassis fans are setup for passive cooling, but I prefer to use active cooling, so the FAN-0094L4 fans are swapped out for something that draws a little less current and run a little slower. I've got the Evercool EC8038HH12BP fans in there now (this was a tip from PigLover from the "taming the C6100" thread, which worked superbly on my C6100's). There's still plenty of airflow to keep my drives and cards cool. - Motherboard: Supermicro X9SCM-F
- CPU: Xeon E3-1230v2 with standard cooler
- Memory: 4x Samsung M391B1G73AH0-YH9 (PC3L-10600E, 1.35v)
- NIC: Intel Pro/1000 VT Quad Port (Dell YT674)
- Infiniband: Mellanox Connectx-2 MHQH29
- RAID: LSI 9261-8i
I have the 9261 flashed with the 0048 FW as I think its the last one that supported the "power save" feature to spin down the drives after 30 min. I have also learned to use the LSI "remote battery kit" to mount the battery off the board and allow overall better airflow/cooling for both the battery and the card. - OS SSD: Seagate ST240FN0021 240GB
- VM SSD: Crucial M500 240GB
- Data Store: 6x Toshiba 3TB DT01ACA, configured in RAID6
- After boot at login screen, ~103W
- Running Crystal Disk Mark to exercise the disks ~118W
- Running Prime95 to exercise the CPU (max heat method) ~184W
- Idle after 30min (the 9261 spins down the 6x 3TB drives) ~77W
Just for grins, a couple of other data points ..
A SM836 chassis configured with the same SM PWS-563-1H psu as above, SM X9SCM-F, Xeon E3-1220 with active cooler, 2x 2GB Samsung PC3-10600E, Sandisk 240GB OS SSD after boot at login screen draws ~31W. This is the older 32nm Sandy Bridge E3, 4C4T but this shows that idle draw is minimal.
And, just arrived, a SM SYS-5018D-MTF 1U chassis, based on a 350W gold psu and x10SLM-F motherboard. This is for a basic low power 24/7 file server use case.
I dropped in a G3220 2C2T with 2x 2G Samsung PC3-1060oE and Sandisk 240GB OS SSD after boot at login screen draws ~17W. No doubt about it, the latest generation of Intel Silicon provides a very "green" base on which to build !
The next project will be to add 4x drives and give either FreeNAS or Xpenology a spin (both new to me), with the goal of a low power NAS.
The majority of my drive failures over the years have been when spin down was enabled.
My server has 24 2TB Hitachis which I started purchasing shortly after they were released (took about a year to reach 24 drives). All of them are set to spin down, and to this day not one of these drives have ever failed in 5 years. I also have six 3TB 5400 RPM and six 3TB 7200 RPM Hitachis. Although these are set to spin down as well, they rarely do because of the data I have on them is frequently used. None of these drives have failed either.
My experience with Seagates and Western Digitals, unfortunately, is not the same. I'm quite familiar with their RMA process.
FWIW, all these Hitachis are hanging off an Areca 1882i controller in a Supermicro SC847 case with the SAS expander backplane (don't remember the exact model number).