New Build - E3 1220v5 vs E3 1230v5

[Tekz]

Hi all,

I purchased a new build that I'm putting together right now - still waiting on a few parts to arrive. It contains a Xeon E3-1220v5 that has been removed from the original packaging and installed into the mobo. However, yesterday I realized that the 1220v5 does not have hyperthreading and only has 4 threads instead of the 1230v5's 8. I originally was planning to build with a 1230 but bumped the build down to save $50. I thought at the time that it was just a frequency drop, not any substantial change - and now I find I was wrong and should have paid more attention.

The server is meant to be a CIFS share server for this office as it's only function. My question is: Will the 1220v5 function well enough, or should I look to sell it as used for a markdown on eBay and pick up the 1230v5? I want to get this figured out before I start installing - don't want to run into issues down the road and have to redo everything. Altogether this system is running around $4500 with the disks included so I'm not concerned if I should be spending $50 more. Just a an oversight on my part.

I ran a few searches on the forums here and elsewhere but didn't find much about multithread performance.

Reference links:
1220v5: http://www.newegg.com/Product/Product.aspx?Item=9SIA24G3RH8300
1230v5 http://www.newegg.com/Product/Product.aspx?Item=N82E16819117613

Rest of build:
Case: 3u NORCO RPC-3216 http://www.newegg.com/Product/Product.aspx?Item=9SIA3913UN3435
Mobo: SuperMicro X11ssh-LN4F-O http://www.newegg.com/Product/Product.aspx?Item=N82E16813182996
PSU: Corsair CS550M 550W ATX12V http://www.newegg.com/Product/Product.aspx?Item=N82E16817139059
Memory: 4x Samsung M391A2K43BB1-CPB 16GB DDR4-2133 http://www.amazon.com/dp/B018CZWVVU/?tag=ozlp-20
SAS Expander: M1015 LSI 9220-8i PCI-E SAS RAID 46M0831 http://www.amazon.com/dp/B0034DMSO6/?tag=ozlp-20
OS storage: 2x Sandisk 16GB CZ43 flash drives
CPU Cooler: Noctua NH-L9x65

Drives:
5x WD Re SAS 2TB 7200RPM
6x Seagate Constellation 2TB SAS Enterprise (1 hotspare)

 

[Spearfoot]

FWIW, the CPU mark for the E3-1220 v5 @ 3.0GHz is 7191; for the E3-1230 v5 @ 3.4GHz it's 9366. Benchmarks should be taken with a grain of salt, but your CPU appears to have roughly 75% of the faster chip's performance.

Regarding cores & threading, my understanding is that the SAMBA/CIFS service isn't multi-threaded in FreeNAS, so it might not matter much that you'll only have half the logical cores you were expecting. Other factors enter it to it, of course: how much RAM you have (you seem to have plenty), how many users will be attaching to the server, how much traffic you anticipate they'll generate, etc., etc.

You'll probably do fine with the slower chip, but if it were me? I'd spring for the faster one...

 

[Dice]

The server is meant to be a CIFS share server for this office as it's only function.

How many users are in this office?
For only CIFS usage, Xeon will probably never move above 5-10% CPU usage, before the drive setup is becoming the bottleneck.

As some point of reference, on my i3-6100 and CIFS share, I've attempted to stress the hell out of the FreeNAS box by moving thousands of files around, doing copies, just stressing the system at 10-15 massive 'file requests at once' (which does not resonate as normal usage per user what so ever). At 16Gb RAM, I experienced ram deficiency where operations came to a crawling halt, before CPU usage ever touched 20%, not to mention the level of oversubscription on that single 1Gbit interface during testing... An i3-6100 is according to benchmarks roughly half a E3-1230v5 in terms of horsepower.
Though, I have no experience of corporate sized user bases.

If you've a SATAslot to spare, I'd consider getting a cheap SSD for boot device.

Cheers /

 

[Tekz]

How many users are in this office?
For only CIFS usage, Xeon will probably never move above 5-10% CPU usage, before the drive setup is becoming the bottleneck.

As some point of reference, on my i3-6100 and CIFS share, I've attempted to stress the hell out of the FreeNAS box by moving thousands of files around, doing copies, just stressing the system at 10-15 massive 'file requests at once' (which does not resonate as normal usage per user what so ever). At 16Gb RAM, I experienced ram deficiency where operations came to a crawling halt, before CPU usage ever touched 20%, not to mention the level of oversubscription on that single 1Gbit interface during testing... An i3-6100 is according to benchmarks roughly half a E3-1230v5 in terms of horsepower.
Though, I have no experience of corporate sized user bases.

If you've a SATAslot to spare, I'd consider getting a cheap SSD for boot device.

Cheers / Dice

Thanks for the input.

Drive setup is going to be RAID10. 15-ish users currently and will probably go up to 20 or so in the next few years. Right now for the boot device I'll be setting up those two USB flash drives in a mirror. We have multiple ESXi servers in our environment running on USB drives without issues as well so I don't any significant benefit from going to SSDs for the boot drive.

 

[Tekz]

FWIW, the CPU mark for the E3-1220 v5 @ 3.0GHz is 7191; for the E3-1230 v5 @ 3.4GHz it's 9366. Benchmarks should be taken with a grain of salt, but your CPU appears to have roughly 75% of the faster chip's performance.

Regarding cores & threading, my understanding is that the SAMBA/CIFS service isn't multi-threaded in FreeNAS, so it might not matter much that you'll only have half the logical cores you were expecting. Other factors enter it to it, of course: how much RAM you have (you seem to have plenty), how many users will be attaching to the server, how much traffic you anticipate they'll generate, etc., etc.

You'll probably do fine with the slower chip, but if it were me? I'd spring for the faster one... :)

So even if CIFS isn't multi-threaded, the performance would likely be better for the sole reason that the 1230v5 has a higher base clock by 400MHz and higher turbo of 300MHz...

 

[Dice]

We have multiple ESXi servers in our environment running on USB drives without issues as well so I don't any significant benefit from going to SSDs for the boot drive.

I don't think it's fair to compare ESXi host running off USB to FreeNAS running off USB.

USB's have been recommended in the past, but there are reasons for this recommendation to 'turned to SSD'.
- Since a while back the boot drives on FreeNAS are running ZFS. Which has the implication of significantly higher wear on the device.
- Dropping SSD prices with immensely increased relaibility compared to USBsticks
- USB sticks not capable of sustain R/W required by ZFS leading to shortened lifespans.

Cheers /

 

[joeschmuck]

My two cents...

1) The CPU differences will not make a difference for what you have specified the purpose of your FreeNAS will be used for, and you have lots of overhead for many other things if you should want that in the future.
2) You have the RAM you need which will make a huge difference in performance.
3) Your bottleneck will be the way you configure your drives and/or the network connectivity, not the CPU.

Now for the boot device... Since this is for an office then I'd advise you to use a SSD as the boot device simply due to reliability. USB Flash drives die and SSDs rarely die a premature death. If you must go with a USB flash drive, get a good USB 2.0 version. USB 3.0 versions tend to overheat, but if you have a few ESXi servers running off of USB flash drives, you may already have many that are reliable to use. And always make a backup of your configuration file once you have the system configured, it will make restoring it easy.

 

[Tekz]

Ok, you've all convinced me. I've ordered a couple small SSDs for the mirror with the hope that they're more reliable than the USB drives I already picked up. All the parts should be here by Tuesday - once they get here and I've finished running burn-in tests I'll double-check the CPU usage under load.

 

[Bidule0hm]

Even one SSD is better than two USB sticks ;)

 

[joeschmuck]

once they get here and I've finished running burn-in tests I'll double-check the CPU usage under load.

Curious what you will consider an appropriate CPU load test. CIFS won't do much to your CPU if that was your test, not using a 1Gb Ethernet port. Maybe if you had a few 10Gb ports all running you could make the CPU jump. That's what I figure you will need to see anything appreciable. I'm not saying you won't see maybe up to 30% usage (likely 20%) but I doubt you will hit 50% using CIFS.

Also, do not encrypt your hard drives, not unless you are certain you have an idea of the pain you could be in for. This has nothing to do with CPU utilization, it's more about drive replacement, it can be a nightmare if you don't have your ducks in a row.

 

[Tekz]

Curious what you will consider an appropriate CPU load test. CIFS won't do much to your CPU if that was your test, not using a 1Gb Ethernet port. Maybe if you had a few 10Gb ports all running you could make the CPU jump. That's what I figure you will need to see anything appreciable. I'm not saying you won't see maybe up to 30% usage (likely 20%) but I doubt you will hit 50% using CIFS.

Also, do not encrypt your hard drives, not unless you are certain you have an idea of the pain you could be in for. This has nothing to do with CPU utilization, it's more about drive replacement, it can be a nightmare if you don't have your ducks in a row.

Basically just don't want to see the CPU bottlenecking anything/too high usage/too high temps due to usage. Wasn't planning to encrypt, so no problems there.

 

[joeschmuck]

The stock CPU fan will keep the CPU cool and it's quiet too. Of course your case layout will affect cooling as well. Try to keep the drive temps below 40C, meaning have proper air flow in the case.