BUILD Intel S1200V3RPL vs Supermicro X10SLH-F

[kroko]

Ok. Then we can agree that argumentation choosing Intel over Supermicro or vice versa would not be IPMI based. :)
As this thread has not outlined any other "differences", then presumably for anyone else interested - choose whichever board makes you feel better subjectively, as no objective difference that can be attributed to FreeNAS performance can be found and they cost basically the same, the difference is based on wether you are in US or EU. Except for Intel not chocking on easily obtainable Kingston RAMs, while Supermicro has lots of warnings and List A/B by @Ericloewe .

 

[marbus90]

Intel boards would choke as well if you give them crap food (=bad Kingston DIMMs with swapped chips) -> go Samsung or Micron/Crucial. Supermicro boards are easily obtainable in US and EU.

however most Intel boards require an addon card for IPMI to work, that's integrated in Supermicro boards.

 

[kroko]

Sure, any board will choke on bad RAM. IMHO, Kingston ECC value RAM should not be considered a bad RAM. And it should't become flagged as such because Supermocros have issues with it. It is just plan specific board - RAM incompatibility thing. Think of it - would you flag some RAM as bad that performs well across all market per se, just because some boards has issues with it? And any supplier may have bad bunches, probably not allowable for high end server RAMs, but it happens.
Yes, Intel boards require addon card if IPMI desired. Without this addon card S1200V3RPL costs ~36 EUR+21%VAT (44 EUR inc VAT) less than X10SLH-F here in Latvia from certified dealers with warranty (no ebay thing). And adding AXXRMM4LITE will cost you exactly this difference. I chose a bit more expensive AXXRMM4 which gives dedicated 3rd NIC.
If it sounds I'm trying to convince Intel is better - I'm not. Just publicly logging my reasoning and experience with building this on top of Intel board. Isn't what this forum is about? :) See how it goes.

 

[danb35]

uhmm, pretty much anything you'd want to do from a phone can be done from Supermicro's IPMI web interface...

SuperMicro's iKVM requires Java, which doesn't run on iPhones/iPads. The rest of the monitoring, power control, etc. could likely be done through the web interface, but you wouldn't have the remote console.

Since I have two SuperMicro-based servers, and I run OSX as my primary client OS, my experience might be relevant. I have not yet been able to use the web-based remote console under OS X, whether with Google Chrome, Firefox, or Safari. The IPMIView application works without great difficulties. The iKVM Java app, launched from the IPMIView application, works fine once it gets started, but that sometimes takes a few tries. However, there's a significant difference between the iKVM launched from IPMIView and the iKVM available through the IPMI web interface: the latter lets you mount a local disk image as a virtual disk to the server, while the former requires a network path to the image file. The result is that when I need to boot an ISO on one of my servers, I use Parallels on my Mac to fire up Windows 7, load up Firefox in that VM, and use it to connect to the IPMI web interface. Fairly cumbersome, really, but fortunately I don't need to do it very often.

 

[marbus90]

the SLH is a waste of money, you don't need the C226 chipset. supermicro have dedicated RJ45 ports for IPMI as well.

 

[Bidule0hm]

IMHO, Kingston ECC value RAM should not be considered a bad RAM. And it should't become flagged as such because Supermocros have issues with it.

The problem is that Kingston changed the RAM chips model without changing the stick model so you effectively have two different sticks of RAM with the same part number, one compatible and one not compatible. You'll tell it's not 100 % the fault of Kingston?

Plus with the SSD and USB sticks problems I just don't trust Kingston at all anymore (note that a few years ago I would recommend Kingston products...).

 

[mjws00]

I have the Intel S1200rps which is the cheaper version without the rmm option. I also have the x10sl7. I even have the very unpopular Kingston RAM. All are completely rock solid and suffer none of the hassles often discussed. Both boards will allows uptime measured in years. Intel is easier for me to get local vendor support (stock in hand) on so I picked one up before grabbing a Supermicro. Same with Kingston RAM, always in stock minutes away.

Yes IPMI is awesome, yadda yadda. But frankly I might log into it once a year. I can get up and touch a power switch that many times. Different story in a datacenter, but home storage is typically accessible.

I have yet to see any downside to the Intel board at all. BUT will likely only buy the Supermicro going forward. There are just more options and configurations available with slightly better bang for the buck, imho. At some point I may throw AsRock a bone as well simply for first hand knowledge. Also interested in seeing who puts together the nicest Broadwell-DE packages.

Two bits from one who has to balance lack of local availability and owns em all. If SM was on the shelf here all the time. Instant pick. But it is not. Same for Samsung/Hynix/Micron RAM.

Chances are you'll love the Intel rig. Truth is you stick it in a box and in most cases forget about it. It just works.

 

[Z300M]

Sure, any board will choke on bad RAM. IMHO, Kingston ECC value RAM should not be considered a bad RAM. And it should't become flagged as such because Supermocros have issues with it. It is just plan specific board - RAM incompatibility thing. Think of it - would you flag some RAM as bad that performs well across all market per se, just because some boards has issues with it? And any supplier may have bad bunches, probably not allowable for high end server RAMs, but it happens.

So why did Kingston even discontinue their "Elpida F" modules altogether? If it was simply a matter of incompatibility with (some?) Supermicro motherboards, they would have kept producing them for use in other motherboards. Why did months go by when Kingston had no 8GB modules at all that they recommended for many of the Supermicro X10 motherboards? And what about their alleged substitution of slower chips in their SSDs after they'd received glowing reviews?

 

[Whattteva]

And what about their alleged substitution of slower chips in their SSDs after they'd received glowing reviews?

Kingston is not alone on that one. PNY also got caught red-handed doing that. Bait-and-switch seems to be the norm in the SSD industry.

 

[marbus90]

sure, Intel, Crucial, Micron and Samsung do that all the time. As do all of them but Intel with DIMMs. warning, irony...

 

[kroko]

Put together yesterday. Network data copy speeds out of box.

From OSX 10.10 via AFP (zero)
dd
1048576000 bytes transferred in 9.890715 secs (106016197 bytes/sec)
106.016 MB/s

From OSX 10.10 via AFP (real file)
rsync
sent 1.07G bytes received 42 bytes 69.28M bytes/sec
total size is 1.07G speedup is 1.00

From OSX 10.10 via SMB (zero)
dd
1048576000 bytes transferred in 9.686584 secs (108250339 bytes/sec)
108.25 MB/s

From OSX 10.10 via SMB (real file)
rsync
sent 1.07G bytes received 42 bytes 74.06M bytes/sec
total size is 1.07G speedup is 1.00

Form Ubuntu 14.04 via NFS3 (zero)
dd
2097152000 bytes (2.1 GB) copied, 18.851 s, 111 MB/s

Form Ubuntu 14.04 via NFS3 (real file)
dd
1073741824 bytes (1.1 GB) copied, 11.1947 s, 95.9 MB/s

NFS is mounted with params
nfsvers=3,resvport,async,soft,rsize=32768,wsize=32768,timeo=900,retrans=3

i do not see any difference between "resvport" and "proto=tcp,port=2049". this ubuntu will have only one NFS mount, so all ok with one priviledged port. "intr" is depreciated as of kernel 2.6.25. (Ubuntu runs on 3.16.0).