Update: I have raised this directly with EVGA support. My query was responded to by somebody who can read the manual, who was able to confirm that, as I already knew (because, you know, I can read the downloadable PDF manual, too):
 
1) X99 Classified hasn't been certified to work with the E5-2699 v3 CPU, so no guarantees of it working
2) No ECC memory of any type was ever certified to work with the X99 classified so even with a Xeon, no guarantees that will work
3) No 64-bit BAR support is listed in the manual
 
And, of course, no intention was even remotely insinuated in supporting any of the above in future BIOS versions.

Well, it sure seems that it isn't what EVGA motherboards are made for.

Call me old fashioned, but I expect a motherboard with socket X to support all CPUs designed to fit socket X, and to support all features that the CPUs for socket X supports (such as ECC RAM in case of Xeons).
 
And yes, I can confirm that life is full of disappointment when your expected standard of engineering excellence is stuck in the 20th century.

Cool GTX
Design limits go well beyond fall well short of the sum of the parts.

FTFY

Cool GTX
Are you interested in building a workstation computer or a server with ECC RAM & 18 cores? 

 
Obviously.
 

Cool GTX
What software will you be using: OS & programs ?

Linux, with a VM (or two), each with GPU passthrough (gaming capable), so 100% working and bug-free VT-d and VT-x are essential. You don't want to know just how broken those are and what kind of bodgy patches I had to write for Xen's hvmloader 2 years ago when I embarked upon a similar task with the SR-2 to make such a configuration work. Thankfully, relatively recently a feature was added to mainline QEMU that can be abused to work around the hardware bug in question more cleanly.
 
The other reason I am looking at options WRT replacing the SR-2 with something less buggy and less requiring of an enormous case is that it would be very handy (and cost effective) to be able to use a Xeon Phi for certain workloads (e.g. video transcoding), and those require working 64-bit BARs (PCIe memory apertures bigger than 4GB and thus implicitly mapped above the 4GB 32-bit boundary).
 
As for ECC requirements, since a few years back I refuse to use anything more advanced than a laptop without ECC memory (and I grudgingly accept using a non-ECC laptop due to the fact that there is no such thing as an ECC laptop that doesn't weigh as much as a desktop). Life is far, far too short for weeks-long OCCT/minerd and other memory I/O subsystem stability testing, especially when those even when they error out (eventually) don't tell you which physical DIMM is at fault. Combine that with the fact that there are 8-12 DIMMs in the machine (and marginal signal quality/interference issues may not manifest without all the DIMMs installed), and you are lining yourself up for a whole world of pain if you want verifiable stability and your time is worth more than $0/day.
 
So yes, a full fat server board is probably what I'll have to get. It's just that I also put extremely high value on good customer service and warranty, which is why I would have preferred to get an EVGA board (because in comparison, most other hardware manufacturers do awfully poorly in that department).

Thanks DaveB for giving clarity to this post
 
 
EVGA = great customer service and warranty & a forum where they make getting answers easy.

E5-2697 V3, E5-2698 V3, E5-2699 V3 work on Rampage 5 Extreme.
But with desktop motherboard I would try first with ASUS X99-E WS USB 3.1.
They advertised as Powerful Home Server motherboard and almost all Xeons are supported list.
More models than on Rampage 5 Extreme.
Except support for them she have and very nice options for multi GPU configuration x16 3-4 way SLI.
Linus installed 18 core Xeon on ASRock X99E-ITX/ac Mini ITX motherboard with only two memory slots.
 

They are not worse than GIGABYTE or MSI.
Only solution is to find seller who treat customers better as Amazon, I hear they allow to people back hardware if something not work as advertised.
If Xeon 18 core work on some desktop motherboards that's ASUS X99-E WS/USB 3.1, ASUS X99-E IPMI and ASRock X99 WS-E...
I think all of them support 128GB DDR4 ECC and RDIMM.
You can find list of supported modules on their sites... You can't find better today. 
 

OK, but for now no something as you want, models above are only you can find with X99 chipsets compatible with all Xeon models and server memories.
This is list of supported and there are other list with much more models but for desktop...
 
 
4x 16GB Samsung M393A2K40BB1-CRC is probably OK, DDR4 2400 ECC/REG...
https://www.compuram.de/e...k40bb1-crc,samsung.htm

Maybe for you is EVGA support good, but for me was worse example, I had better experience with ASUS.
EVGA can't make such good thing to force me to buy that ever again. 
Same was before few days on ASUS Forum guys from UK had problem with burned card, after few days he got a new.
If you can build own motherboard with X99 chipset to support anything as server board than OK. If not and if you want there are not 50 options.
EVGA can't be competitive in desktop segment, not to optimize Xeon and ECC for desktop motherboards... That would be tragedy probably.
Expensive tragedy for people. And than it's same if someone send you 5 boards and all of them had some flaws or you wait other company longer to send you second normal board.
 
 

I can confirm a Xeon E5-2699 v3 will work in a X99 Micro2 but I'm using standard desktop non-ECC memory. So I can't confirm ECC works.

RetardedTiming
I can confirm a Xeon E5-2699 v3 will work in a X99 Micro2 but I'm using standard desktop non-ECC memory. So I can't confirm ECC works.

Thanks for sharing that! I'm actually considering snagging one (albeit an Extra Spicy/QS version), but I am trying to hold off for EVGA's V4 support and 14 nm.

Aiwandia

EVGATech_DaveB 
We have no plans currently on BIOS updates designed to give full ECC support, I apologize for the inconvenience.
 Our boards are meant for gaming, competitive clocking and desktop apps.  The analogy that comes to mind is you have a decked out Mustang GT350 with every power adder you can safely add to pull a heavy load, while the Mustang has the torque this is NOT what it was designed for, sounds like you need an F350.

There you have it, EVGA SR-2 ECC support was a fluke result of unlocked Xeons.

Wait why? And why was it a fluke of unlocked Xeons, you say that as if there isn't Unlocked Xeons right now? Looks at the entire line up of Unlocked x99 Xeons. 
 
Over half of the Xeons in the 16** range are fully unlocked and support ECC. That includes the Sandy Es, Ivy Es, and Haswell Es. As well as the 8 cores, in all 3. 
 
The 10 and 14 cores may also in fact be unlocked but you cant buy them so there is that. 
 
I only say that as your post leads me to believe you think that the unlocked Xeons were a fluke? that isnt the case. 

Couple things I want to mention for what it's worth:
 

1. Server, workstation, gaming, office, etc -- these are all uses, they are not functions limited to application specific hardware; you can use anything for anything, the only difference is what you expect versus a resulting performance.  All Intel's processor designs for the same architecture are the same; Intel only changes the core and cache counts, the number of PCIe and RAM channels enabled, and all other supported features including ECC based on the target market.  For the most part, the performance per core and per clock is identical whether we're talking about a Pentium G3220 or a Xeon E5-2699 v3.  Except the higher count Xeons have the advantage of the larger pool of share cache even while running single-threaded versus the Pentium.  
2. Ultimately, support for a particular feature is more about marketing territory when referring to a brand:  In the case of EVGA, the SR-2/SR-X products were catered to "Prosumers", which were not regular professionals who would find use for ECC while not regular consumers who wouldn't find use of more than one CPU.  One could say that Xeon support in Intel's HEDT platform is catering exclusively to those that want more cores than the flagship i7 model.  Furthermore, the fact that EVGA caters to overclockers in terms of board design could be why you may not see official ECC support despite seeing Xeon support:  Overclocking generally destabilizes a system, while anyone that prefers ECC generally wants absolute stability and endurance.  
   
   That said, you can have your cake and eat it too by simply choosing another board maker that openly supports more feature combinations.  ASUS and Gigabyte are among the largest manufacturers in term of global sales, the chances are much higher they may have what you want.
3. Some (albeit a few) Xeons are unlocked because they share the microcode with those i7 models meant for Intel's HEDT platforms, they are of course the E5-1600 series.  Pre-Sandy Bridge, there were no unlocked Xeons, per se.  The key is that Intel had the clock generator externally integrated into the mainboard, which gave motherboard makers the freedom to decide which processors could overclock.  This was why EVGA's SR-2 worked so well to overclock those Gainstown and Gulftown pairs of Xeons.  But when the Sandy Bridge/Sandy Bridge-E generation debuted, Intel integrated the clock generator into the processor die, which meant they had control and prevented all processors but those with K/X labeling from overclocking.  Recently, and ironically, it was Supermicro that allowed non-K overclocking on their Skylake platforms using an i3 which was quickly supported by a number of other vendors, most notably ASRock's SkyOC-- and then Intel reacted and some of those vendors created new BIOS to take that option back.  The reason that was possible is that apparently Intel put a secondary clock generator in the platform which was able to override the primary, thus leading to Bclk overclocking of non-K like the old days.  It is unknown, mainly due to unavailable data, whether anyone has attempted to overclock Xeon E3's in the Skylake platform, same whether the circumstance exists on the -E platforms.  
4. From what little rumors I've found on the subject thus far (because I pay attention as if I'll buy it), Intel's Skylake-E might have a different socket than Skylake-EP, i.e. LGA2011 v4 versus LGA3467, so prosumers can't just stick the upcoming 28-core Xeons in a proverbial X190 board when they come.  So if you want an 18-core Haswell-EP Xeon (or a 20-core Broadwell-EP) in an X99, it is now or never.  Core i7's might be limited to a 10-core early on, perhaps Cannonlake-E could be a 12-core.  

 
 

Warboy

arestavo

Warboy
...They unofficially support Xeons. One of the reasons why my X79 Dark has a Xeon in it right now. Only problem is the memory dividers don't work correctly. So it's limited to 1600mhz DDR3. So yours might be limited to 2133mhz DDR4. But I'm not sure, I don't know if they fixed that for X99.

Really? My X79 Dark has a 1650V2 running at 4.5 GHZ with 32GB (8 sticks) of 2400MHz RAM running at 2133MHz with just a slight RAM voltage bump. Or were you referring only to the locked 2XXX series?

And both the X79 and X99 officially, not unofficially, support Xeons.All of them may not be on the QVL, but a lot of them are. Unless that's what you meant?

V1s encounter the problem. Also double check CPU-Z to see if it's really running.
 
Because V1s claim to be running 2400mhz in bios, but only 1600mhz in windows. This was also proven by SuperPI 32M numbers.

I have checked CPU-Z when I was doing a bit of testing using the pre-DX12 version of Rise of the Tomb Raider to see how much RAM speed affected minimum FPS. It most assuredly works on my 1650V2 and X79 Dark.