X99 on mITX: ASRock X99E-itx/ac

I think it's just poor wording/translation...

The Narrow ILM brackets screw onto the heatsink base just as any of the other mounting systems do - the only hassle is going to be screwing the four screws into the mounts on the motherboard.

Have you talked to Noctua about it?
If not, could be worth sending them a e-mail, they've been very helpful when I've contacted them. ([email protected])
 
I think it's just poor wording/translation...

The Narrow ILM brackets screw onto the heatsink base just as any of the other mounting systems do - the only hassle is going to be screwing the four screws into the mounts on the motherboard.

Could be, but yeah it's worth confirming as the person right above me suggested.

Since Narrow-ILM is almost 100% servers it would make sense they would only produce narrow versions of their small server-compatible coolers. Packing it in with their enthusiast coolers wouldn't really sell.
 
I'd highly suggest using a noctua u9s heatsink with this, it's probably the most silent non water based cooler you can get that's silent and compact, and that actually keeps the system cool.

Good find, I'll have a closer look thanks.
 
I have a Xeon 2676 V3 12 core coming next week. Hopefully it works with this mobo!

Do report how this works! This CPU looks like a beast, 12 core, turbo to 3.0 GHz, and cheap! From a quick web search it looks like these are custom CPUs for Amazon.
 

Awesome! By any chance do you know with these strange models that aren't available on retail if they can be overclocked? I could imagine a situation where you might want to disable half of the cores and clock higher for better single threaded performance, but then keep all your cores enabled and stock clock for a multi-threaded kick.
 
Xeon E5 1xxx series chips (i.e. chips without QPI and destined for single-socket boards) can be overclocked.
Xeon E5 2xxx (with QPI, intended for dual-socket boards) cannot.
Xeon E7 CPUs cannot.
Xeon E3 CPUs cannot.
 
Xeon E5 1xxx series chips (i.e. chips without QPI and destined for single-socket boards) can be overclocked.
Xeon E5 2xxx (with QPI, intended for dual-socket boards) cannot.
Xeon E7 CPUs cannot.
Xeon E3 CPUs cannot.

what about these "engineering samples" that you see pop up on eBay?
 
The E5-2676v3 is presumably an ES chip. Hopefully you receive one of the later generation ones as the earlier steppings aren't always identical to the production silicon. Generally ES CPUs intended for dual CPU systems cannot be overclocked. I would also like to clarify that not all E5-16xx CPUs can be overclocked. The multiplier on my E5-1620v3 won't go past 36 (unless I'm doing something wrong?).
 
The E5-2676v3 is presumably an ES chip. Hopefully you receive one of the later generation ones as the earlier steppings aren't always identical to the production silicon. Generally ES CPUs intended for dual CPU systems cannot be overclocked. I would also like to clarify that not all E5-16xx CPUs can be overclocked. The multiplier on my E5-1620v3 won't go past 36 (unless I'm doing something wrong?).

The turbo of your CPU is 3600, so unless it's unlocked, I'm assuming that's the max multiplier?
 
Correct. Might only be the E5-1650 and up that are unlocked. I opted for the quad core one because it was cheap and I wanted an ITX system with 32GB of RAM and support for the Intel 750 SSD. Might need to upgrade after all.
 
Correct. Might only be the E5-1650 and up that are unlocked. I opted for the quad core one because it was cheap and I wanted an ITX system with 32GB of RAM and support for the Intel 750 SSD. Might need to upgrade after all.

Get an E5-2676v3. I got mine off eBay for 375 with "Make an offer". 12 cores, 24 threads, and 3.0 Max Turbo. Not as fast as yours in single threads though but it should be plenty for me.
 
Still need the single threaded performance for games and the likes unfortunately. I did manage to buy a E5-2699v3 on the cheap, but it was mislabeled and wound up being a 14 core CPU, so it didn't work out. Too bad none of the single socket ES CPUs ever appear since Intel has produced the higher core variants.
 
To help those of us decide between crucial 2133mhz ecc 16GB and corsair non ecc 2600 (which seems to need a painful bios update needing ddr4 dimm before post?) has anyone tried both to see the performance difference?
 
I'd generally hew on the side of ECC, purely for the stability benefits. 25 years ago, the BER just from cosmic rays was estimated at 1 error per 256MB per month, or 1.418×10^-15 errors per second per byte. For 16GB, that's about 2 errors per day. And that was for 250nm; smaller process scales are more vulnerable (as you need less energy input to flip a bit).
 
Without ECC I've had near 100% stability with my desktop the last couple of years. And if the few minor issues I've seen through the years weren't my own doing, it was some wonky app causing it. But I would probably also get ECC when my board would support it just for the hell of it. If it wasn't much more expensive.
 
To help those of us decide between crucial 2133mhz ecc 16GB and corsair non ecc 2600 (which seems to need a painful bios update needing ddr4 dimm before post?) has anyone tried both to see the performance difference?

To answer the question directly:

2666MHz non-ECC DDR4 will outperform 2133MHz ECC DDR4.

Depending on the timings the slower memory might have marginally faster "true" latency but it will still get beaten in benchmarks.

AFAIK ECC confers no actual performance benefits.
 
Unless your doing mission critical stuff (which gaming is not...), there's no point in going for ECC ram.

Clock vs clock it will perform worse.

So save some money and get faster ram :)
 
Clock-per-clock (and with the same timings), ECC and non-ECC will perform identically.
 
Worked like a charm! Thanks Blue Fox! Indeed, mine were non ECC 2x16GB corsairs (vengeance LPX). Machine is finally posting and booting :D

I have the same modules (of the 2666MHz variety).

I noticed ASRock has BIOS 1.40 up but don't explicitly mention improved RAM compatibility. I wonder if the beta 1.35 RAM compatibility is baked-in or whether it's from a different chain.
 
They seem to fork things a bit. The L1.35 firmware I was provided didn't have full PCIe bifurcation support for example, but P1.20E did. Hopefully they've merged the branches and have everything in a single version now. I also might be able to test 64GB DIMMs in it soon just for fun. I already verified 32GB DIMMs work fine.
 
They seem to fork things a bit. The L1.35 firmware I was provided didn't have full PCIe bifurcation support for example, but P1.20E did. Hopefully they've merged the branches and have everything in a single version now. I also might be able to test 64GB DIMMs in it soon just for fun. I already verified 32GB DIMMs work fine.

Yea, hopefully 1.40 has the PCI-E bifurcation support!
 
For those wondering if the Noctua Narrow ILM brackets can be used on their other heatsinks.... they CAN!

I used the bolts from the included Dynotron cooler as the Notcua ones are longer clash with the heatpipes, but otherwise they're a straight swap over and currently cooling my 5960X.

Excuse the poor quality phone pics, but here are some shots:


1.jpg

2.JPG

3.JPG

4.JPG
 
Amazing! How did you order the Narrow-ILM mounting kit? It's not in the order form on the Noctua website...
 
I just added the request to the order form to have the narrow ILM specific LGA2011 mounts - simple!
 
For those wondering if the Noctua Narrow ILM brackets can be used on their other heatsinks.... they CAN!

I used the bolts from the included Dynotron cooler as the Notcua ones are longer clash with the heatpipes, but otherwise they're a straight swap over and currently cooling my 5960X.

Excuse the poor quality phone pics, but here are some shots:

Which fan are you using on your C-14S?
 
Using the included Noctua but switching to a Noiseblocker PWM shortly
 
For those wondering if the Noctua Narrow ILM brackets can be used on their other heatsinks.... they CAN!

csd, you, sir, have just made my day! That's great news, thanks!

Was screwing it in as tough as we thought it might be? Did you have to use a special screwdriver, or were you able to reach the screws with a regular (straight-handled) one?

I'm putting in my order for this board and a 5820K immediately. Looking forward to seeing what kind of performance/temps/noise I can squeeze out of this setup. X99 in mini ITX, now with solid air cooling options! It's a niche within a niche, I know, but if somebody at Noctua or ASRock is reading this thread, you should send csd an award.

EDIT: Props to Xandr, too, who I believe first suggested the combo in this thread. Thanks!
 
Last edited:
So this is absolutely the only X99 mITX board huh? Does anyone know of any others in development? I'm surprised there aren't more coming to market with Intel aiming the next 6 and 8-core refresh at LGA2011-3 before the eventual socket changeover in mid-2016.

I'm really just looking for something Thunderbolt-capable, but I do realize it's somewhat of a tall order.
 
Even after the release of the ASRock board showed it's possible, Asus (and I think EVGA too) has said that it's not worth their time and effort since the market is so small.
 
csd, you, sir, have just made my day! That's great news, thanks!

Was screwing it in as tough as we thought it might be? Did you have to use a special screwdriver, or were you able to reach the screws with a regular (straight-handled) one?

I'm putting in my order for this board and a 5820K immediately. Looking forward to seeing what kind of performance/temps/noise I can squeeze out of this setup. X99 in mini ITX, now with solid air cooling options! It's a niche within a niche, I know, but if somebody at Noctua or ASRock is reading this thread, you should send csd an award.

EDIT: Props to Xandr, too, who I believe first suggested the combo in this thread. Thanks!

Glad I could help :)

Three of the screws were easy to do, the last was a bit more of a hassle - the one nearest the NIC's... it's probaby not as tight as can be, but seems clamped fairly well as the temps are fairly reasonable.

I've got a new PWM fan tonight so will see how OCing goes on this heatsink!
 
Even after the release of the ASRock board showed it's possible, Asus (and I think EVGA too) has said that it's not worth their time and effort since the market is so small.

Shame really... Asus would do a god job.

That being said, I am REALLY happy with the quality of the Arcock board so far and good on them for doing something outside the box. Much better thought through than my old EVGA X99 Micro too...
 
AquaComputer will also make you a Narrow ILM bracket for their coolers, if you ask nicely :)

Mine's just arrived and is awaiting the rest of my WCing parts:

IMG_6392.jpg
 
I am REALLY happy with the quality of the Arcock board so far and good on them for doing something outside the box.

Indeed, and I feel that we already have a markedly different picture of the board than when it was released. It reviewed well and garnered several awards, but at least my takeaway from many of the reviews was "this is a marvel of engineering, but the lack of cooling options make it difficult to work with and ultimately limit its appeal." A sampling:

Tom's Hardware: "We found several big-air coolers that provide enough clearance, but can’t be screwed down due to the unusual mounting pattern… The X99E-ITX/ac also requires a special CPU cooler though, and your few cooler upgrade options appear slightly overpriced for the money."

HardOCP: "The CPU socket takes up a huge amount of real estate. This necessitates the need for custom cooling as the mounting holes normally associated with this socket type would be spaced further out making this form factor impossible to use… While it is a great overclocker, you are likely going to have to be creative with your cooling configuration."

Tweaktown: "Boards with the extra pins in the socket can overclock the cache further; on the X99E-ITX/ac I was able to OC the cache to 4.4GHz, while the core was 4.7GHz. Cooling was the issue; I could have hit 4.5GHz if I could have used my watercooling."

Kitguru: "As with all SFF hardware, cooling presents a major challenge for the X99E-ITX/ac."

Either ASRock wasn't proactive enough about illuminating the available cooling options ahead of launch (Corsair AIOs! Noctua air coolers!), or they genuinely didn't know. The H100i and the NH-C14 are well represented on this forum, and they can both work with minor adjustments, hardly anything I would call a "major challenge."

I feel that dual-channel memory and the single PCI-E slot were logical compromises that people could live with (must, in the latter case), but going with narrow ILM was a gutsy move from ASRock. Maybe they deserve more credit. This could be the future. Maybe narrow ILM won't be "the server cooling option," but rather "the enthusiast SFF cooling option." Already on this thread, we've seen portable Xeon rendering stations, a 36-threaded encoding monster from Linus, and now an 8-core air-cooled overclockable beast. If this catches on, you can bet that Asus, Gigabyte, Noctua, Corsair, and everyone else will sit up and take notice.
 
Last edited:
Well, at least Tom's Hardware did illustrate back in June that Noctua NH-D15 was compatible with X99E-ITX with the Narrow ILM brackets that come with their DX series:
Noctua_NH-D15_installed.jpg


You can also deduce from this and the socket compatibility chart that the same set of brackets works for all of their coolers except L9i/L9a, L12, and the DO series. Noctua simply doesn't provide the Narrow ILM brackets free of charge to non-DX customers.
 
Back
Top