News ASRock Unveils the X299E-ITX/ac: Mini ITX + X299 + Quad-channel Memory

Mod edit:

Detailed overview of what we know about the X299E-ITX/ac thus far here: https://smallformfactor.net/news/asrock-x299e-itxac-little-monster-detailed

Original:

ASRock did it! Finally, there's an Intel HEDT platform motherboard with full quad-channel DDR4 memory. The new X299E-ITX/ac is for those who need up to 18 CPU cores and up to 64 GB of quad-channel DDR4 memory in their SFF machines for reasons. The board manages its limited PCB real-estate by going vertical. It features two riser cards, one with a few onboard controllers, and a pair of 32 Gb/s M.2 slots), and the other riser with SATA 6 Gb/s ports, a third M.2 slot, and the headers such as USB 3.1. The board draws power from 24-pin ATX and 8-pin EPS connectors, conditioning it for the LGA2066 CPU using a 7-phase VRM. The lone expansion slot is a PCI-Express 3.0 x16, memory is handled by four DDR4 SO-DIMM slots. Connectivity includes two Intel I219-V driven gigabit Ethernet interfaces, 802.11ac WLAN, and Bluetooth 4.1.

Source

 

[EdZ]

So I had been trying to figure out how the X299E-ITX/ac handles the triple M.2 with Kaby Lake-X since they only have x16 lanes off the CPU, yet the M.2 here are wired to the CPU so that wouldn't leave enough for the PCIe slot.

I was thinking maybe there's some kind of complicated switching going on where if a Kaby Lake-X is used the M.2 get wired to the chipset instead, but it's actually much simpler:

The X299E-ITX/ac does not support Kaby Lake-X. Problem solved

Welp, that makes CPU choices easier, I guess.

 

[QuantumBraced]

Good, that would have been a total waste of engineering, cost, etc. The only reason to buy Kaby Lake-X is for collectors' value as it will be the least sold Intel CPU in history. Probably <0.5% of X299 users will buy it, and quite possibly literally zero people would have considered it for this specific board. ASRock could have not even said anything and they would have gotten zero complaints.

 

[EdZ]

as it will be the least sold Intel CPU in history

Desktop Broadwell could probably give it a run for its money.

 

[jeshikat]

Good, that would have been a total waste of engineering, cost, etc.

Exactly, and I told them that.

Because if you want 6-8 cores but didn't need quad-channel memory and triple M.2 then you would get AM4. If you didn't even need at least a hexacore then you'd go 1151 or still AM4 or . So other than those hypothetical people who can't scrape together an additional $150 on a HEDT build to actually take advantage of the platform, no one will miss this.

And I find it endlessly amusing that (as far as I know) the only other motherboard hardcore enough that it can't support Kaby Lake-X is the flagship ASUS Rampage VI Extreme.

 

[Phuncz]

Desktop Broadwell could probably give it a run for its money.

I was looking for one to replace the i5-4670K in my previous main config, but those prices just didn't drop. Especially since they enabled VT-d which I was after at the time.

 

[BirdofPrey]

Man, Intel's new HEDT platform sure is a clusterfuck.
I wonder how much this debacle is coing to cost not only Intel, but also the motherboard manufacturers (you can be sure some compatibility issues will be blamed on the board maker rather than on Intel for their platform decisions and not actually giving board makers any time to come up with good products)

 

[QuantumBraced]

Desktop Broadwell could probably give it a run for its money.

Hey watch it, a proud 6800K owner here in none other than the X99E-ITX/ac. Or did you mean the 1150 chips?

 

[MarcParis]

Here is best review of skylake-x!!!

 

[Biowarejak]

Here is best review of skylake-x!!!

I would love to play that.

 

[VegetableStu]

Hey watch it, a proud 6800K owner here in none other than the X99E-ITX/ac. Or did you mean the 1150 chips?

He's serious. Go look up Broadwell, not Broadwell-E.

 

[jeshikat]

So apparently no one here was planning to use Kaby Lake-X anyway, so what chips are you all going to get for this board?

I'm thinking the octa-core 7820X since it can actually turbo 500MHz higher than the hexacore so that'd come in handy for single-threaded stuff like my CAD program.

 

[Kmpkt]

Eighteen core because who needs to eat, right?

 

[jeshikat]

I'd really like to see clocks on that. Having used a 14-core Xeon for the last few months that never seems to go above 2.5GHz despite supposedly having 3.0GHz max turbo, it's kinda rough unless ALL your workloads are heavily multi-threaded.

 

[EdZ]

So apparently no one here was planning to use Kaby Lake-X anyway, so what chips are you all going to get for this board?

I'm thinking the octa-core 7820X since it can actually turbo 500MHz higher than the hexacore so that'd come in handy for single-threaded stuff like my CAD program.

With KBL-X off the table, the 7820X and 7900X do look to be the highest stock-clocking chips. I'll probably wait for some benchmarking to see if the extra cache of the 7900X makes any appreciable difference in VR performance (avoiding cache misses adding latency), though I doubt it'd be close to $400 faster.

 

[|||]

With KBL-X off the table, the 7820X and 7900X do look to be the highest stock-clocking chips. I'll probably wait for some benchmarking to see if the extra cache of the 7900X makes any appreciable difference in VR performance (avoiding cache misses adding latency), though I doubt it'd be close to $400 faster.

Interestingly, I've seen estimates for the 6-core Coffee Lake i7 processor to have up to 12MB of L3 cache, more than the 6-core 7800X (8.25MB) & 8-core 7820X (11MB), but still shy of the 10-core 7900X (13.75MB).

 

[jeshikat]

though I doubt it'd be close to $400 faster.

Yeah, the price on the 7900X really bothers me, it's almost twice the price for just two more cores. Sure you could say it's partly justified because there's also more PCIe lanes, until you realize that 40+ PCIe 3.0 lanes will cost you $1000 this generation compared to $480 last gen!

I really hope Threadripper annihilates sales of anything higher than the 7820X and forces Intel to scrap their pricing structure and start over.

 

[QuantumBraced]

For the previous platform, the 6800K was the only one that was within consumer range, so that's what I got. This time around we also have the 7820X. That's what I would get if I was buying today. People are complaining about the 28 PCIe lanes, but it's also $400 cheaper, so I think that's a fair tradeoff.

It also makes sense because I think $600 is the upper limit of overkill-consumer hardware shopping, i.e. something a regular consumer/hardware enthusiast would get because it's awesome, not because they necessarily need it. And those people (myself included) don't need more than 28 lanes. When you step up to $1000+ CPUs, the only people buying are those who are making a very conscious investment into a workstation that they actually need, and those guys may need 44 lanes. Plus, for us it's perfect because we can't even use 28 lanes with this board anyway. Overall, I am pretty happy with X299 -- yeah Kaby Lake-X is stupid, but you can just ignore it. I bet they won't make the same mistake again. I wonder how much an effect the 1MB of L2 cache will have.

 

[MarcParis]

well, tell that to sli customer...switching from 16x/16x on X99 to 16x/8x on X299...what a progress...

 

[tbronzwaer]

what chips are you all going to get for this board?

I'm looking for a nice balance between single- and multithreaded performance; I want to be able to do high end gaming on this thing, but I also want a high (double precision) total throughput for stuff like custom ray tracers. I had been looking at the 10-core model which seems like a nice balance, and it offers the higher PCI lane count, but my thoughts on that are shifting (see below).

Having used a 14-core Xeon for the last few months that never seems to go above 2.5GHz despite supposedly having 3.0GHz max turbo, it's kinda rough unless ALL your workloads are heavily multi-threaded.

Interesting - can you tell us what type of cooling solution you used in that build? I would hope that with something like a good AIO cooler, those chips would at least come close to their max turbo speeds. I love the idea of a 14 core chip, but 2.5 isn't great. Your experiences are making the 8 core chip look more appealing.

Plus, for us it's perfect because we can't even use 28 lanes with this board anyway. Overall, I am pretty happy with X299

Is that true? If one fills this board up to the theoretical maximum, does that still not exceed 28 lanes? I hadn't thought of this, but it may influence my CPU choice...

 

[EdZ]

CPU choice

There is one x16 slot, and 3 x4 m.2 slots, for a total of 28 lanes form the CPU. The DMI 3.0 link to the PCH does not use PCIe lanes (Ryzen, and presumably Threadripper, have the PCH and inter-chip links share exclusive PHY links with some PCIe lanes).

I'd like to learn more about the mechanics of Turbo Boost 3.0/TurboMax/VTECjustkickedinyo when in moderately and heavily threaded workloads. From what I understand, during CPU validation the two cores that can clock the highest are identified and selected. But can these cores 'clock up' under any workload, or only two-core workloads? Is there a threshold for other core utilisation at which these cores will 'throttle back' to the normal Turbo Boost behaviour, or will they always be available to run faster all the time? Does the CPU indicate to the host OS to prioritise these cores for assigning processes above other cores? It's be rather annoying for the theoretical 4.5GHz two-core speed to be unattainable in any practical workload.