ASUS ROG Dominus Pictured, Core i9 XCC Confirmed to Feature 6-channel Memory

[btarunr]

This Tuesday at its Computex presser, Intel unveiled an unnamed 28-core/56-thread HEDT (client-segment) processor that's capable of being bench-stable at 5.00 GHz. The chip is a client-segment implementation of the Skylake XCC (extreme core count) silicon, which features 30 Mesh Interconnect "tiles," of which 28 are cores and two integrated memory controllers. The XCC silicon features a 384-bit wide (6-channel) DDR4 memory interface, and it turns out that whatever SKU Intel is planning, will require a different motherboard from your X299 board that can handle up to 18 cores and 4-channel memory. It will require a client-segment variant of the LGA3647 enterprise socket from the Purley platform. One of the first of these is the ASUS Republic of Gamers (ROG) Dominus.

Clearly bigger than ATX, in being either E-ATX or SSI form-factor, this board draws power from two 24-pin ATX, two 8-pin EPS, and three 6-pin PCIe, and has a gargantuan 16-phase VRM with two fan-heatsink blocks. Six DDR4 DIMM slots flank the socket, three on either side, each with its dedicated 64-bit wide path to the socket. The XCC silicon features a 48-lane PCI-Express gen 3.0 root complex, and so the board could feature at least two PCI-Express 3.0 x16 capable of full bandwidth, among a boat load of PCIe based storage connectivity, and onboard devices.



Update: This motherboard may have been a quick modification of the WS C621E SAGE, by removing one of its sockets, and modifying the rest of the board accordingly. Prototyping a board like that, for a company with ASUS' resources, would barely take 2-3 weeks by our estimate.

View at TechPowerUp Main Site

 

[Caring1]

And clearly needs a wind tunnel to remain cool.

 

[Th3pwn3r]

People will love hearing those little fans.

 

[CheapMeat]

Damn, I want this. I like unique EATX and larger boards. I have a few halo products just because I find them interesting. I doubt it'll be for sale exactly as it is though or ever.

 

[iO]

The amount of fans to keep those poor mosfets alive is ridiculous
But at least they have actual prototype hardware this time around and it doesn't look quite as rushed as last year when they had nothing but vage and incomplete specs to show.

 

[las]

People will love hearing those little fans.

Haha yeah

 

[CheapMeat]

The board Intel used for the 5GHz benchmark is more ridiculous.

 

[Vya Domus]

Wait , I though that was just a board put together just so they can showcase the new lineup. It just sort of looks ... makeshift ... like a server board mended to work as a desktop replacement.

 

[dj-electric]

This platform seems a little forced, a bit too over the top even for us blood thirsty enthusiasts.

 

[AltCapwn]

Jesus Christ that's an impressive beast! Why didn't they watercooled the VRM at this point?

 

[kastriot]

Overkill, too much PSU-s, fans, big ugly but hey it's intel

 

[CheapMeat]

Also, that block of fins is copper, not aluminum, according to AnandTech. That alone must've cost a good chunk. https://www.anandtech.com/show/12907/we-got-a-sneak-peak-on-intels-28core-all-you-need-to-know

 

[Valantar]

Jesus Christ that's an impressive beast! Why didn't they watercooled the VRM at this point?

Impressive? With that incredibly poorly optimized layout? No, that's not impressive.


Looks like they took a dual socket server motherboard, stripped out one socket, routed the VRM from that socket to the remaining one, rejiggered the PCIe to match, and called it a day. Lazy/rushed engineering if I ever saw it.

 

[bonehead123]

I must be missing something, but I only see 3 dimm slots on either side of the socket, not 6 like the headline says ????

 

[Valantar]

I must be missing something, but I only see 3 dimm slots on either side of the socket, not 6 like the headline says ????

Not all boards have 2 DIMMs per channel - this one has one per channel, hence 6 DIMM slots only. Which also supports my theory that this is a server 2P board with a single socket removed, as 1DPC is far more common in 2P servers due to space constraints.

 

[bogami]

So good hardware that everyone would have. Then, one does not know how to correctly position the pipes, which in this setting dampen the flow of fluid through the CPU.

 

[Hood]

E-peen at it's finest. The upcoming Threadripper 2 will have 32 cores. Intel i9 XCC will counter with 36 cores. And so on... A pissing contest that nobody can win. Great for headlines, but the question is, can bragging rights generate more profits for either company? Probably, the way people think about brand loyalty these days.

 

[efikkan]

E-peen at it's finest. The upcoming Threadripper 2 will have 32 cores. Intel i9 XCC will counter with 36 cores. And so on... A pissing contest that nobody can win. Great for headlines, but the question is, can bragging rights generate more profits for either company? Probably, the way people think about brand loyalty these days.

So far, both Intel and AMD are just bringing their high-end server CPUs to the consumer market. Hopefully this will finally bring down the prices of these chips. Even Broadwell-EP offered up to 22 cores at 2.2 GHz base / 3.6 GHz boost, so high-core chips are not new.

But I share some of your concern. While competition is good, hopefully this wouldn't lead to a core race (like we've had the GHz race before), focusing on hype and specs rather than real performance.

 

[Assimilator]

Seems like some management type at Intel decided that 5GHz is the magical frequency that'll make people forget about ThreadRipper. Well, tough luck guys, because a ridiculous system like this isn't exactly setup for the mass market, and coupling a binned Xeon with a server board tweaked to allow overclocking on that CPU isn't going to do much more than remind us all that you deliberately gimped overclocking on your server platforms.

C'mon Intel - stop dicking around, get that 10nm node done, and start innovating in the CPU space again.

 

[Kohl Baas]

E-peen at it's finest. The upcoming Threadripper 2 will have 32 cores. Intel i9 XCC will counter with 36 cores. And so on... A pissing contest that nobody can win. Great for headlines, but the question is, can bragging rights generate more profits for either company? Probably, the way people think about brand loyalty these days.

What we really need is consoles to get at least 8-10 cores with 16-20 threads or else there will be no game on earth that can profit from more cores. You can use your 28-core/56-thread HEDT to catch the console performance through a (developer wise) more convenient (aka poor) porting or feed some performance-consuming, irrelevant effect developed only to justify the otherwise totally useless extra performance. Of course there will be the option to pump up the resolution and the refresh rate/fps but the basic restrictions will be there, especially in multi-platform titles.

Well, at least there are some competition going on so it's time to cheer up. If the competitors want to beat each other with an ever-growing e-peen, then go on! I'll stay on mainstream and enjoy gaming and modding my level. Of course there is an urge in me to get the newest and coolest stuff, but until that urge meet the need to do so, it will stay an urge. I think this fight, this resistance is the burden of living in a consumer society. We all have our fight at some point/level.

 

[Valantar]

Why on earth are people here talking about gaming? These aren't meant for gaming in any way, shape or form. These are workstation chips, meant for CAD, 3D modelling, design, video editing and other heavy workloads. Sure, some incredibly rich streamer will no doubt buy one, and Linus Tech Tips will probably build an "X gamers, 1 CPU" rig with it, but other than that, only idiots buy chips like this for gaming. For any significant amount of gaming, you'd be better off with <12 cores and an OC, as games need high clock speeds to perform well. Of course you might run the odd game on it if you have a $20 000 workstation at home, but if you have that kind of cash, chances are you'll buy a dedicated gaming rig instead - as it'll perform better.

But please, stop discussing this as if it's meant for gaming. If that's what you're thinking of, you're missing the point.

 

[Kohl Baas]

Why on earth are people here talking about gaming? These aren't meant for gaming in any way, shape or form.

Because with ROG, it is branded up for gaming and overclocking, like it or not. Asus has it's own WS linup, yet it's ROG. Overclocking is also something you won't find on WS-line because at working, stability is more important than momentary performance. Loosing your entire work because instability is worse than having an extra hour or two to compute it.

 

[Valantar]

Because with ROG, it is branded up for gaming and overclocking, like it or not. Asus has it's own WS linup, yet it's ROG. Overclocking is also something you won't find on WS-line because at working, stability is more important than momentary performance. Loosing your entire work because instability is worse than having an extra hour or two to compute it.

Sure, but extreme OC on these chips is likely to be limited to the finance world where both total compute power and raw speed are roughly equally important. For workstations, a mid-range OC would be fine (and entirely stable), or they'd just buy them for the savings over buying Xeon.

As for Asus mixing up OC and Gamer branding into one mess, that's on them. Even if it's a ROG Board, saying it's for gamers is still just plain dumb. It simply doesn't make sense for gaming - an 8700k would outperform it in 99.9% of games, and likely an 8600k too.

As for this not being WS branded, that's likely a) for marketing purposes and brand building (of the "MOAR COARS ZOMG WE PWN YOU 4THE LULZ" KIND), and b) because they're not even close to running this through QA and validation for that lineup. And, of course, to cater to the XOC crowd, which has also been pushed into the extreme ROG SKUs in later years.

 

[efikkan]

Overclocking a workstation with a CPU at >$2000 is just silly, sacrificing warranty, stability and longevity for ~5% more performance. And as several people have pointed out, these CPUs are not relevant for gamers.

 

[cadaveca]

Overclocking a workstation with a CPU at >$2000 is just silly, sacrificing warranty, stability and longevity for ~5% more performance. And as several people have pointed out, these CPUs are not relevant for gamers.

Maybe it is to you, because your budget could not afford to replace such a system should it die, but I know many who simply can't get enough, and a $50,000 machine is no big deal, and if it breaks, they'll just toss it and buy another. This market is actually quite large, but there aren't any products for this type of user as an enthusiast, so I am very happy to see these CPUs get a platform that has the clocking restrictions removed so I can sell such rigs and make 5k profit on a single PC.

I'd be disappointed here is there wasn't a proper VRM phase for each core. The VRMs are huge, but there are other reasons for this than the CPU consuming all that power... I guess people forget all those mainstream boards with VRMs capable of 3000 W +. I'd say we are looking at the 700-850 range for a good 5 GHz chip, while some poorer examples will need maybe up to 1000W. If you are delivering 1000W, you want a VRM capable of twice that, IMHO, so these boards look like they should, to me.