Threadripper mATX motherboard in the near future ?

[cmvrgr]

I wonder if we will see mATX x399 motherboards for threadripper.

That is a huge minus in comparison to Intel's x299 that mATX will be available soon.

 

[VirtualLarry]

Why would you want a PC with 8 memory slots, and 64 PCI-E lanes, with only a mATX board? C'mon, get real. If you want a smaller board, go with Ryzen 7 and an AB350 mATX.

 

[cmvrgr]

Because I need to make portable workstations for work. I need max 32 gb of ram but need the 16 core and 32 threads.

Size is very important for my project.

x299 has a motherboard coming. X399 should do to. Some of us need only the core count not the other options available.

 

[Ancalagon44]

Difference is that it is possible to buy a CPU for X299 that does not have quad channel memory or even loads of PCI slots. So the mATX X299 boards might be the right thing for a KBL-X build - even if they can also fit the LCC and HCC chips.

The ThreadRipper platform has no such CPUs - everything supports quad channel memory and loads of PCI slots, so there is less incentive for them to release a mATX board.

 

[cmvrgr]

That's no good. I will have to move to a larger chassis.....

 

[StefanR5R]

Why would you want a PC with 8 memory slots, and 64 PCI-E lanes, with only a mATX board? C'mon, get real. If you want a smaller board, go with Ryzen 7 and an AB350 mATX.

For a single-CPU + single-GPU workstation, the µATX form factor is quite appropriate.

Why should we be limited to 8 cores in the µATX form factor? We already have up to 22 cores for µATX on X99. (And even for mITX, but then only with dualchannel RAM which would be absurd.)

Besides, the currently existing µATX boards for Ryzen 7 are rather limited in terms of PCIe lane allocation: All you can have is 16+4 from the CPU, but not 8+8+4 which would still have uses in some µATX applications.

 

[Solarion]

TR4 is a desktop version of a server socket. Expecting motherboard makers to plant that massive socket on a tiny motherboard seems unrealistic. I mean if you actually look at a threadripper board, you'll notice that just the CPU socket + the "no-go" zone around the socket will take up 1/3 of a mATX board. Then there's the significant VRMs required to feed such a powerful CPU. Heck, Asus hasn't even released a standard ATX motherboard for threadripper yet, preferring instead to stick with E-ATX designs. Again, if you actually look at an ATX threadripper motherboard you can see that fully half of the thing is taken up by just the CPU, DIMM slots, VRMs, and I/O shield. That's most of a mATX board right there and that's without a chipset, PCIe slots, SATA, USB, etc. The compromises to cram all that into such a small form factor would be brutal and the market tiny. It shouldn't be a surprise that motherboard makers aren't in a hurry to make that happen.

...and while x299 was mentioned for tiny motherboards, it'll be interesting to see what happens when someone plants a 7960x or a 7980xe on one of those things. Guessing it won't go great, but if someone wants to park their $2k CPU on a micro motherboard then hey...moar power to them. I just hope they have room for a kilowatt power supply in that tiny case if they want to use the CPU anywhere near its potential.

lol Perhaps this too should be mATX form factor:

 

[Topweasel]

There are numerous issues outside just space of the socket as well. Things like where to the M.2 slot (have to have at least 1) and Wireless card. Which are staples of the mATX ecosystem. Another is cooling. The ability to cool a 180w CPU just plainly without perfect cooler tunneling and heatsinks that extend waaay past the socket, cooler height for the most part rules out mATX SSF solutions since it would need to extend way past the limits of a thin ATX case.

Biggest issue is market.
Let's say that Intel can sell 1 million X299 systems. 1% of those guys want an mATX solution. It would make sense for Asus to make a mATX board even if it doesn't make sense, because that is still 10k in motherboard sales. Now let's look at AMD, for several reasons, but the biggest being that the R7 takes the edge out of the needs of a high core platform (since intel caps the base platform at 4c), Threadripper is going to sell at a smaller pickup rate than the AMD vs. Intel ratio. But even if we give AMD a nice 20% ratio and hold onto the X299 ratio for X399. We would be talking about 200k X399 systems. 1 percent of that would be 2k, So what you end up with is the same if not more time and resources developing a similar board, but with 1/5th the actual market to spread the cost. Not only that but because you have to offer it at a more expensive price, the market shrinks because for many that would like an mATX X399 that added price of the niche board would make it not worth it vs. a mid tower.

So the question becomes. Would it be worth it to Asus to make a $350 mATX board with maybe 1500 purchasers available?

This doesn't just apply to mATX X399 either. X399 as a whole or even AM4 sees the same thing. Look at the counterparts to X299 and 1151. Even with companies like Asrock with their relatively large selection of AM4 options are still relatively small in comparison to their Intel alternatives. With a smaller market means less desire to stretch themselves thin to avoid unsold stock.

 

[StefanR5R]

TR4 is big because the physical design of SP3 has been re-used. SP3 supports octa-channel RAM and 128 PCIe lanes, but TR4 only half of that.

Have a look how ASRock, Gigabyte, and MSI (i.e. not ASUS) have organized their TR4 ATX boards. Then mentally reduce DIMM sockets from 2x4 to 2x2, and take away the lowermost PCIe slots of course. It's certainly not easy to lay out a TR4 µATX board, but not out of this world either. It's hardly different from 2011(-3) on µATX.

Regarding the relationship of the µATX big-socket market relative to ATX:
The list of X99 mainboards at geizhals.eu for example currently contains

1x m-ITX with dual-channel DIMMs,
6x µATX,
44x ATX,
23x E-ATX,
74x total.​

Interpolating from that, chances for a µATX TR4 board coming to market are low but non-zero.

And C612 mainboards for single-socket servers and workstations:

1x m-ITX with quad-channel SO-DIMMs,
6x µATX,
20x ATX,
3x proprietary,
30x total.​

Difference is that it is possible to buy a CPU for X299 that does not have quad channel memory or even loads of PCI slots. So the mATX X299 boards might be the right thing for a KBL-X build - even if they can also fit the LCC and HCC chips.

Incidentally, the first announced X299 µATX board (MSI X299M Gaming Pro carbon AC) does not support KBL-X.

Things like where to the M.2 slot (have to have at least 1) and Wireless card. Which are staples of the mATX ecosystem.

MSI X299M as an example has three M.2 slots: two for storage, one for the included WLAN module.

The ability to cool a 180w CPU just plainly without perfect cooler tunneling and heatsinks that extend waaay past the socket, cooler height for the most part rules out mATX SSF solutions since it would need to extend way past the limits of a thin ATX case.

Not different from 2011(-3) with moderate OC headroom.

 

[Solarion]

That's the socket we have to deal with though...and I'm not having difficulty visualizing an ATX form factor Threadripper board, because I'm staring at an MSI x399 now. The chipset + heatsink alone is 3" x 2" and it gets toasty even at stock clocks. This is the busiest, most densely populated motherboard I've ever owned and I have a 16 dimm 2P opteron system on hand...it's not nearly as densely packed. Even if you cripple the platform by tossing out half the dimm slots, all but 1 m.2, and all save one x16 slot...it still won't fit without more sacrifices.

x299 = 2066 pins while x399 = 4094. From a board real estate perspective...they're not comparable and again we don't know what's going to happen when someone sticks a 16c/32t processor in a shrunken x299 board. One would hope that the motherboard makers were able to take that into account, but we won't know for sure until it happens.

 

[Yakk]

Probably better off with thin client workstations running off a centralized server for the price then imho.

 

[Topweasel]

It's certainly not easy to lay out a TR4 µATX board, but not out of this world either.

MSI X299M as an example has three M.2 slots: two for storage, one for the included WLAN module.

I was a little surprised with X299M. I still don't see how a 4k pin socket with its required pin outs (probably still a lot more than X299) the 180w power requirements would be all that manageable in an mATX layout.

But the heart of the matter is what I already outlined. There isn't money in it. Take the MSI selection. Not only do they have two mATX X299 motherboards (one is KL-X only). But they have 11 total X299 boards available. For a CPU that came out about 3 weeks earlier than ThreadRipper. How many TR4 boards? 1.

Even Abit who made out like a bandit as probably the prefered Ryzen system due to their deep launch and post launch product selection and value in products like the Taichi. They launched with 4 X299 boards. They only have 2 for X399 and they are the same boards with a couple of chips either removed or replaced with cheaper alternatives. Asus having the most bandwidth has the biggest X399 selection, but again uses the same board for two products, with 6 X299 options. Gigbyte pulls a MSI with 10 X299 boards to 1 TR4. X399 has almost the same amount of boards altogether as there were x99 mATX boards.

 

[Solarion]

These guys aren't going to engineer a whole bunch of different boards, unless they see a demand. Perhaps that demand increases with the release of the 1900x in a few days and possibly even cheaper threadripper CPUs later, but it's probably never going to rival mainstream Ryzen boards. AMD has tiny market share compared to Intel, and Threadripper is a tiny portion of AMD CPU sales...so you're looking at a subset...of a minority here.

 

[cmvrgr]

The truth is between 2 factors. Market Demand vs profit.

I read all theories that everyone mentioned about bla bla demand, Intel greater market share bla bla and mb manufacturer's profit.

How are you sure guys that mATX for threadripper is a small market?
How do you know buyer choices ?
How do you know that HEDT will prefer to buy double priced Intel cpus ? I am sure that in the enthusiasts and in business oriented tasks category threadripper will sell more. Anyone prefer to spend the difference of that extra cash that Intel demands for their overpriced cpus to a better GPU or ram or storage.

mATX is a great option for tr4 socket in the market and if it is available will sell for sure and not only at 1500 buyers that someone mentioned. MB manufacturers are just pressed in the corner by Intel in any aspect that can cut sales from AMD.

Only a fanboy would build an Intel system or someone that pumps oil from his backyard and money is not an issue

 

[Topweasel]

I read all theories that everyone mentioned about bla bla demand, Intel greater market share bla bla and mb manufacturer's profit.

Those aren't theories. Intel has a much greater market share and motherboard manufacturers don't run charities so the goal of all products they release is to make a profit.

How are you sure guys that mATX for threadripper is a small market?

Well that is easy. Compare the numbers of mATX boards for X99 to ATX boards. It's roughly 1 tenth. Chances are that the numbers are skewed even worse towards ATX when you understand volume sellers vs. even more costly enthusiasts solutions. But ignoring that you still have a good rule of thumb, less options equals less demand.

How do you know buyer choices ?

Product availability is one part of the equation. But this market maybe a first for AMD but has been around for over half a decade. You can pretty much comfortably map out buying trends by using BW-E and SKL-X purchasing habits. Which you can also gleam from looking at the selection listed.

How do you know that HEDT will prefer to buy double priced Intel cpus ? I am sure that in the enthusiasts and in business oriented tasks category threadripper will sell more. Anyone prefer to spend the difference of that extra cash that Intel demands for their overpriced cpus to a better GPU or ram or storage.

Again look at the SKL-X build thread and the TR build thread. I am not saying all of the SL-X purchasers are making great choices. But the amount of people posting their SKL-X build in those threads is much much much higher than TR.

mATX is a great option for tr4 socket in the market and if it is available will sell for sure and not only at 1500 buyers that someone mentioned. MB manufacturers are just pressed in the corner by Intel in any aspect that can cut sales from AMD.

The 1500 wasn't supposed to be an exact number. I don't know what the profitability point in MB volume is supposed to be. I used those numbers to show what the effect of trying to make a niche product for what is already a niche market for what is sadly still a secondary CPU manufacturer would be like. I am not saying TR won't or isn't a successful product. But Intel can't and hasn't forced the Mobo manufacturers not to make AMD motherboards. This isn't 1998 and they aren't selling these boards in white boxes (Praise the K7M). You can set a clock to the expectations of the motherboard manufacturers by their selections, the 7 TR boards is very telling. This isn't to say that one company isn't going to take a chance to eat up all the potential pent up demand. But there is a big difference for their being a market for a product and there being a viable market for the product.

Only a fanboy would build an Intel system or someone that pumps oil from his backyard and money is not an issue

That's just silly and absolutely untrue.

 

[cmvrgr]

Those aren't theories. Intel has a much greater market share and motherboard manufacturers don't run charities so the goal of all products they release is to make a profit.

Well that is easy. Compare the numbers of mATX boards for X99 to ATX boards. It's roughly 1 tenth. Chances are that the numbers are skewed even worse towards ATX when you understand volume sellers vs. even more costly enthusiasts solutions. But ignoring that you still have a good rule of thumb, less options equals less demand.

Product availability is one part of the equation. But this market maybe a first for AMD but has been around for over half a decade. You can pretty much comfortably map out buying trends by using BW-E and SKL-X purchasing habits. Which you can also gleam from looking at the selection listed.

Again look at the SKL-X build thread and the TR build thread. I am not saying all of the SL-X purchasers are making great choices. But the amount of people posting their SKL-X build in those threads is much much much higher than TR.

The 1500 wasn't supposed to be an exact number. I don't know what the profitability point in MB volume is supposed to be. I used those numbers to show what the effect of trying to make a niche product for what is already a niche market for what is sadly still a secondary CPU manufacturer would be like. I am not saying TR won't or isn't a successful product. But Intel can't and hasn't forced the Mobo manufacturers not to make AMD motherboards. This isn't 1998 and they aren't selling these boards in white boxes (Praise the K7M). You can set a clock to the expectations of the motherboard manufacturers by their selections, the 7 TR boards is very telling. This isn't to say that one company isn't going to take a chance to eat up all the potential pent up demand. But there is a big difference for their being a market for a product and there being a viable market for the product.

That's just silly and absolutely untrue.

You sound very logical. But no offense it sounds like tech science fiction novel in favor of Intel.

Because one or two topics in a forum have more subscribers doesn't prove nothing that high end users are preferring Intel products and for that reason mb manufactures backs mATX boards for x299.

That specific topics was build before the first ryzen cpus was announced and is logical to have that amount of followers and as we know before ryzen AMD had become nobody !

Ryzen forced Intel to compress pricing and bring multi core cpu s in the market. We all should thank AMD for this.

I can see around me that 9 to 10 workstations are switching to AMD 1950x and I am in a business related to computer systems.

I was impressed that many high end enthusiasts are moving too and they are not waiting the 16 core i9 cpus. The reason price vs performance. Only a crazy person would prefer to spent +700$ for a 16 core Intel i9 CPU even if it will have a small performance gap.

For me personally if AMD manage to steal back a respectable account of buyers it will force Intel to drop even more their prices and they both will make better and faster products.

mATX for x399 is a smaller market but not as niche as you believe.

I am waiting from the mb manufacturers their reply for the tr4 mATX boards and we will discuss it further. I made many questions with economical and technical nature to them.

I do not have time now to analyze more and reply to the sections that you replied and I am sorry for that. I will do it during weekend.

 

[Ratman6161]

Folks, don't forget we are talking about 180 Watt do US here. Just the cooling for such a processor will take room and not be very conducive to small cases. I'd be very surprised to see many syllable x SFF SYSTEMS either. The motherboards maybe used for i7 7740x or the i5 7640x.

 

[thecoolnessrune]

You sound very logical. But no offense it sounds like tech science fiction novel in favor of Intel.

Because one or two topics in a forum have more subscribers doesn't prove nothing that high end users are preferring Intel products and for that reason mb manufactures backs mATX boards for x299.

Motherboard manufacturers, and pretty much all businesses like making money. The uATX workstation board concept is not a new one. The concept is known. Boards for this market have been built before. It's not a novel idea awaiting discovery.

So we can safely say mobo manufacturers know this is a possible market, and they have been doing those boards for long enough that they should know what size the market is, and what room there may be for an AMD board.

So please, instead of shooting down reason after reason, why don't you tell us why mobo manufacturers are not currently making uATX Threadripper boards? What would convince a mobo manufacturer to abandon all that money just waiting to be sent to them if they made such a board?

 

[thesmokingman]

TR4 and cpu will probably take 3/4ths of the matx form factor, lol.

 

[Topweasel]

Motherboard manufacturers, and pretty much all businesses like making money. The uATX workstation board concept is not a new one. The concept is known. Boards for this market have been built before. It's not a novel idea awaiting discovery.

So we can safely say mobo manufacturers know this is a possible market, and they have been doing those boards for long enough that they should know what size the market is, and what room there may be for an AMD board.

So please, instead of shooting down reason after reason, why don't you tell us why mobo manufacturers are not currently making uATX Threadripper boards? What would convince a mobo manufacturer to abandon all that money just waiting to be sent to them if they made such a board?

People are stuck on, "I want one which proves that there is a market and must be bigger than you think, probably because you don't want one."

It's funny everyone wants to pretend they are a unique snowflake except the times they actually are.

There is room though. Threadripper and X299 sales are primarily retail and not big box purchases. X299 seems to be under selling and potentially under TR. I don't know if both combined are doing worse than x99 or X79, but if TR is selling well above Mobo manufacturers expectations board selections should increase and the potential for X399 mATX should increase as well.

Though I still caution that TR's strength are in things that a mATX setup limits like connectivity and M.2 expandability.

 

[StefanR5R]

To those who have doubts about the technical problems which socket TR4 poses in the µATX form factor:
Supermicro already offers Intel's socket 3647 on three variants of an µ-ATX board.

For 165 W TDP CPUs, 6-channel RAM, 10 GbE, 2280 M.2, 12x SATA, and everything.*

https://www.supermicro.nl/products/motherboard/Xeon/C620/X11SPM-TPF.cfm

Spoiler: photo

https://www.supermicro.nl/products/motherboard/Xeon/C620/X11SPM-TF.cfm

Spoiler: photo

https://www.supermicro.nl/products/motherboard/Xeon/C620/X11SPM-F.cfm

Spoiler: photo

--------
*) except RGB lighting

 

[cmvrgr]

2 mATX boards for x299 ....

https://www.scan.co.uk/shop/compute.../intel-x299-socket-2066-microatx-motherboards

Mini itx for the 18 core i9.

http://www.tomshardware.com/news/asrock-x299-itx-mini-itx-intel-computex,34586.html

Everybody understands that TR4 can not be fitted in mini itx but it can in matx.

There is demand from HEDT users for smaller form factors. Not everyone likes bulky systems.

 

[pcunite]

Came here to ask about this very subject.

I want an mATX board with Threadripper for the official ECC memory support. Might put in 32GB of ram and two NVMe drives (vertical position is fine). I don't need a monster case to hold a sea of hard drives or video cards. Just want the core count.

 

[FaaR]

Because I need to make portable workstations for work. I need max 32 gb of ram but need the 16 core and 32 threads.

Size is very important for my project.

Just saw your thread, and unless you've already decided on what to buy, my recommendation would be to go ATX format mobo and put it into a very small chassis, such as the Fractal Design Define C. Threadripper on uATX would be nuts, you could probably only fit 2 DIMM sockets, and the motherboard would packed with components, and it would likely be very hard to cool both CPU and its VRMs properly. TR is a big, thirsty CPU which sucks down power and comes back begging for more. It's not really suited for uATX cases.

x299 has a motherboard coming. X399 should do to. Some of us need only the core count not the other options available.

Core count is of little use if you don't have memory bandwidth to sustain performance. You might just end up wasting money and power on a CPU which is starved for data because a tiny mobo can't physically fit a full set of memory channels. Worst case, you end up with two DIMM channels hooked up to one Ryzen core, leaving the other core having to mooch off the first one for data. That'd be bad for performance.

 

[cmvrgr]

Just saw your thread, and unless you've already decided on what to buy, my recommendation would be to go ATX format mobo and put it into a very small chassis, such as the Fractal Design Define C. Threadripper on uATX would be nuts, you could probably only fit 2 DIMM sockets, and the motherboard would packed with components, and it would likely be very hard to cool both CPU and its VRMs properly. TR is a big, thirsty CPU which sucks down power and comes back begging for more. It's not really suited for uATX cases.


Core count is of little use if you don't have memory bandwidth to sustain performance. You might just end up wasting money and power on a CPU which is starved for data because a tiny mobo can't physically fit a full set of memory channels. Worst case, you end up with two DIMM channels hooked up to one Ryzen core, leaving the other core having to mooch off the first one for data. That'd be bad for performance.

No I haven't yet decided. I am thinking to go for an x299 MSI mATX board pay 700$ more for the 16 core i9 and give the finger to AMD

x399 can play nicely on mATX size board it's just a couple of pci slots less. Same memory slots can be fitted.

The only problem is with the motherboard manufacturers. They are not convinced that will sell well. Everything is about money...

AMD should have found a solution to have at least one available from one MB manufacturer.