Mini-STX: A new form factor

I am excite. Now I know what my P4400 is going into once I upgrade my current build. Media/Streaming PC built into a Fallout Anthology Mininuke without having to 3D print an extension, yesplz.
 
Cool - new smaller form factor but where's the pci-e or mxm? It kind of defeats the point of making something between the itx and NUC if there's no support for gpu expansion.
 
I think I heard that there may be GPU expansion possibilities with USB-C. I think I spot one of the board. I personally like this idea.
 
The problem is, the GPU enclosure would be bigger than the computer itself and would require its own cord. Don't get me wrong its still awesome, but I think ITX inside an ultra SFF case like the S4 Mini is more exciting to me personally because internal GPU.
 
I think I heard that there may be GPU expansion possibilities with USB-C. I think I spot one of the board. I personally like this idea.
Yeah, but that's completely different approach to use another box with external gpu that's supposed to go for a laptops.
 
Personally, if I were to get something without dGPU support, the ability to swap CPU wouldn't matter.
 
Yeah, but that's completely different approach to use another box with external gpu that's supposed to go for a laptops.

Exactly, if you have to provide an external chassis plus PSU for the GPU, you might as well have built it internal and gone slightly bigger with Mini-ITX.

The lack of expansion slot, and the incredibly limited use case plus other complications is the same thing hat killed Thin Mini-ITX.

Via had success with the Nano and Pico-ITX extensions to the standard because they targeted embedded devices. Dual m.2 slots is targeting one incredibly high-performance I/O subsystem, for an embedded system :D
 
Seeing as it is an AsRock board and there will be two M.2 slots, hopefully at least one of them will be Ultra M.2 (4x PCI 3.0). If that is the case, just use one of these bad boys:

http://www.bplus.com.tw/ExtenderBoard/P4SM2.html

It's already been done here (PCIe 2.0 x4 in this example):

http://hardforum.com/showthread.php?t=1858980

But that's not the point here. They are introducing another form factor and once again intel tries to push their igp while not thinking about that it will kill the whole concept.

I can only hope that some day they'll figure out that we need a form factor with right angle pci-e slot like this (on the opposite from where the slots are now):

8015-4467233.jpg
 
Why not just throw this out there:

Release M.2 GPU cards. Seriously. Put a GPU on the M.2 form factor. The factor has just enough room for a mid-range chip and a few blocks-o-RAM.
 
No, it doesn't. While I can see fitting chip + 2 ram blocks you still need some space for power components. Look at mxm:

Best-NVIDIA-GeForce-GTX-780M-GTX780M-GDDR5-4GB-256-Bit-MXM-Graphics-Video-Card-for-Dell.jpg


That also would mean amd and nvidia would have to make smaller package base which isn't happening or they would need to make whole card themselves by putting die directly on m.2 pcb and that's not their business model either.
 
No, it doesn't. While I can see fitting chip + 2 ram blocks you still need some space for power components. Look at mxm:

Best-NVIDIA-GeForce-GTX-780M-GTX780M-GDDR5-4GB-256-Bit-MXM-Graphics-Video-Card-for-Dell.jpg


That also would mean amd and nvidia would have to make smaller package base which isn't happening or they would need to make whole card themselves by putting die directly on m.2 pcb and that's not their business model either.

What if the M.2 slot was rotated to resemble a typical PCI-E expansion slot. Then you could essentially have little 'mini cards' that resemble much smaller graphics cards, only with a male M.2 set of fingers, just drop it in like typical card. Viva la SFF.
 
Didn't think about that. Good point - just like some mini-pcie slots for wifi placed vertically. But then what would be the point vs making pci-e x1/x4 cards? Only to support those unique boards with no pci-e slot...
 
This form factor needs a motherboard with thunderbolt 3.

It also needs 14x14cm GPU's that are thunderbolt 3 connectable.

Nox
 
I feel like mSTX will slowly replace thin mITX in most scenarios. The problem with thin mITX was that it required too many things to be on the board, like an LVDS connection and headers for display backlight, because it was meant to be used in AIO PCs. Most people that bought such boards put them in passively cooled cases or small ITX-only cases with no expansion slots, so all of these complicated features that kept a lot of mainboard manufacturers from making boards, were never used at all.

mSTX is meant to satisfy system builders that want a NUC with a socketed CPU, and nothing more. NUC is too proprietary to be successful on a grand level, the number of third-party components is low, and the layout isn't standardised enough. mSTX offers all the benefits of thin mITX like the fixed socket position, external DC input, fixed zones for components and gets rid of all the disadvantages while becoming smaller.
Instead of being tied to a standardised cooling solution, the idea with mSTX is that cases will always be bundled with a cooler in order to maximise cooling performance.

I understand that MXM would've been a fitting addition, but this form factor is not aiming for the high performance market and the implications for cooling would've made case design much more complicated. If you want to go small and powerful, mITX will be the way to go for quite some time to come.
 
I was really hoping what Dell did with the Alienware Alpha was going to become a thing.

For those not familiar, it's a UCFF motherboard with the GPU on the motherboard along with the CPU.

If we could get a socketed version of that, I'd be happy.
 
I understand that MXM would've been a fitting addition, but this form factor is not aiming for the high performance market and the implications for cooling would've made case design much more complicated. If you want to go small and powerful, mITX will be the way to go for quite some time to come.

I disagree. We've already got a full-desktop GTX 980 in a MXM form factor. Sandwich that back-to-back with an STX mobo that has a connector on the backside with the pins parallel to the board. Slap some proper blower fans on each side, one for CPU, one for GPU. Or go full fan+heat sink mode and slap some low profile cooling solution on either side. Top down view would look like this:

____CPU______
|______________
+++++GPU

EDIT: forgot to include the rest of my argument:

I think STX should be aiming for high-middle performance. We're all ready to downsize from mITX. The tech has been around for some time now, just that nobody has put it all together in the same package and made it play nice yet. I would jump on an STX with the aforementioned connector in a heartbeat and never ever look back at my M1. Two nice ultra low profile heatsinks on either side, and two thin 140mm fans cooling them would be an amazing set up. I personally would settle for full size Noctua IPPC fans, but it'd make the case probably close to 100mm wide. I'm thinking of Intel's 5x5 board though, with the horizontal RAM modules.
 
Last edited:
I don't think going for mxm would be a good choice since we still don't have the option to pick off-shelf mxm cards for general use.

I agree with the layout you're proposing - it's essentially what dondan did in his a4 and IF intel made a form factor with such angled pci-e connector on the back of the motherboard it could be a game changer.
 
it could be a game changer.

That's what I'm saying! If only these chumps would go for it. This sub would be all over this stuff, I think.

If MXM becomes an off-the-shelf component. If it becomes feasible to make Thunderbolt eGPU MXM enclosures for the laptop crowd. If Intel stops banking on their stupid iGPUs. So many 'ifs' that are holding us back from enjoying a truly high performance UCFF system. It's so frustrating.

I have a hard time believing anybody who says the market is too niche. I really do. I think the reason PC sales are dwindling is because nobody wants a giant ass tower in their room anymore. They take up too much space, look often as if they've been designed by a CoD-binging 14 year-old, are overpriced, and wholly unnecessary in terms of I/O for most use-cases. I think the SLI and overclocking crowd is MUCH more of a niche market than what I'm talking about. But maybe I'm out of touch and too biased.
 
I'd also argue that if some vendors really started pushing MXM, MXM would take off.

MSI's getting behind it and putting it in desktops, now, though, and apparently guaranteeing a couple years of GPU upgradeability for their platforms that use it.

Another thing with eGPU boxes, though, for 5x5 (what ASRock is calling Mini-STX) and NUCs and such, though, is that if you design the 5x5 board to be USB-C powered, you can power the board off of the eGPU box.
 
I'd also argue that if some vendors really started pushing MXM, MXM would take off.

Agreed.

Another thing with eGPU boxes, though, for 5x5 (what ASRock is calling Mini-STX) and NUCs and such, though, is that if you design the 5x5 board to be USB-C powered, you can power the board off of the eGPU box.

Well, I think it'd make more sense to have STX systems that could integrate an MXM GPU, as opposed to doing an eGPU solution for something that's going to stay on your desktop. Same reason people are mostly like "???" with Intel's Skylake i7 NUC announcement where they said that you could plug it into an eGPU. It doesn't really make much sense for a desktop application.

Where I want to see MXM TB3 eGPUs is for the laptop crowd. That's where it makes the most sense. We don't need full size GPUs anymore. NVIDIA has crammed a desktop 980 onto an MXM board. Dedicated graphics NEEDS to continue in that direction.
 
No vendor will start "pushing" the mxm. The reason is simple - warranty and service checks are quite a pain and unless there's clearly visible damage done to the device like from water or mechanical damage then vendors will replace, fix device or reinburst the client. Because of that shipping an mxm card that doesn't have standard cooling solution and giving warranty for that is not a thing fitting current business model for the vendors.
 
No vendor will start "pushing" the mxm. The reason is simple - warranty and service checks are quite a pain and unless there's clearly visible damage done to the device like from water or mechanical damage then vendors will replace, fix device or reinburst the client. Because of that shipping an mxm card that doesn't have standard cooling solution and giving warranty for that is not a thing fitting current business model for the vendors.

How is that any different from third party CPU coolers? I don't see this as an argument against making MXM available off-the-shelf. Ship them with a boxed cooler like Intel does with their CPUs. If people want to change their coolers out, let them.
 
Vendors aren't intel or amd - they are smaller and they are using licensed technology (not fully their own know-how) so they are limiting the risks to the minimum.

Note that gpu's are a bit different in tdp and how they peform temperature-wise and that there are other components apart from chip that are picked by vendors that need to be cooled.

There were some recent problems with radeon drivers that made cooling turn off and fry the gpu while on the other hand cpu's shut down immediately when temps are dangerous. At this point I fairly remember having damaged cpu later than pentium 3 era while on the other hand motherboards and graphic cards are all that dies quickly.
 
I disagree. We've already got a full-desktop GTX 980 in a MXM form factor. Sandwich that back-to-back with an STX mobo that has a connector on the backside with the pins parallel to the board. Slap some proper blower fans on each side, one for CPU, one for GPU. Or go full fan+heat sink mode and slap some low profile cooling solution on either side. Top down view would look like this:

____CPU______
|______________
+++++GPU

EDIT: forgot to include the rest of my argument:

I think STX should be aiming for high-middle performance. We're all ready to downsize from mITX. The tech has been around for some time now, just that nobody has put it all together in the same package and made it play nice yet. I would jump on an STX with the aforementioned connector in a heartbeat and never ever look back at my M1. Two nice ultra low profile heatsinks on either side, and two thin 140mm fans cooling them would be an amazing set up. I personally would settle for full size Noctua IPPC fans, but it'd make the case probably close to 100mm wide. I'm thinking of Intel's 5x5 board though, with the horizontal RAM modules.

While that solution is interesting and I would like to see the market move there, but we're already getting into problems.

1. Cooling. This isn't standardised for MXM, so as opposed to the cooling of the CPU, which the case manufacturer can trim to their case by the design of the formfactor, this isn't true for the GPU. MXM cards don't ship with cooling solutions like regular GPUs do, so a case manufacturer who was willing to support MXM cards in his case, had to supply a cooling solution with it that can support all and every single card you could possibly put on there, including a GTX980, which puts out more than double the heat of the strongest CPU mSTX supports. Additionally, the mSTX board would have to be able to control that cooling system, which means additional logic and circuitry on a board that might never use it, which is the exact problem Intel faced with thin mITX.

2. Power. mSTX is supposed to be a low-energy platform, powered by 19V laptop bricks (maybe 12V-24V or something like that, I don't know exactly). As it has been reported by users of the S4 Mini and QinX' research for his new project Kees KogelMier, even the strongest laptop bricks have issues dealing with the enormous power draw spikes pretty much all GPUs experience. This issue occured with a R9 Nano, which draws 175W and multiple power bricks, all supplying north of 300W of power. A GTX 980 can draw up to 250W. This poses serious problems when implementing a system like you're proposing. As the mSTX board contains it's own power converters to supply 12V, 5V and 3.3V, it would have to handle the whole load of the GPU on it's own. If Intel limits the CPU power draw to 65W, you can imagine how good such a board could supply 250W to a GPU. The power circuitry would have to be beefed up significantly, alongside with Intel developing a power brick strong enough to supply the needed power.
The alternative would be to limit the power draw of the card to 75W, which is what a 750Ti would draw. Then everyone would be shitting on them for not going all-in.


There is no good way of supplying that much power via an external brick. Zotacs SN970 or the Alienware Alpha can do this sort of stuff because everything inside them is custom-designed, and even those don't have a GTX 980 inside them. mSTX isn't custom-designed, it's a standard. The only good and reliable way of supplying lots of power to a PC is via an internal PSU right now, and mSTX is supposed to make PCs small and simple. Using MXM GPUs make everything extremely complicated.
 
I'm headed out the door, so I can't respond to everything right now, but what about a separate brick just for the GPU?
 
I'm headed out the door, so I can't respond to everything right now, but what about a separate brick just for the GPU?

In my eyes, that completely defeats the point of the exercise. The two bricks will are close to being as large as the whole PC at that point, and while it might be the solution for some private projects, it seems like an unreasonable choice for a commercial product to me.
 
I'm thinking the GPU could supply the PC, although you've still gotta supply quite a lot of power to keep a GPU happy.

Oh, and MXM can be reasonably compatible, I thought, if you stuck to a 100 W or a 55 W (depending on MXM-B or MXM-A) TDP limit - the idea being that you could use any MXM card in a MXM-B chassis+heatsink, or any 55 W or less MXM card in a MXM-A chassis+heatsink. It's really just the GTX 980 for Notebooks that violates Nvidia's own MXM specs on TDP...
 
I think the whole reason MXM is not an 'off the shelf' solution is because it does not have the demand to be one, and it doesn't have a self-contained cooling system. If there was a MXM-PC spec that included guidelines for a self-contained cooling system, we could essentially have the framework for mini-GPUs to be used in Uber-SFF systems.
 
Absolutely agree. A spec for a self-contained cooling system would be the first step for MXM to become a proper form-factor for desktop GPUs.
 
I'm eager to see whether this comes to market or not. It really looks like a good concept, but I'm not sure how they could solve the availability problem of MXM cards. Maybe they'll just offer them through their own store? That way they can also make sure that all of the cards line up with their cooling solution and don't draw too much power for the power brick.
 
It looks like the MXM3.0 daughter board has his own video output and two connections cables. One for power and one for data. If they use m.2 standard for data it would be amazing.

Agreed, I'd buy the shit out of this. TAKE MY MONEY GIGABYTE. :p

Seriously though, that could be completely spectacular. An "MXM enclosure" with a standardized cooler, case, and offsets for attaching a 35W NUC/mSTX motherboard. The motherboard would be powered through the brick for the GPU. Or simply add a nano ATX internal PSU space a la HDPlex or Streacom.

That seems like it could be a completely viable and profitable product for a sizeable niche of SFF enthusiasts like myself.
 
A little off topic, but has anybody here tried or purchased the Zotac SN970? Seems like a pretty good portable media/lan, secondary pc to have once you install Windows on it.
 
I think you reach a point where there are too many advantages to a proprietary design. When you want to make something as small as say a NUC or something, going DIY with generic components and trying to utilize space to that level of efficiency is a losing battle.

ITX still makes sense because you're using a variety of standard desktop components, including a stand alone GPU. So, thanks, but no thanks.
 
I think you reach a point where there are too many advantages to a proprietary design. When you want to make something as small as say a NUC or something, going DIY with generic components and trying to utilize space to that level of efficiency is a losing battle.

ITX still makes sense because you're using a variety of standard desktop components, including a stand alone GPU. So, thanks, but no thanks.

The form factor is just getting started. ITX was pretty janky when it first came out, and many of the boards first available had integrated CPUs rather than swappable ones. I don't think it will be too long before someone puts a PCI-E connector on the reverse side or where the DIMM slots are or something.
 
The form factor is just getting started. ITX was pretty janky when it first came out, and many of the boards first available had integrated CPUs rather than swappable ones. I don't think it will be too long before someone puts a PCI-E connector on the reverse side or where the DIMM slots are or something.

PCI-E on the bachkside would actually be a great idea since it would allow for super compact Dan A4-like layouts without needing a riser card.
 
I'd rather see manufacturers cram more stuff onto ITX boards, specifically M.2 and DIMM slots.
 
It is going to be dependent how well it sells vs. the NUC devices already out. If it gains steam fast, I could definitely see possible big leaps in RnD for MxM Desktop cards. If PCI-E is implemented on the backside as Aircookie stated, this could be the future for the desktop market. Of course, there will still be a market for DIY PC builders that want to OC and water-cool. I am still not quite sold on M.2 PCI-E SSDs as the temps they run at makes me wary of their long-term use/reliability.
 
Back
Top