Awesome mobo... I hope it can be purchased separately... or if they (and others) will make more boards like this. The day of a million add-on cards are over for most, 1080s can be fit onto MXM size cards, lots of great M.2 SSD options, can get 1x16GB SO-DIMM sticks, power draw is low and can use external power bricks.
That's pretty cool, but I question the benefit of MXM over, say, a right-angle PCIe slot for a low-profile GPU. MXM is just too damn rare. The worst part is finding a heatsink for MXM cards when adding one to an OEM system since most cards are sold without them, and the designs are all super proprietary.
MXM Allows for a full range of GPU choices (all the way to the GTX 1080). HHHL cards take up more space, and the biggest/best you can get there is either an RX 460 or a GTX 1050. That would be a far, far worse choice.
Also, this might be one of the things that makes MXM cards more available to consumers. If nobody tries, they'll stay OEM only.
The market for MXM just isn't here yet. And MXM cards were designed for severely z-height limited systems like laptops. Insisting on fitting that into a desktop with (relatively) plenty of head-room is a pointless exercise. And since MXM cards are designed with laptops in mind 99% of the time you'd be hard pressed to find one that's actually speced like a desktop card. Cooling limitations ensure that it will never behave like one even if it had the same specs.
The sales for PCs are declining for years now and the segment for "really tiny but oh so powerful systems" is probably nowhere near the size needed to convince manufacturers to put out good and cheap MXM cards. So you're stuck with expensive modules that only run on some systems, that need custom cooling only useful in those particular systems and very limited upgrade options.
And what's the real benefit of an MXM card when compared to something like the R9 Nano? Yeah, the Nano is a little larger by a couple of cm on each side but it shows there's little need to implement yet another standard in the PC space that would only fragment the market even more and drive prices up. You can simply optimize and shrink what you already have. There's no reason future generations of GPUs couldn't be fit onto PCIe cards and installed horizontally in a "micro" case without having to worry that the card will be useless outside of that very custom system.
1) The R9 Nano uses HBM, which is how they got it that tiny. Are you saying SFF builds should be limited to HBM-equipped GPUs, or GPUs low-end enough to require very little space? That makes little sense to me.
2) Sure, MXM was designed for low Z-height. But it mainly was designed to take up less area than standard PCIe. Would it be possible to make "desktop only" MXM cards with slightly taller (and probably cheaper) power delivery components? Absolutely. And the "less area" thing is what matters here. Again, you can get a _fully specced_ GTX 1080 in MXM form - only limited by cooling. Sure, it won't OC as well as the desktop cards, but that really doesn't matter.
3) "99% of the time you'd be hard pressed to find one that's actually speced like a desktop card." No. The entire mobile GTX 10XX series is ~identical to their desktop counterparts, with minor optimizations for power savings. Heck, the 1070 has _more_ cores than the desktop version (albeit at lower clocks). Performance differences stem mainly from lack of proper cooling. Which would be a very solveable problem in a case like this.
4) Sales for PCs are declining, sure. Sales for gaming PCs in every single possible form factor is growing. Including tiny "console replacement" PCs.
1) The GTX 1080 is so good in mobile devices because it's a very efficient chip (performance per watt). Are you saying SFF builds should be limited to very efficient chips? Is this your argument? o_O How can limiting them to MXM boards sound better? MXM has virtually 0 availability for desktops right now and even for laptops it rarely met its purpose. How long do you think it will take for high end GPUs (GTX 1080 equivalent) to have HBM? How long before MXM becomes mainstream and affordable?
2) It's very possible to make a standard PCIe card much smaller. And that's exactly the point. Why work with yet another standard that's incompatible with almost everything in the desktop space right now? With PCIe you'll be able to move your tiny card in a full size PC should you want that.
3) Absolutely identical save for the lower clocks and the fact that TDP limitations will always force that chip to drop clocks even further when used in a thermally constrained environment (like a laptop or typical MXM cooling). Long story short there is no practical MXM implementation right now that can compete with a desktop card implementation because of its intrinsic limitations.
4) How many games are sold is irrelevant. I was talking about a discreet GPU market that becomes relatively small and then you go on and split it between two completely incompatible formats: PCIe and MXM. Every other format goes for universal standardization (at every layer) so why should graphic cards be an exception? Just for the sake of exotic hardware?
There's a tradeoff to be made here and it might not be worth it when it comes at the price of compatibility. Memory modules and SSDs can be made even smaller but they're not for compatibility's sake. You have the wrong answers because you're asking the wrong questions. The first one should have been "why invest into building a whole new hardware ecosystem instead of optimizing the one you have?".
1) Yes. Or, well, I'm not arguing that they _should_ be, I'm simply pointing out that due to the laws of thermodynamics and/or current cooling technology, SFF builds _are_ limited to one of three things: low-power GPUs, very efficient GPUs, or some sort of water cooling implementation. I'd LOVE for my Fury X to fit in a tiny chassis (heck, I'm moving it into a custom loop in an ITX case some time this year). But it can't. Below a certain volume, displacing >250W of heat simply becomes impractical.
2) Sure. But that would drive costs up massively - after all, miniaturization is as much of a reason for the cost of MXM cards as scarcity. SFF costs more, and will always cost more. I for one am willing to pay somewhat of a premium for that. Also, MXM _is_ PCIe. Making a PCIe-to-MXM adapter in the case of proliferation of these cards would be no problem whatsoever. Gigabyte already has one in their BRIX UHD. This would make for _more_ standardization and interchangeability, not less - you could migrate the GPU from your laptop to your desktop, or vice versa. Upgrade either, or both. Keep, and reuse, a part that would previously go to waste. And so on.
3) Lower clocks? Only due to TDP limitations. Nvidia is quite flexible with this, but they don't want people to fry their laptops simply because they've heard that overclocking is cool af. In a case like this, with sufficient cooling, there's no reason for the board to deliver equivalent clocks to stock cards. Will the power delivery handle similar OCs to a full PCIe board? Probably not. But neither would the power delivery on your proposed tiny PCIe card.
4) Who mentioned games? I sure as f*ck didn't. Please read before you comment. I said gaming PCs. Gaming PC sales are _rising_. Sure, AIB GPU sales are droppiing off, because sales of low-end (think GT 730, Radeon R7 240) are tanking. Cards like that used to represent HUGE sales numbers (with near zero margins, but nvm), but are now being replaced by capable iGPUs. Sales of midrange and high-end GPUs are either stable or rising. Sales numbers for >$200 GPUs are massively up the last 5-10 years.
And no, I'm not asking the wrong questions here. If anyone is, you are. I'm talking about _increasing_ compatiblity. MXM is currently irrelevant because the market for gaming laptops used to be tiny, and when laptop makers had to design the graphics card custom any way, they might as well integrate it onto the motherboard. This is no longer the case, although integration is still done to a large degree for the sake of miniaturization (as thin-and-light gaming laptops are very popular). Still, the popularization of good gaming laptops could quite easily reinvigorate the MXM standard. Also, of course, the MXM-SIG needs to step up its game.
But, you might ask, how can MXM cards increase compatibility? That's pretty easy. -They can be adapted to PCIe with relatively simple, low-cost adapters (that can carry over between generations, unless you need new display outputs). As mentioned, see the Brix UHD. -They allow for standardized coolers. You keep arguing for compatibility, but ignore the fact that GPUs are a free-for-all when it comes to cooling solutions. I'd love to see you fit a high-end cooler from the previous generation to a current-gen card, even from the same maker. Not a chance. Want custom cooling? Great, you can replace that every upgrade too! The clusterf*ck that is GPU coolers would benefit greatly from standardized component placements. A market where GPU coolers can be chosen like CPU coolers can is a market I welcome with open arms. And yes, this can of course be done with standard PCIe cards. But there's no incentive to do so, as added components (bigger power delivery and such) is considered a value-add, and standardized component placements could jeopordize this. In the SFF space, the incentive is there, as there's limited room for huge coolers, and fitting the cooler to the case is a more pressing issue. -They increase interchangeability. A single card could feasibly fit both laptops, SFF PCs and regular desktop PCs.
Also, the PCIe form factor standard places pretty severe limitations on how small you can make cards. -IO slot sizes are standardized, and not optional. You're stuck with either HH slots or FH slots, and both have bulky and awkward form factors and mounting mechanisms for SFF. This isn't surprising, since this aspect of the standard is made to be compatible with AGP, PCI, ISA (and so on) slots, with dimensions worked out in the 80s and early 90s meant for servers and full-tower desktops. -PCIe power delivery is big and clunky, and has no place in SFF PCs. Of course, power connectors are adaptable. But for SFF use, a smaller standard (with higher quality requirements to make up for a smaller connector) is pretty much necessary.
And the answer to your last question is: Because the one we have is reaching its limits in a number of significant ways*. This _should_ prompt a re-think. But then again, dumb people kept using Windows XP for more than a decade. I guess you can't beat'em all.
*With the caveat that this isn't a new hardware ecosystem, but rather a compatible, adaptable, optimized implementation of an already-open standard.
And here I was thinking the point couldn't have been made clearer. There's some dissonance in what you're saying. Like how expensive it is to shrink a PCIe card just a little while being able to maintain compatibility with every desktop out there but thinking that making an almost inexistent standard in the PC market popular with adaptors and reengineering everything on the market must somehow be dirt cheap...
In tomorrow's feature: microMXM for USFF, EMXM for larger but not so large builds. Also microSODIMM, microATX power connector, micro USB headers, micro M.2 slot, micro screw holes, and why not... micro CPU socket. Some of there require almost 0 re-engineering so maybe breaking compatibility is a small price to pay to get a slightly smaller footprint. And I'm sure you could easily just buy some adapters to make these compatible with everything else. I mean why would it make sense to do this with GPUs but no other component?
P.S. Yeah, not a completely new hardware ecosystem... just one that barely registers on any radar even in the laptop market let alone the desktop market where I think it exists exclusively in a handful of OEM builds. But hey, if it sounds cool and exotic why not. I remember having an Asus sk. 479 adapter for the desktop. The next big thing.
I am definitely in favor of some new form factors with GPU support. Mini-ITX leads to issues. The way the GPU hangs off the edge in points down (in vertical cases), means either:
You make the case a lot bigger to have space, or the GPU is cramped for air, or you put in a giant dust admitting cutout to help with cramping. I don't particularly like any of these options.
So if I was building today, I would either stick with Micro ATX if using a GPU, or skip GPU and go with an ultra small case.
There just doesn't seem to be a good small + GPU form factor option that I like. There are some kludgy horizontal adaptors for the Mini-ITX slot, but they leave the GPU point the CPU at the bottom of the case, so again, something of a problem.
A real horizontal MXM GPU form factor that got widely used might be an answer, though I am not sure this is exactly the right one. I am thinking more Mini ITX + MXM, and don't mess with the standard CPU area to shave a few mm.
I'm pretty sure the SATA connectors are the two odd-looking compact connectors next to the m.2 slots. They look similar to connectors like this seen in NUCs and other UCFF PCs that combine SATA+power into a single low-profile connector.
"Personally, I’d prefer a solid copper cooler on there for a Core i7, but a Core i5/i3 would do OK with the stock cooler."
Of the 7 Skylake and Kaby Lake i7's in intel's desktop stack, 5 are 65w or below. Perhaps this sentence would be better worded by talking about K-series CPU's, instead of focusing on i7 vs i5/i3, which isn't very relevant in 2017.
I have to say this looks really, really good. Of course, the board could do with better power delivery. But it's still pretty awesome. With some better cooling, you could have a coffee table book-sized i7-7700K+GTX1080 PC. That would be amazing. Also, upgradeability and flexibility makes this very interesting as a concept.
Looks ok, but not sure the use case. Wouldn't any of the low profile and/or short gtx1060 and related cards manage this in a one of the small cases (mini-itx, micro-atx, and related)? MxM seems pretty expensive and limited options compared to the availability new smaller cards that are in wide availability from multiple vendors at approximately $0 price premium.
well, a short 1060 isnt exactly in the same ball park as a 1080. You would also have a lot more z height to deal with.
A board like this will be super useful for NUC style HTPC and gaming PCs. I know I'd love one for my TV. Between this and zotac's e boxes, there is a market for slim gaming pcs that are book sized.
I had a similar idea in mind. Have a socket for the GPU (analog to the CPU socket) and the GPU being a single chip like (would only work GPUs with HBM memory). The video output can be on the MB (pci-e or something specific for the connection). This way the CPU and GPU can share the power delivery, and can use a single cooling system that gets on top of both. Then again, just join the 2 and get an APU...
I think you'd still run into compatibility problems due to different voltage needs from one GPU generation/family to the next that would limit future compatibility in the same way that LGA1156/1155/1150/1151 have meant you're unable to keep swapping in new generations of processor despite CPU IO mostly staying the same from generation to generation.
The desktop market needs this soo badly! Full ATX and even mATX is just huge and awkward. We've come so far at making things smaller and faster, why are we still hanging on to full length PCIe expansion cards that are wasteful of materials while encouraging GPU companies to keep bloating up on TDP despite getting access to new manufacturing processes.
Agreed. The MXM-SIG should really step it up - this time period is a golden opportunity for them, which they seem to be wasting. The Zotac E boxes, Alienware Alpha and so on would really benefit from the upgradeability that MXM could bring to them.
The trouble with APUs and similar Intel non-Iris iGPUs isn't reall a lack of graphics processing power. There's a much more substantial bottleneck in contending for limited access to relatively slow system memory. Whereas modern GPUs with their own memory have a big dedicated pipe (even the lowly GTX 950 has over 105 GB/s), iGPUs end up dealing with less total bandwidth and contending with the CPU plus other components for a slice of that smaller pie. That can have adverse effects on the texture sizes, maximum playable resolutions, and everything else regardless of how powerful the iGPU is. Intel's eDRAM is tiny, but offers 50 GB/s of dedicated bandwidth to mitigate the problem. Modern texture compression can help too, but that only gets you so far before you end up starving the GPU cores. Then there's the problem of a shared thermal budget and close proximity of numerous hot running CPU cores that would otherwise be on an entirely different IC somewhere else on the motherboard.
Besides that, the current top end AMD A10 has 512SP so though it's a pretty good iGPU, it's not competitive with a MXM 1060. We'll have to wait and see what Ryzen brings to the table, but I think even with a node shrink to 14nm, there's just not going to be a big enough leap to make a Ryzen APU into a GTX 1060 killer even whent hat 1060 is crammed into a MXM card in a small form factor case.
7nm? Sure, in 2019 or 2020. Have fun with your APU in 2-3 years. I guess you'll be playing Super Mario Run on your phone until then?
Also, 125W is low for anything supposed to perform well with a mid-range GPU plus a 4c8t CPU on board. Remember that the more densely you pack your heat generating components, the harder it becomes to keep them all operating within reasonable temperatures. Besides, you'd still need a rather huge cooling solution for this. Nvidia has a decent efficiency lead on AMD (for now), but a laptop GTX 1060 + a 45W i7 should be comparable to this thermally. Laptops with this need pretty hefty coolers - and that's with _two_ sizeable dies across which to run heatpipes. With just one die, this becomes significantly harder.
I'm not saying I'm against this - not at all, I'd love for AMD to make a high-end APU like this. I just don't see it as quite feasible yet.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
33 Comments
Back to Article
colinstu - Wednesday, January 18, 2017 - link
Awesome mobo... I hope it can be purchased separately... or if they (and others) will make more boards like this. The day of a million add-on cards are over for most, 1080s can be fit onto MXM size cards, lots of great M.2 SSD options, can get 1x16GB SO-DIMM sticks, power draw is low and can use external power bricks.Samus - Wednesday, January 18, 2017 - link
That's pretty cool, but I question the benefit of MXM over, say, a right-angle PCIe slot for a low-profile GPU. MXM is just too damn rare. The worst part is finding a heatsink for MXM cards when adding one to an OEM system since most cards are sold without them, and the designs are all super proprietary.Valantar - Wednesday, January 18, 2017 - link
MXM Allows for a full range of GPU choices (all the way to the GTX 1080). HHHL cards take up more space, and the biggest/best you can get there is either an RX 460 or a GTX 1050. That would be a far, far worse choice.Also, this might be one of the things that makes MXM cards more available to consumers. If nobody tries, they'll stay OEM only.
close - Thursday, January 19, 2017 - link
The market for MXM just isn't here yet. And MXM cards were designed for severely z-height limited systems like laptops. Insisting on fitting that into a desktop with (relatively) plenty of head-room is a pointless exercise. And since MXM cards are designed with laptops in mind 99% of the time you'd be hard pressed to find one that's actually speced like a desktop card. Cooling limitations ensure that it will never behave like one even if it had the same specs.The sales for PCs are declining for years now and the segment for "really tiny but oh so powerful systems" is probably nowhere near the size needed to convince manufacturers to put out good and cheap MXM cards. So you're stuck with expensive modules that only run on some systems, that need custom cooling only useful in those particular systems and very limited upgrade options.
And what's the real benefit of an MXM card when compared to something like the R9 Nano? Yeah, the Nano is a little larger by a couple of cm on each side but it shows there's little need to implement yet another standard in the PC space that would only fragment the market even more and drive prices up. You can simply optimize and shrink what you already have. There's no reason future generations of GPUs couldn't be fit onto PCIe cards and installed horizontally in a "micro" case without having to worry that the card will be useless outside of that very custom system.
Valantar - Thursday, January 19, 2017 - link
1) The R9 Nano uses HBM, which is how they got it that tiny. Are you saying SFF builds should be limited to HBM-equipped GPUs, or GPUs low-end enough to require very little space? That makes little sense to me.2) Sure, MXM was designed for low Z-height. But it mainly was designed to take up less area than standard PCIe. Would it be possible to make "desktop only" MXM cards with slightly taller (and probably cheaper) power delivery components? Absolutely. And the "less area" thing is what matters here. Again, you can get a _fully specced_ GTX 1080 in MXM form - only limited by cooling. Sure, it won't OC as well as the desktop cards, but that really doesn't matter.
3) "99% of the time you'd be hard pressed to find one that's actually speced like a desktop card." No. The entire mobile GTX 10XX series is ~identical to their desktop counterparts, with minor optimizations for power savings. Heck, the 1070 has _more_ cores than the desktop version (albeit at lower clocks). Performance differences stem mainly from lack of proper cooling. Which would be a very solveable problem in a case like this.
4) Sales for PCs are declining, sure. Sales for gaming PCs in every single possible form factor is growing. Including tiny "console replacement" PCs.
close - Thursday, January 19, 2017 - link
1) The GTX 1080 is so good in mobile devices because it's a very efficient chip (performance per watt). Are you saying SFF builds should be limited to very efficient chips? Is this your argument? o_O How can limiting them to MXM boards sound better? MXM has virtually 0 availability for desktops right now and even for laptops it rarely met its purpose. How long do you think it will take for high end GPUs (GTX 1080 equivalent) to have HBM? How long before MXM becomes mainstream and affordable?2) It's very possible to make a standard PCIe card much smaller. And that's exactly the point. Why work with yet another standard that's incompatible with almost everything in the desktop space right now? With PCIe you'll be able to move your tiny card in a full size PC should you want that.
3) Absolutely identical save for the lower clocks and the fact that TDP limitations will always force that chip to drop clocks even further when used in a thermally constrained environment (like a laptop or typical MXM cooling). Long story short there is no practical MXM implementation right now that can compete with a desktop card implementation because of its intrinsic limitations.
4) How many games are sold is irrelevant. I was talking about a discreet GPU market that becomes relatively small and then you go on and split it between two completely incompatible formats: PCIe and MXM. Every other format goes for universal standardization (at every layer) so why should graphic cards be an exception? Just for the sake of exotic hardware?
There's a tradeoff to be made here and it might not be worth it when it comes at the price of compatibility. Memory modules and SSDs can be made even smaller but they're not for compatibility's sake. You have the wrong answers because you're asking the wrong questions. The first one should have been "why invest into building a whole new hardware ecosystem instead of optimizing the one you have?".
Valantar - Friday, January 20, 2017 - link
1) Yes. Or, well, I'm not arguing that they _should_ be, I'm simply pointing out that due to the laws of thermodynamics and/or current cooling technology, SFF builds _are_ limited to one of three things: low-power GPUs, very efficient GPUs, or some sort of water cooling implementation. I'd LOVE for my Fury X to fit in a tiny chassis (heck, I'm moving it into a custom loop in an ITX case some time this year). But it can't. Below a certain volume, displacing >250W of heat simply becomes impractical.2) Sure. But that would drive costs up massively - after all, miniaturization is as much of a reason for the cost of MXM cards as scarcity. SFF costs more, and will always cost more. I for one am willing to pay somewhat of a premium for that. Also, MXM _is_ PCIe. Making a PCIe-to-MXM adapter in the case of proliferation of these cards would be no problem whatsoever. Gigabyte already has one in their BRIX UHD. This would make for _more_ standardization and interchangeability, not less - you could migrate the GPU from your laptop to your desktop, or vice versa. Upgrade either, or both. Keep, and reuse, a part that would previously go to waste. And so on.
3) Lower clocks? Only due to TDP limitations. Nvidia is quite flexible with this, but they don't want people to fry their laptops simply because they've heard that overclocking is cool af. In a case like this, with sufficient cooling, there's no reason for the board to deliver equivalent clocks to stock cards. Will the power delivery handle similar OCs to a full PCIe board? Probably not. But neither would the power delivery on your proposed tiny PCIe card.
4) Who mentioned games? I sure as f*ck didn't. Please read before you comment. I said gaming PCs. Gaming PC sales are _rising_. Sure, AIB GPU sales are droppiing off, because sales of low-end (think GT 730, Radeon R7 240) are tanking. Cards like that used to represent HUGE sales numbers (with near zero margins, but nvm), but are now being replaced by capable iGPUs. Sales of midrange and high-end GPUs are either stable or rising. Sales numbers for >$200 GPUs are massively up the last 5-10 years.
And no, I'm not asking the wrong questions here. If anyone is, you are. I'm talking about _increasing_ compatiblity. MXM is currently irrelevant because the market for gaming laptops used to be tiny, and when laptop makers had to design the graphics card custom any way, they might as well integrate it onto the motherboard. This is no longer the case, although integration is still done to a large degree for the sake of miniaturization (as thin-and-light gaming laptops are very popular). Still, the popularization of good gaming laptops could quite easily reinvigorate the MXM standard. Also, of course, the MXM-SIG needs to step up its game.
But, you might ask, how can MXM cards increase compatibility? That's pretty easy.
-They can be adapted to PCIe with relatively simple, low-cost adapters (that can carry over between generations, unless you need new display outputs). As mentioned, see the Brix UHD.
-They allow for standardized coolers. You keep arguing for compatibility, but ignore the fact that GPUs are a free-for-all when it comes to cooling solutions. I'd love to see you fit a high-end cooler from the previous generation to a current-gen card, even from the same maker. Not a chance. Want custom cooling? Great, you can replace that every upgrade too! The clusterf*ck that is GPU coolers would benefit greatly from standardized component placements. A market where GPU coolers can be chosen like CPU coolers can is a market I welcome with open arms. And yes, this can of course be done with standard PCIe cards. But there's no incentive to do so, as added components (bigger power delivery and such) is considered a value-add, and standardized component placements could jeopordize this. In the SFF space, the incentive is there, as there's limited room for huge coolers, and fitting the cooler to the case is a more pressing issue.
-They increase interchangeability. A single card could feasibly fit both laptops, SFF PCs and regular desktop PCs.
Also, the PCIe form factor standard places pretty severe limitations on how small you can make cards.
-IO slot sizes are standardized, and not optional. You're stuck with either HH slots or FH slots, and both have bulky and awkward form factors and mounting mechanisms for SFF. This isn't surprising, since this aspect of the standard is made to be compatible with AGP, PCI, ISA (and so on) slots, with dimensions worked out in the 80s and early 90s meant for servers and full-tower desktops.
-PCIe power delivery is big and clunky, and has no place in SFF PCs. Of course, power connectors are adaptable. But for SFF use, a smaller standard (with higher quality requirements to make up for a smaller connector) is pretty much necessary.
And the answer to your last question is: Because the one we have is reaching its limits in a number of significant ways*. This _should_ prompt a re-think. But then again, dumb people kept using Windows XP for more than a decade. I guess you can't beat'em all.
*With the caveat that this isn't a new hardware ecosystem, but rather a compatible, adaptable, optimized implementation of an already-open standard.
close - Tuesday, January 24, 2017 - link
And here I was thinking the point couldn't have been made clearer. There's some dissonance in what you're saying. Like how expensive it is to shrink a PCIe card just a little while being able to maintain compatibility with every desktop out there but thinking that making an almost inexistent standard in the PC market popular with adaptors and reengineering everything on the market must somehow be dirt cheap...In tomorrow's feature: microMXM for USFF, EMXM for larger but not so large builds. Also microSODIMM, microATX power connector, micro USB headers, micro M.2 slot, micro screw holes, and why not... micro CPU socket. Some of there require almost 0 re-engineering so maybe breaking compatibility is a small price to pay to get a slightly smaller footprint. And I'm sure you could easily just buy some adapters to make these compatible with everything else. I mean why would it make sense to do this with GPUs but no other component?
P.S. Yeah, not a completely new hardware ecosystem... just one that barely registers on any radar even in the laptop market let alone the desktop market where I think it exists exclusively in a handful of OEM builds.
But hey, if it sounds cool and exotic why not. I remember having an Asus sk. 479 adapter for the desktop. The next big thing.
fanofanand - Wednesday, January 18, 2017 - link
Don't run your high-end system in single channel memory mode.....bigboxes - Wednesday, January 18, 2017 - link
Just stick to your tablet.TheinsanegamerN - Thursday, January 19, 2017 - link
and what tablet would get anywhere close to a 1080's performance?If this takes off, we could have a whole range of NUC sized mini boxes with serious hardware. That would be awesome.
guidryp - Wednesday, January 18, 2017 - link
I am definitely in favor of some new form factors with GPU support. Mini-ITX leads to issues. The way the GPU hangs off the edge in points down (in vertical cases), means either:You make the case a lot bigger to have space, or the GPU is cramped for air, or you put in a giant dust admitting cutout to help with cramping. I don't particularly like any of these options.
So if I was building today, I would either stick with Micro ATX if using a GPU, or skip GPU and go with an ultra small case.
There just doesn't seem to be a good small + GPU form factor option that I like. There are some kludgy horizontal adaptors for the Mini-ITX slot, but they leave the GPU point the CPU at the bottom of the case, so again, something of a problem.
A real horizontal MXM GPU form factor that got widely used might be an answer, though I am not sure this is exactly the right one. I am thinking more Mini ITX + MXM, and don't mess with the standard CPU area to shave a few mm.
xenol - Wednesday, January 18, 2017 - link
This would be great if we can find MXM boards for a decent price, rather than say $800 for a GTX 970M or 980M on eBaynathanddrews - Wednesday, January 18, 2017 - link
I agree, it would be great. If more boards come out with this type of connectivity, maybe they will drop in price as more people buy them?TheinsanegamerN - Thursday, January 19, 2017 - link
it could also drop if asrock provides a single place to buy new ones. Supply would have a chance to catch up with demand.keeepcool - Wednesday, January 18, 2017 - link
1070N from MSI that will work on regular MXM3.0b slots are around 800€ on ebay and other specialty laptop mxm sellers.DanNeely - Wednesday, January 18, 2017 - link
Reading the specifications I wonder if the sata ports are done via M.2 to SATA dongles.Valantar - Thursday, January 19, 2017 - link
I'm pretty sure the SATA connectors are the two odd-looking compact connectors next to the m.2 slots. They look similar to connectors like this seen in NUCs and other UCFF PCs that combine SATA+power into a single low-profile connector.Cygni - Wednesday, January 18, 2017 - link
"Personally, I’d prefer a solid copper cooler on there for a Core i7, but a Core i5/i3 would do OK with the stock cooler."Of the 7 Skylake and Kaby Lake i7's in intel's desktop stack, 5 are 65w or below. Perhaps this sentence would be better worded by talking about K-series CPU's, instead of focusing on i7 vs i5/i3, which isn't very relevant in 2017.
cilvre - Wednesday, January 18, 2017 - link
The 2 sata ports are to the left of the m.2 slots. it looks like a proprietary cabling similar to laptop cables that lock down onto the pcb.Valantar - Wednesday, January 18, 2017 - link
I have to say this looks really, really good. Of course, the board could do with better power delivery. But it's still pretty awesome. With some better cooling, you could have a coffee table book-sized i7-7700K+GTX1080 PC. That would be amazing. Also, upgradeability and flexibility makes this very interesting as a concept.spikebike - Thursday, January 19, 2017 - link
Looks ok, but not sure the use case. Wouldn't any of the low profile and/or short gtx1060 and related cards manage this in a one of the small cases (mini-itx, micro-atx, and related)? MxM seems pretty expensive and limited options compared to the availability new smaller cards that are in wide availability from multiple vendors at approximately $0 price premium.TheinsanegamerN - Thursday, January 19, 2017 - link
well, a short 1060 isnt exactly in the same ball park as a 1080. You would also have a lot more z height to deal with.A board like this will be super useful for NUC style HTPC and gaming PCs. I know I'd love one for my TV. Between this and zotac's e boxes, there is a market for slim gaming pcs that are book sized.
xyf - Thursday, January 19, 2017 - link
I had a similar idea in mind. Have a socket for the GPU (analog to the CPU socket) and the GPU being a single chip like (would only work GPUs with HBM memory). The video output can be on the MB (pci-e or something specific for the connection). This way the CPU and GPU can share the power delivery, and can use a single cooling system that gets on top of both. Then again, just join the 2 and get an APU...DanNeely - Thursday, January 19, 2017 - link
I think you'd still run into compatibility problems due to different voltage needs from one GPU generation/family to the next that would limit future compatibility in the same way that LGA1156/1155/1150/1151 have meant you're unable to keep swapping in new generations of processor despite CPU IO mostly staying the same from generation to generation.TheinsanegamerN - Thursday, January 19, 2017 - link
so, instead of having to swap mobos every 5 years for a new CPU, I'd have to swap them every time I want a newer GPU generation?We dont need more forced obsolescence built into our stuff.
BrokenCrayons - Thursday, January 19, 2017 - link
The desktop market needs this soo badly! Full ATX and even mATX is just huge and awkward. We've come so far at making things smaller and faster, why are we still hanging on to full length PCIe expansion cards that are wasteful of materials while encouraging GPU companies to keep bloating up on TDP despite getting access to new manufacturing processes.TheinsanegamerN - Thursday, January 19, 2017 - link
Agreed. This is like the zotac e boxes, but more upgradeable. I'd love one for the TV as a HTPC gaming rig.Valantar - Friday, January 20, 2017 - link
Agreed. The MXM-SIG should really step it up - this time period is a golden opportunity for them, which they seem to be wasting. The Zotac E boxes, Alienware Alpha and so on would really benefit from the upgradeability that MXM could bring to them.ruthan - Thursday, January 19, 2017 - link
Can normal mortal buy a MXM modul or its still for OEM gods?Lolimaster - Thursday, January 19, 2017 - link
High end AMD APU's pretty much kill this attemps.If AMD wanted. they could release a 1536SP+4c/8t 125w Ryzen based Raven Ridge APU. Or a 2048-2304SP one using the 7nm.
BrokenCrayons - Thursday, January 19, 2017 - link
The trouble with APUs and similar Intel non-Iris iGPUs isn't reall a lack of graphics processing power. There's a much more substantial bottleneck in contending for limited access to relatively slow system memory. Whereas modern GPUs with their own memory have a big dedicated pipe (even the lowly GTX 950 has over 105 GB/s), iGPUs end up dealing with less total bandwidth and contending with the CPU plus other components for a slice of that smaller pie. That can have adverse effects on the texture sizes, maximum playable resolutions, and everything else regardless of how powerful the iGPU is. Intel's eDRAM is tiny, but offers 50 GB/s of dedicated bandwidth to mitigate the problem. Modern texture compression can help too, but that only gets you so far before you end up starving the GPU cores. Then there's the problem of a shared thermal budget and close proximity of numerous hot running CPU cores that would otherwise be on an entirely different IC somewhere else on the motherboard.Besides that, the current top end AMD A10 has 512SP so though it's a pretty good iGPU, it's not competitive with a MXM 1060. We'll have to wait and see what Ryzen brings to the table, but I think even with a node shrink to 14nm, there's just not going to be a big enough leap to make a Ryzen APU into a GTX 1060 killer even whent hat 1060 is crammed into a MXM card in a small form factor case.
Valantar - Friday, January 20, 2017 - link
7nm? Sure, in 2019 or 2020. Have fun with your APU in 2-3 years. I guess you'll be playing Super Mario Run on your phone until then?Also, 125W is low for anything supposed to perform well with a mid-range GPU plus a 4c8t CPU on board. Remember that the more densely you pack your heat generating components, the harder it becomes to keep them all operating within reasonable temperatures. Besides, you'd still need a rather huge cooling solution for this. Nvidia has a decent efficiency lead on AMD (for now), but a laptop GTX 1060 + a 45W i7 should be comparable to this thermally. Laptops with this need pretty hefty coolers - and that's with _two_ sizeable dies across which to run heatpipes. With just one die, this becomes significantly harder.
I'm not saying I'm against this - not at all, I'd love for AMD to make a high-end APU like this. I just don't see it as quite feasible yet.