-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
SFF.Network ASRock Z270M-STX MXM Micro-STX Motherboard Pictured
- Thread starter jeshikat
- Start date
re: the 3xM.2 slots... y'know what would be fun? 3x960 Pros in Raid 5. (Or Raid 0 if you naasty...)
Could you safely do a 6 pin PCIe to JST adapter using say 20 or 22 AWG wires (I assume 18AWG wouldn't work with the small connector type). I'm just thinking forward to how we'd power the 1070/1080 on this unit.
Actually, the first pic for "1080 MXM" shows a 4-pin but looking more shows that the 1080 seems to actually have a 8-pin power connector. And they're colored differently so I'm not sure if it's just pure 12V or a mix of 12V, 5V, and 3.3V, or some other voltage entirely.
If we can get a 1070 on there I'll buy it. I haven't been this excited about a product release in years. I've signed up to these forums just to discuss this product.
Not to keep hyping it, but if we can do 8 pin or SATA to the 1080 or 1070 connector the DC-DC board I'm developing with Larry will power this very easily.
I don't know how famous we are, but thank you anyway
...I'm afraid we'll have to hold off until we see some traction in the market. I can only hope that you and other established companies like Silverstone have some success with the form factor, and we do see some aftermarket develop in time.
I share Necere's concerns, but I tend to be the more optimistic one between the two of us.
Necere's responsible for design, I'm responsible for everything else (business, mfg, logistics, support and etc.). The challenges with developing around this platform (price, moq, availability, support) are all areas that I would have to deal with.
To build up the market for this platform = indie case designers have to sell a reasonable volume = there is price ceiling for the barebone kits, which relates to MOQ = upfront costs. None of us have built a new market before, we just sell cases within an existing market.
I guess I'm more interested in moq and pricing for both the motherboard and MXM cards. Again, I believe the challenge of building around this platform hinges on selling in volume. If the economics works out to be something us indie case makers can handle (crowdfunding, investor maybe), then we can try.
I know some of the information I'm asking may require a NDA, so I have also sent you a PM in case we need to move to an email conversation.
Last edited:
Here are some problems that I have with MXM.
In the most situation you can't use a MXM card of a different vendor on your MXM board.
MXM was not designed to upgrade your system it was designed to make it easier for a vendor to support differen GPUs in one Notebook series. So the vendor does not have to make a PCB desing for every different GPU.
But what are the problems with using a MXM board from e.g. Alienware or Clevo on this Asrock STX-MXM board?
1. The Motherboard have to know the ID of the card and have to know the fan temp sensore on it. This is a big problem in partice because Alienware use a different temp sensor on their board as Clevo in the same series. For MXM the fan header is on ther motherboard and not on the MXM board so the motherboard will control the fan. If the board can not regocnize the temp sensor the fan will spin at 100%.
2. Nvidia and AMD have chips that control video output between MXM and iGPU of the CPU. So if there are changes on this it could be that newer cards will not work with your board. This happend in the past with the switch form AMD HD 6970m to 7970m with the "Enduro" functions.
3. Plug and Play will be not possible if your motherboard did not know the card but why? The AMD and NVIDIA driver will get the vendor and gpu id through the motherboard bios. If the bios did not know the card you have to mod the gpu driver. Otherwise you can't install it. This is the reason why you can't use for many vendors the drivers that are on the nvidia website and you have to use the driver on the vendor website.
4. There are often changes to the MXM layout (height of mosfets). So there will be no heatsink that will work also with cards in some years. If you look on therinferno there are hundrets of mod guids how to mod the heatsink that it will support the new gpu generation.
5. MXM boards are risky because you have voltage control, RAM and GPU on a very small board. The operating temp is very high. So the faulty rate could be high.
6. Upgrading cards is dangerous for customers, because they can easily destroy something. You have to put the MXM board in a perfect angle into the port and then press it down. If you do it not right the card could be mounted incline, that can destroy the card. Furthermore you can't install a card while the heatsink is on it. So the customers have to paste it and put heatpads on it.
So you see MXM is not as easy to handle like the most of you think. It is different from what we know with PCIe. You can use a PCIe Gen3 card in a Gen1 slot. Try to use a GTX 1080m in the first MXM 3.0 generation boards. The most mxm boards only support newer cards because there are bios modder like the legendary user Prema. (https://biosmods.wordpress.com/)
In the most situation you can't use a MXM card of a different vendor on your MXM board.
MXM was not designed to upgrade your system it was designed to make it easier for a vendor to support differen GPUs in one Notebook series. So the vendor does not have to make a PCB desing for every different GPU.
But what are the problems with using a MXM board from e.g. Alienware or Clevo on this Asrock STX-MXM board?
1. The Motherboard have to know the ID of the card and have to know the fan temp sensore on it. This is a big problem in partice because Alienware use a different temp sensor on their board as Clevo in the same series. For MXM the fan header is on ther motherboard and not on the MXM board so the motherboard will control the fan. If the board can not regocnize the temp sensor the fan will spin at 100%.
2. Nvidia and AMD have chips that control video output between MXM and iGPU of the CPU. So if there are changes on this it could be that newer cards will not work with your board. This happend in the past with the switch form AMD HD 6970m to 7970m with the "Enduro" functions.
3. Plug and Play will be not possible if your motherboard did not know the card but why? The AMD and NVIDIA driver will get the vendor and gpu id through the motherboard bios. If the bios did not know the card you have to mod the gpu driver. Otherwise you can't install it. This is the reason why you can't use for many vendors the drivers that are on the nvidia website and you have to use the driver on the vendor website.
4. There are often changes to the MXM layout (height of mosfets). So there will be no heatsink that will work also with cards in some years. If you look on therinferno there are hundrets of mod guids how to mod the heatsink that it will support the new gpu generation.
5. MXM boards are risky because you have voltage control, RAM and GPU on a very small board. The operating temp is very high. So the faulty rate could be high.
6. Upgrading cards is dangerous for customers, because they can easily destroy something. You have to put the MXM board in a perfect angle into the port and then press it down. If you do it not right the card could be mounted incline, that can destroy the card. Furthermore you can't install a card while the heatsink is on it. So the customers have to paste it and put heatpads on it.
So you see MXM is not as easy to handle like the most of you think. It is different from what we know with PCIe. You can use a PCIe Gen3 card in a Gen1 slot. Try to use a GTX 1080m in the first MXM 3.0 generation boards. The most mxm boards only support newer cards because there are bios modder like the legendary user Prema. (https://biosmods.wordpress.com/)
So what you're saying is that MXM would have to be more tightly standardised in order for this to work? I mean, ASRock could still sell MXM cards for their system, but would it even be possible to get to a point where MXM is as flexible as PCIe as an interconnect? Or is MXM in general completely different? I just assumed it was PCIe with a different physical connector and additional pins for temperature sensing and higher power draw.
Yes it is different. E.g no mxm card has the video i/o on the mxm board it is on the mainboard. So you are limited on that what the motherboard provide. So the mxm slot includes also the video stream back to the motherboard.
The problems with MXM are not completely insurmountable: most of the BIOS-related issues could be resolved by the motherboard itself being designed to accept a wider range of cards and passing the correct IDs (the ASRock board already accepts at least 3 cards from two vendors), and the vendor committing to support that. With laptops, the vendor is incentivised to not only not provide support, but actively attempt to prevent user upgrading of MXM modules (because every MXM upgrade is a lost opportunity to sell a whole new laptop).
Actual board-level changes are a more difficult problem. The confines of even a small chassis are larger than that of a laptop, so a more discrete heatsink (e.g. large heatsink for the die, small individual heatsinks for the VRMs and other components when needed) would be viable to account for component changes.
I/O differences would be the most difficult to handle: if a vendor implements a different communication method after launch (e.g. if Alienware decides that they don't want their MXM modules to work with the ASRock board going forward) there's little recourse. If the change is down to protocol or signalling rather than a more radical electronic change (e.g. using different pins entirely) then this could be updated via BIOS, otherwise an 'approved MXM module list' would still be needed. Consumer pressure on MXM designers could have some effect here, but only really on MXM vendors who do not also have laptop distribution arms.
Actual board-level changes are a more difficult problem. The confines of even a small chassis are larger than that of a laptop, so a more discrete heatsink (e.g. large heatsink for the die, small individual heatsinks for the VRMs and other components when needed) would be viable to account for component changes.
I/O differences would be the most difficult to handle: if a vendor implements a different communication method after launch (e.g. if Alienware decides that they don't want their MXM modules to work with the ASRock board going forward) there's little recourse. If the change is down to protocol or signalling rather than a more radical electronic change (e.g. using different pins entirely) then this could be updated via BIOS, otherwise an 'approved MXM module list' would still be needed. Consumer pressure on MXM designers could have some effect here, but only really on MXM vendors who do not also have laptop distribution arms.
Does that really matter, though? You can't get Alienware MXM modules anyway for a sane price, so nobody would really care whether they do work with "open MXM" boards, and vendors that do want to sell MXM to the consumer directly (like ASRock) would care greatly about mutual support between their cards and boards, so they'd hopefully put effort into exactly that.
The most important step is to establish one good standard before someone gets the idea that they could restrict users to their ecosystem by only supporting their own cards. Once two manufacturers components are compatible with each other, the price you pay for going into a locked system is much higher.
I just had a weird thought... but would an MxM module be able to be put into crossfire with a standard GPU? ...
There's examples of laptops with 2 MXM modules in SLI, but i don't think the odds of combined PCI and MXM are great. The DX12 MultiGPU standard could be a safer bet, though.
I need someone to test this... for science. If memory serves me correct, only PCIe 3.0 x4 is needed for crossfire. MxM 480 + 2x standard 480s... Brought together by a lovely 300w hdplex setup... Yes its lucrative and stupid, but it could be done. 2 GPUs laid against the motherboard on the back with PM4S2 boards and a few extensions, hdplex between them, and 2 power bricks for the whole system (1 for motherboard, 1 for GPU). But it might be possible to get a case to around 5-9L... a 5-9L Tri-Fire setup. Its so crazy I want to do it just to try it.
A lot of those incompatibilities with MXM is down to the manufacturer, really.
On the soft/firmware side, some mnufacturers use custom vBIOS on the motherboard keyed to the id of the GPU or have some other such system that prevents the system from working without a GPU specifically for that product line. Oftentimes this is because they use proprietary MXM modules and are more interested in the ease of manufacture than end-user maintenance and upgrade, though I am sure some do it to force people to buy a new system or at least buy their upgrades. It's fully possible for the motherboard to be GPU agnostic where the BIOS is concerned.
As for heatsink mounting, it's true that they tend to change the keep out zones and component height limitations every so often as a new revision, and different revisions aren't guaranteed to be compatible, but it should still be feasible to support a fair range of GPUs per revision. Again, where incompatibilities are concerned, it tends to have to do with how it's manufactured. There's some systems that can take a good range of GPUs with the heatsink it came with, there's others that have to have a specific heatsink for each different module. Again, this tends to involve the use of proprietary modules that are going to require a specific heat dissipation solution, but also to the tight confines of a small chassis where getting a universal heatsink to fit is a challenge..
As far as connectivity, there can be SOME difficulty there. The standard has signal lines for 4 Dual mode DP ports, 1 LVDS port and 1 Analog port. As an option the GPU module can support DVI over the LVDS, gang the DP ports for dual-link DVI, and run stuff like S-Video and SCART over the analog port, and it can run 4 DP ports or put DVI and HDMI ports on the DP lines (since they are dual mode), but the only thing that's actually required is for the GPU to support is 2 DP, 1 LVDS and 1 VGA. That does have the potential to interfere with upgradability, but you're at least guaranteed that the motherboard has access to the three external ports and an integral display.
On the soft/firmware side, some mnufacturers use custom vBIOS on the motherboard keyed to the id of the GPU or have some other such system that prevents the system from working without a GPU specifically for that product line. Oftentimes this is because they use proprietary MXM modules and are more interested in the ease of manufacture than end-user maintenance and upgrade, though I am sure some do it to force people to buy a new system or at least buy their upgrades. It's fully possible for the motherboard to be GPU agnostic where the BIOS is concerned.
As for heatsink mounting, it's true that they tend to change the keep out zones and component height limitations every so often as a new revision, and different revisions aren't guaranteed to be compatible, but it should still be feasible to support a fair range of GPUs per revision. Again, where incompatibilities are concerned, it tends to have to do with how it's manufactured. There's some systems that can take a good range of GPUs with the heatsink it came with, there's others that have to have a specific heatsink for each different module. Again, this tends to involve the use of proprietary modules that are going to require a specific heat dissipation solution, but also to the tight confines of a small chassis where getting a universal heatsink to fit is a challenge..
As far as connectivity, there can be SOME difficulty there. The standard has signal lines for 4 Dual mode DP ports, 1 LVDS port and 1 Analog port. As an option the GPU module can support DVI over the LVDS, gang the DP ports for dual-link DVI, and run stuff like S-Video and SCART over the analog port, and it can run 4 DP ports or put DVI and HDMI ports on the DP lines (since they are dual mode), but the only thing that's actually required is for the GPU to support is 2 DP, 1 LVDS and 1 VGA. That does have the potential to interfere with upgradability, but you're at least guaranteed that the motherboard has access to the three external ports and an integral display.
These were my favorite items out of CES. Someone posted about them at hard, which is what made me come here and register.
I hope we get to see some more STX boards utilizing Z270, multiple m.2, and MXM. It will be an SFF dream if we start seeing more MXM and hopefully quieter cooling solutions for them.
I hope we get to see some more STX boards utilizing Z270, multiple m.2, and MXM. It will be an SFF dream if we start seeing more MXM and hopefully quieter cooling solutions for them.
Before we have quiet MXM coolers, we need them to become widely adopted by slightly larger systems. Being used in small systems means you have a small fan and small heatsink, so much of the noise comes from how fast the fan has to spin to cool effectively.
Similar threads
-
- Article
