Concept Flex-ITX motherboard form factor concept

[Analogue Blacksheep]

@duynguyenle - Asus H110M-PIO?

 

[duynguyenle]

@duynguyenle - Asus H110M-PIO?

Very similar, yes. I think the board I saw was a B360 chipset not H110, but same concept

 

[mtacx]

nice..cant wait to see the finished product

 

[pwnedundies]

Looks neat, but I worry that various sandwich layout cases in the future
will have to be set on the PCIE riser PCB length without versatility.

Not really. These all seem to be adapters. So you could use the stand adapter and then use any riser that floats your boat.

As for the interested riser you would absolutely be stuck with the specific design due to the riser being at the top and causing the GPU to be upside down.

I'm not sure if it's a possible configuration but seeing essentially an "U" configuration would be interesting as well. Instead of running the riser up and behind to make it clean run in down and straight back and straight up to the GPU.

Basically this but for the pcie slot. I think this would be very clean and still accomplish what you need.

GPU VGA 8Pin 6Pin PCIE U Turn 180 Degree Angle Connector Adapter Board

Buy GPU VGA 8Pin 6Pin PCIE U Turn 180 Degree Angle Connector Adapter Board for $4.99 with Free Shipping Worldwide (In Stock)

www.moddiy.com

 

[pwnedundies]

@Necere Interesting concept, however where are you planning on putting the m.2 ports? Part of the reason why is if you have one on the back and you have a riser cable covering it wouldn't that cause thermal issues for the m.2?

When I still had the ZS-A4 DC, I spent some time looking at ways of moving the rear mounted M.2 away to another part of the case. I did look at an M.2 to OCuLink based solution but I never got the chance to test it as my motherboard didn't fit the case. There are also cables like this m.2 to u.2 cable which can also work. @fabio used one in his MI-6 build if you are looking for examples.

Then again, if we had some more motherboards with some native OCuLink or SAS ports like the recent Asrock Rack X570-I board that could solve the problem.

M.2 should be on the front above the pcie slot still. Especially in a sandwich style case, if you have fans, you'll get more than enough airflow on the drive.

If you do want to put them on the back of possible put it parallel to the riser cable and closer to the one end of the board. Unless it's an nvme any fan will be able to keep it cool. Having it parallel allows for the entire drive to get air pulled over it without interference from the riser.

I had a Ghost and a Steck with a perpendicular sata m.2 and as long as I had the offset fan in the bottom it was still enough to easily stay in the 40's°c

 

[Brude27]

I just woke up, so I don’t have any feedback to offer at the moment, but I will say I think that this could be incredible! I will be watching this one closely, and will read it again once I’ve had coffee, haha.

 

[smitty2k1]

I think that's a neat little idea you go there. Any indication that Intel is doing anything with custom PCBs for the new NUC form factor they teased earlier this year?

 

[arthasdklol]

Very interesting idea. Ideally I would love to standardize the crap out of this design in terms of socket, IO and RAM positions/dimensions, so that a vendor can create an air cooler that covers the complete motherboard
I think there is a potential for some extra optimizations by using m2 slots on the back of the mobo and the case's spine as a giant heatsink...

 

[Thehack]

I do not like this current iteration as it doesn't really push the envelope forward. The increased cost in R&D has to have a net benefit.

1. Ncase M1 would be negative benefit since you have increased board cost + mandatory right angle adapter.

2. Sandwich cases you still have to run a backside flexible riser. Higher costs overall.

3. It only shaves a few mm off sandwich cases to clear the flexible riser.


What I would like to see:

- AIO board: opposite side horizontal pcie slot

- customized power solution using new Intel standards

- customized layout with triple m.2 by getting rid of all the SATA ports (2x full pcie m.2, and 1x short sata m.2).

- custom design with no useless heatsinks and better SFF layout

- 6x6 custom design with full slot, plenty of IO but laptop memory instead.

- option for slimmer rear IO size using USB-C to reduce size

- 12v only capable input but still atx compliant

I am open to also collaborate if you need some help dishing out these features.

 

[chyll2]

customized layout with triple m.2 by getting rid of all the SATA ports (2x full pcie m.2, and 1x short sata m.2).


This please. We need to free up sata ports space and use it for something else.

 

[arthasdklol]

customized layout with triple m.2 by getting rid of all the SATA ports (2x full pcie m.2, and 1x short sata m.2).


This please. We need to free up sata ports space and use it for something else.

SATA ports don't use that much space, you would be better off by switching 24->10pins as it decreases the cabling, routing headers and other cables "smarter" and making more m2s on the back of the mobo (I suspect routing traces will be an issue and pcie slots available if you are on intel )
+
reducing/removing all the weird stuff nobody uses anyway, such as super-duper soundcard with headers, weird bulky IO section...
Just leave 4 USB-A, add couple of USB-C/thunderbolt, wifi, ethernet and call it a day on the IO part (+one small usb header for the front panel)

 

[Thehack]

SATA ports don't use that much space, you would be better off by switching 24->10pins as it decreases the cabling, routing headers and other cables "smarter" and making more m2s on the back of the mobo (I suspect routing traces will be an issue and pcie slots available if you are on intel )
+
reducing/removing all the weird stuff nobody uses anyway, such as super-duper soundcard with headers, weird bulky IO section...
Just leave 6 USB-A, add couple of USB-C/thunderbolt, wifi, ethernet and call it a day on the IO part (+one small usb header for the front panel)

Yeah, the foot print of sata port wouldn't translate necessarily to a short m.2 but the point really is to get rid of unnecessary stuff and rearrange the layout to fit what we want.

 

[DrHudacris]

SATA ports don't use that much space, you would be better off by switching 24->10pins as it decreases the cabling, routing headers and other cables "smarter" and making more m2s on the back of the mobo (I suspect routing traces will be an issue and pcie slots available if you are on intel )
+
reducing/removing all the weird stuff nobody uses anyway, such as super-duper soundcard with headers, weird bulky IO section...
Just leave 4 USB-A, add couple of USB-C/thunderbolt, wifi, ethernet and call it a day on the IO part (+one small usb header for the front panel)

SATA ports don't take up much room on the motherboard, true, but the drives and cables for the drives connected to them do take up a lot of room. I would be in favor of a short m.2 SATA over traditional SATA ports.

 

[gwertheim]

Not only that I would keep optical out for a clean audio output. You can get a cheap dac/amp combo to convert it for headphones. Also the wifi antenna would be nice to keep

 

[arthasdklol]

Yeah I'm looking at X570 Aorus https://www.aorus.com/upload/Album/2019062610315505ff4a5d2efcf8b626b8162eb7c933cccb_big.png and main offenders are:
- ATX 24pin plug (dream case scenario: replace with 10pin)
- extra 8 pin in the corner (remove?)
- (potentially) 2 SATAs to be removed (I would like to keep 2 SATAs, there are sometimes weird use cases like 2,5inch 4tb ssd 'cold storage' someone might want to have...)
- socket cooler mounts (can use different standard)
- chipset shroud (potentially remove, but it's also used by m2 as a heatsink)
- IO shroud (remove shroud, leave wifi antenna, 4 usb-a, ethernet, potentially keep one optical audio out and/or add 2xusb-c/thunderbolt[3/4])
- random headers (remove)

There are some elements that have a dual purpose, mostly cooling (VRM, m2 & chipset), that could be removed if the end user had provided some other means of cooling (fan on the bottom etc). What I worry about is that a lot of the elements we would want to standardize (VRM&socket location/size) or get rid of are the main differentiators between motherboard models and/or manufacturers...

 

[nulio]

Amazing initiative and great comments so far!

I don't enough on the topic. I've seen some concepts that would love having the MB and GPU, boards and I/O, both facing the same way (either being parallel or still perpendicular with each other). I think this would be a must to be really "Flex", maybe with a second PCIe edge connector on top. This would probably need the simplified I/O on the board and some other things.

It'd also allow the sandwich cases to just use an U connector.

 

[gwertheim]

The atx 24 pin, what does it power?

It might be something which is critical to the operation of the whole board.

Also it might help if we had an electrical engineer with some background in motherboard design to help us seperate the pie in the sky from the down to earth ideas.

We are talking about stripping down a tesla with all the electronics into a racecar.

We need to consider the core components which are critical parts to keep the motherboard running like any other. You can't just rip the autopilot or the seats out and expect everything to work as normal.

 

[Necere]

Interesting concept. The PCIe riser for sandwich case are hard/pcb?

Yes.

Love the idea. Is there a benifit in adopting a top mounted PCIE (rather than conventional bottom mounted) using the flexible mount approach? A bit like the PIO boards use? This would allow both CPU and GPU fans to face the same way.

Maybe, but the advantage with this concept is that it's fully compatible with existing standards.

would this be possible in console style layout with cpu/gpu facing same way?

i think this would actually be revolutionary, and as long as you got it working no problem with pcie4 (one benefit of pcb risers i can think of is being able to add retimers if needed) everyone should be on board, maybe even oems!

Yes, that would also be possible using a long PCB riser. Good point on being able to add retimers to a PCB riser - another point in favor of this concept over mini-ITX + flex risers.

@Necere Interesting concept, however where are you planning on putting the m.2 ports? Part of the reason why is if you have one on the back and you have a riser cable covering it wouldn't that cause thermal issues for the m.2?

When I still had the ZS-A4 DC, I spent some time looking at ways of moving the rear mounted M.2 away to another part of the case. I did look at an M.2 to OCuLink based solution but I never got the chance to test it as my motherboard didn't fit the case. There are also cables like this m.2 to u.2 cable which can also work. @fabio used one in his MI-6 build if you are looking for examples.

Then again, if we had some more motherboards with some native OCuLink or SAS ports like the recent Asrock Rack X570-I board that could solve the problem.

This form-factor concept isn't intended to define the location of board components, just the PCIe slot. The confined space/lack of cooling for rear-mounted M.2 drives are an inherent issue with sandwich layouts regardless of how you go about it, and they should either be avoided or special care taken by the chassis designer to ensure adequate airflow between the motherboard and back of the GPU.

I believe this already exist (motherboard with edge-aligned PCI express connector). I can't seem to find the link anymore, but I remember seeing one just like the concept you proposed. It's more or less a standard form factor except with the pcie connector mounted on the edge in-line with the plane of the board instead of otlrthogonal to the board surface.

IIRC its quite a popular form factor for Chinese Internet cafes because it allows for thinner case width (allowing them to cram more PC booths into the same floorspace)

I'll try and find that posting again after work

PIO boards are only usable in cases designed specifically for PIO boards. The point with this Flex-ITX concept is it can be used in any existing cases, as well as standardizing and bringing lower cost and higher reliability to alternative layouts.

Not really. These all seem to be adapters. So you could use the stand adapter and then use any riser that floats your boat.

As for the interested riser you would absolutely be stuck with the specific design due to the riser being at the top and causing the GPU to be upside down.

I'm not sure if it's a possible configuration but seeing essentially an "U" configuration would be interesting as well. Instead of running the riser up and behind to make it clean run in down and straight back and straight up to the GPU.

Basically this but for the pcie slot. I think this would be very clean and still accomplish what you need.

GPU VGA 8Pin 6Pin PCIE U Turn 180 Degree Angle Connector Adapter Board

Buy GPU VGA 8Pin 6Pin PCIE U Turn 180 Degree Angle Connector Adapter Board for $4.99 with Free Shipping Worldwide (In Stock)

www.moddiy.com

Yes, that would be trivial to implement. However, the GPU fans would be facing the back of the motherboard, so you'd want to have a good reason for doing it vs. a back-to-back layout.

Very interesting idea. Ideally I would love to standardize the crap out of this design in terms of socket, IO and RAM positions/dimensions, so that a vendor can create an air cooler that covers the complete motherboard
I think there is a potential for some extra optimizations by using m2 slots on the back of the mobo and the case's spine as a giant heatsink...

I like a standardized CPU socket location like thin mini-ITX has, but it's probably not a good idea as it puts too many restraints on the motherboard makers.

 

[Necere]

I do not like this current iteration as it doesn't really push the envelope forward. The increased cost in R&D has to have a net benefit.

The initial R&D and tooling for the slots and risers would cost a bit, but once that's done the marginal cost of adding a single cheap riser would be negligible over the base cost of an equivalent mini-ITX board. The benefit is that it enables alternative layouts at a much lower cost and higher reliability, since there are no expensive or unreliable flexible risers required.

1. Ncase M1 would be negative benefit since you have increased board cost + mandatory right angle adapter.

Added cost would be marginal.

2. Sandwich cases you still have to run a backside flexible riser. Higher costs overall.

Check the illustration again - the riser in the sandwich layout is clearly labeled as a PCB riser. Getting away from the cost and reliability issues of flex risers is really the entire point. Maybe the color threw you off? I'll have to change it.

What I would like to see:

- AIO board: opposite side horizontal pcie slot

- customized power solution using new Intel standards

- customized layout with triple m.2 by getting rid of all the SATA ports (2x full pcie m.2, and 1x short sata m.2).

- custom design with no useless heatsinks and better SFF layout

- 6x6 custom design with full slot, plenty of IO but laptop memory instead.

- option for slimmer rear IO size using USB-C to reduce size

- 12v only capable input but still atx compliant

I am open to also collaborate if you need some help dishing out these features.

You're talking about something else entirely. I'm not trying to create an entirely new form factor, but extend an existing one (mini-ITX) to enable greater flexibility. All that stuff you've suggested is for a different discussion. I like ATX12VO, but that will develop on its own timeline unrelated to this.

 

[arthasdklol]

The atx 24 pin, what does it power?

It might be something which is critical to the operation of the whole board.

Also it might help if we had an electrical engineer with some background in motherboard design to help us seperate the pie in the sky from the down to earth ideas.

We are talking about stripping down a tesla with all the electronics into a racecar.

We need to consider the core components which are critical parts to keep the motherboard running like any other. You can't just rip the autopilot or the seats out and expect everything to work as normal.

It's so big mostly due to legacy reasons. There is this thing: https://custompc.raspberrypi.org/articles/new-psu-standard-to-launch-this-year and this thing:
edit: forgot that the reddit links point actually to this forum https://smallformfactor.net/forum/threads/m2427-cable-management-freedom.10349/page-5