SFF Threadripper motherboard petition thread

Aibohphobia

aka James
Original poster
Feb 22, 2015
4,955
4,681
I know this topic will come up repeatedly so this thread will serve as the gathering point for Threadripper in a SFF-friendly motherboard form factor. Whether that's microATX, Mini-DTX, Mini-ITX, DTX, or really just anything smaller than ATX.
 
Last edited:

alamilla

SFF Lingo Aficionado
Feb 11, 2016
119
112
Got lazy today and drew this. Come to think of it, I'd like to see if Asrock could do Threadripper on an ITX board

If they use SO-DIMM memory as suggested it's a (wishful) possibility.
They could potentially be placed on the rear of the board along with M.2 slots.

The further I learn about the X299 platform the more I become frustrated.
Reduced PCIe lanes for the lower tier SKUs, non-soldered dies, lack of ECC support, sky high prices.
I will definitely be holding out to see what Threadripper brings to the table.
 

VegetableStu

Shrink Way Wielder
Aug 18, 2016
1,769
2,444
They could only fit the single-storey flat SODIMM slot behind, but I'd think it's enough to move two of them behind (and probably the 2 m.2 slots). It's going to be a super tight build-and-hopefully-forget system, because troubleshooting the underside of the mobo might need complete removal...

I just had the thought of a second PCB layer position at roughly the height of a PCIe slot actually. Not to jinx myself, but when I get time I'll do a quick 3D mockup on what I mean
 

Dyson Poindexter

If there's empty space, it's too big!
Moderator
Jun 25, 2015
55
62
I believe SODIMM slots, even laying down are too tall for the back of a motherboard. There's a keep-out spec somewhere.

Why not "lift" the PCB by 2-4mm and have a low profile (shorter) PCIe slot? The rear IO should still fit, and the PCI slot bracket is the only other constraint. Then the rear could hold SODIMMs with no issues.

Routing ~1152 traces though a 8-14 layer PCB might be a bit of a challenge though! :p
 

VegetableStu

Shrink Way Wielder
Aug 18, 2016
1,769
2,444
I vaguely recall some STX boards doing that RAM thing though, that's my I mentioned it. I could be wrong...
 

Aibohphobia

aka James
Original poster
Feb 22, 2015
4,955
4,681
ASRock actually didn't completely dismiss Threadripper on Mini-ITX when I asked, but they made clear that there were absolutely no plans for such a thing anytime soon.

And I got the impression that they didn't foresee there being enough demand to justify the hefty R&D before something new and better came along.
 

Dyson Poindexter

If there's empty space, it's too big!
Moderator
Jun 25, 2015
55
62
PCIe cards 'bottom out' in the slots, so if you raise the board the PCIe card will also rise.
Unless the PCB of the GPU hits the PCB of the motherboard, there's room. It would require a special slot component with an open bottom.
 

|||

King of Cable Management
Moderator
Sep 26, 2015
767
749
And I got the impression that they didn't foresee there being enough demand to justify the hefty R&D before something new and better came along.
Does AMD not have plans to support this socket as long as AM4 (which they claim will be the supported socket out through 2020)?
 

VegetableStu

Shrink Way Wielder
Aug 18, 2016
1,769
2,444
They could only fit the single-storey flat SODIMM slot behind, but I'd think it's enough to move two of them behind (and probably the 2 m.2 slots). It's going to be a super tight build-and-hopefully-forget system, because troubleshooting the underside of the mobo might need complete removal...

I just had the thought of a second PCB layer position at roughly the height of a PCIe slot actually. Not to jinx myself, but when I get time I'll do a quick 3D mockup on what I mean

hahahahha I'll stop. looks damned ridiculous LOL
 

|||

King of Cable Management
Moderator
Sep 26, 2015
767
749
Multiple boards stacked on one another have been done for a long time in servers. They usually go by the moniker of "mezzanine boards."



They do add a fair amount of complexity and signal integrity become an issue with very high frequency I/O signalling.