Modified PCIe 8-pin connector

[Phuncz]

While it's not as innovative as the connector Regack showed below, I was still curious what one could do to reduce the connector size of the PCIe 8-pin (or 6+2-pin) cable.

It started in https://smallformfactor.net/forum/t...rdable-and-elegant-we-need-your-feedback.259/

@Aibohphobia, thanks for posting that [H] link, that's a great read. I can confirm that it is very important to make sure there is clearance for connectors. I learned the hard way...

Doesn't get much lower profile than that

I have a Lian-Li with about one millimeter of slack between the top of the PCB and the case, so I have a point of reference to work from

So while 1mm is a challenge for the future (when I've learned PCB design), I managed to reduce the part that extends past the socket by 50%, from 11mm to 5,5mm.


Don't mind the bad finish and badly sleeved cables, I blame the lens (partly). I also tried a sleeved wire where I put heatshrink just over the clamp that digs into the insulation.

I used a Dremel with a cutting blade and used two files and sandpaper to finish it. I needed to remove crud with a sharp knife and some real needle-nose pincers.

So basically if a company is prepared to get the mold costs paid, we'd have easy low profile GPU connectors.

This took me two hours, of which making pictures was almost half the time.

 

[PlayfulPhoenix]

I've said to @Aibohphobia on multiple occasions that, if the imminent crowdfunding campaign for Project Nova is successful, one of the next things I'd really like to jump on and produce is an ultra-low profile 6+2pin adapter cable. I've actually had one or two conversations with a friend/classmate/electro-mechanical engineer to understand how low-profile you could reasonably get (the answer to that being pretty darn low).

The cost to create a mold for the plug would be a bit steep upfront, but the manufacturing and materials cost would be low. I think that making and selling well-made adapters (sleeved, good materials) for ~$8-10 would be plenty viable.

 

[Phuncz]

I think it's viable too, if a reasonably large OEM signs on. I'm guessing the MOQ is over 10,000 for plastic molding ?
Wouldn't this also conflict with Molex's patents ?

My modified connector does feel sturdy enough, but I'm curious how lower profile versions would hold up.

A possible version that seems to hard to do at home, would be to have the pins be L-shaped, at two lengths, which would allow a sturdy construction but requiring a custom plastic casing. It would also ideally go from 2x4 connector to a 1x8 cable by using interleaving, so ribbon cables can be used, arguably more SFF and OEM-approved than sleeved cabling.

Or maybe just something like the SATA T-connectors.

 

[|||]

I've given this idea some thought as well...with proper modification to the internals in the Bolt II, I can potentially stuff an EVGA Classified card in it, but I'd have only 1.5mm clearance for the PCI-e power connectors.

A few pointers:

• You want the heat shrink to go down further on the pin as you will be exposing a substantial portion of it and you want to prevent direct contact with other things
• You can reduce the connector height even further around the retaining clip...this reduced height allow you to bend the individual wires down and be no higher than the top of the clip
• You definitely do not want sleeved cables...way too bulky; you want to have either bundled or ribbon cables depending on how you are managing them

Molex patents, if applicable, will probably be subject to FRAND restrictions.

 

[iFreilicht]

I agree with everything | | | said, except for sleeved cables, maybe. Sometimes things have to look good as well.

A commercially available cable from KI would probably sell very good, especially when you make it look nice. If you want a clean-as-possible case mod, having a ribbon cable come out a low-profile housing like that would even be interesting to people that don't need one due to space constraints. And it would of course allow some pretty neat case designs, too.

Your current efforts are very good already, but I think you could easily go slimmer.
For example, you could nearly cut the whole casing of the connector away on the bottom of the plug and bend the pins to leave the casing on a right angle. The top part of the casing can't be reduced as much, but maybe still enough to get their pins bent a little bit. You should hot-glue the bottom ones first, though, to avoid short circuits.
Once the top pins are bent, you shave off as much as possible from the rear of the casing and hot-glue the the top pins in place.

 

[jØrd]

pins sticking out a pcb, ribbon cable off the pcb to a full size connector, would be about as low profile as it gets, nfi if its a good idea in practice though.

 

[EdZ]

The lowest profile you could possibly get would be using two 'oversized' conductors (one for 12V, one for GND) connecting to all the pins in a row like a busbar, and exiting at right-angles to the connector. Those conductors would either extend all the way to the other end of the cable (thin, but inflexible to a pain to route) or be split into multiple conductors near the GPU end with the far end terminating as normal.

The downside is this is not only outside of the PCIe spec, it's outside the Mini-Fit Jr mechanical spec, as the pins would only be retained by the solder and whatever overmoulding were in place on the back side of the connector.

 

[iFreilicht]

pins sticking out a pcb, ribbon cable off the pcb to a full size connector, would be about as low profile as it gets, nfi if its a good idea in practice though.

This might be the easiest way in terms of manufacturing and it would make routing safe and repeatable. Maybe the PCB could be designed in such a way that it acts as the back of the connector without exposing any traces.

 

[PlayfulPhoenix]

pins sticking out a pcb, ribbon cable off the pcb to a full size connector, would be about as low profile as it gets, nfi if its a good idea in practice though.

This might be the easiest way in terms of manufacturing and it would make routing safe and repeatable. Maybe the PCB could be designed in such a way that it acts as the back of the connector without exposing any traces.

In my aforementioned conversation(s), this was essentially how we thought it could be done, but we assumed that the back would have to be covered.

Note that this would be much more expensive to make when compared to just doing it the "regular" way, and relying on sharp bends/clever wire arrangement to get a lower profile, however. And that makes it a bit hard to justify since you can already shrink the connector so much by simply making the plug a lot shallower.

 

[Phuncz]

The traces could be all internal when using a multi-layered PCB, but I'm not sure we'd be OK with pumping 150W of power over these.
I'm pretty sure QinX is abscent in the discussion because he's making these ideas in his DestructoLaser etchy thing of Doom and soon beat us all to the punch.

 

[EdZ]

You'd only get an isolated backside if the 'pins' were surface-mount rather than through-hole. SMT pins are notorious for being pulled off boards, as they are retained only by the adhesion of the copper trace to the PCB substrate itself and some friction from the pin sides against the plastic housing. Through-hole pins are more mechanically sturdy (pull-out retention from the backside solder blob) but obviously have live current exposed on the rear face.

 

[iFreilicht]

Couldn't you design the pins in such a way that they are secured between the housing and the PCB? Right now they have retention clips on the side that keep them from falling out the back, I guess you could do it the other way around, too.

 

[|||]

Yes, you can mount the pins blind or back-drill the pins and cover them. Costs a good deal, though (although Asus may have a fairly cheap way to do the blind mounting with the Auto-Extreme tech).

 

[GuilleAcoustic]

I wish we could see more cards like the Firepro V7800P, good way to get the 6 pins out off the way :

 

[PlayfulPhoenix]

I wish we could see more cards like the Firepro V7800P, good way to get the 6 pins out off the way.

Well... I mean, in lots of smaller cases, the plugs present a problem. But at least they are consistently located on that edge of the graphics card, towards the end.

The issue with having a card that places them on the end itself is that it isn't expected. Case designers take for granted where connectors are and account for it in their designs; putting them someplace else means you could have an unexpected length restriction, an issue with the length of your power cables, and so on. Further, if that placement were to catch on and become popular, we would suddenly have to compensate for both possible positions of the plugs, and that would invariably lead to more compromised designs.

Really, what we should wish for are ultra-high-powered PCI slots that negate the need for power ports at all - or even a motherboard that accepts 6 and 8-pin plugs, and just re-routes that through the PCI slots. That wouldn't introduce anything that would have to be accounted for in a case design.

 

[GuilleAcoustic]

Plug located on the end is not uncommon and must be a possible option. Here are examples, to name a few :

Radeon HD5870 :

Quadro K5000 and Firepro W8100 :

A tesla K20 :

We need at least a standard size for GPU with standard power connector placement, just like MXM where everything is standard up to the chip placement and height

Really, what we should wish for are ultra-high-powered PCI slots that negate the need for power ports at all - or even a motherboard that accepts 6 and 8-pin plugs, and just re-routes that through the PCI slots. That wouldn't introduce anything that would have to be accounted for in a case design.

That would be nice, but for multi GPU that would need an impressive power passed to the motherboard.

 

[iFreilicht]

Well... I mean, in lots of smaller cases, the plugs present a problem. But at least they are consistently located on that edge of the graphics card, towards the end.

The issue with having a card that places them on the end itself is that it isn't expected. Case designers take for granted where connectors are and account for it in their designs; putting them someplace else means you could have an unexpected length restriction, an issue with the length of your power cables, and so on. Further, if that placement were to catch on and become popular, we would suddenly have to compensate for both possible positions of the plugs, and that would invariably lead to more compromised designs.

Really, what we should wish for are ultra-high-powered PCI slots that negate the need for power ports at all - or even a motherboard that accepts 6 and 8-pin plugs, and just re-routes that through the PCI slots. That wouldn't introduce anything that would have to be accounted for in a case design.

I feel the problem only arises if the cards are very long AND have the plug on the front. With something like the R9 nano, the whole card is shorter than 170mm by about the length of that connector, so no harm done. I do see the point that it can impact the installation procedure quite a bit, though.

In the end, having powered PCIe interconnects on the board highly impact modularity and compatibility that we currently get from self-built PCs and would limit board space on ITX boards even more, especially as these connectors are commonly through-hole parts, so even the backside of the board would be lost.
The maximum TDP of a GPU your PC can support would not only be relying on the PSU anymore, but also the Mainboard, which effectively just passes the power input through to the GPU anyway. Not a good design decision in my eyes.

The best way to solve this problem is to add a specification for the positioning of PEG connectors to the PCIe standard. This spec would require that the power connectors are recessed into the PCB by ~2-3cm and oriented towards the side of the card, so the power cables can bend inside that space, making the height of the GPU the absolute maximum height you have to account for.

We need at least a standard size for GPU with standard power connector placement, just like MXM where everything is standard up to the chip placement and height

That would be nice, but for multi GPU that would need an impressive power passed to the motherboard.

The problem with that is that it would actively prohibit changes like HBM to be backwards compatible as the cooling solution drastically changes shape. I don't think it's a good idea to connect the graphics card from its cooling, that should go hand in hand like it does now, for the sake of the consumer. If you buy a GPU right now, it will work for sure, even if it's not the most silent or small it could be. But if you required everyone to also buy an additional cooler or case designers had to incorporate the cooling solution into the case, you'd have to set a maximum limit on TDP or rely on the user to be informed about which heatsinks can cool which GPUs, complicating the process of selecting components even further and segregating the market.

Yes, the implications for Quad-SLI or Quad-CFX are a clear killer of this idea.

 

[confusis]

I'd pay good money for low profile 90 degree PCIe 6 pin adapters

 

[GuilleAcoustic]

Following the same idea, we need a Sata data + power combo connector on the motherboard side. That shouldn't be too hard to do and a single cable for both would really reduce the number of cables and help with their management while improving hairflow.

 

[iFreilicht]

Combining SATA data and power seems much more sensible to me, as the number of drives is already limited by the motherboard and they don't need a lot of power anyway. The question is, would such an interface support 5.25" drives as well? Those need more power which would potentially increase the size of the connector, so it might be worth to think about it as an interface only for 2.5" drives and Slimline drives.
That way, you could potentially use something like the eSATAp connector which is quite small and supplies 5V and SATA data, which is already sufficient for slimline drives and should be enough for most SSDs. Maybe it could also be designed in a way that makes 5mm thick 2.5" drives viable again.