reading the conversation that blew up right above your posts might help. Might even be faster than waiting around for days for someone bored enough to come around and repeat it for you.
3 Likes
Need to be more specific.
What 8i stuff?
Will there be a 8i OcuLink module?
Can a 8i OcuLink module be used with an 8 to 4x cable?
There are multiple things you each could be asking about. There are anwsers about several things in the thread, but it is understandable not wanting to scan through 100 or more posts, though in that case you really want to be specific, help someone help you.
I mean the 8i module. I know very well it has been asked before, but it has been several months already, I assume the status is still “waiting for Framework review”? I’m asking again since I’ve ordered and paid it months ago. Josh did offer money back for those not willing to wait a while back, but I’d still like to know if there has been any developments.
2 Likes
I don’t think there has been any new updates.
For anyone who missed why such projects can take this amount of time: OcuLink Expansion Bay Module - #546 by nrp
2 Likes
Indeed, but its important to keep up the pressure. I assume someone at Framework reads these messages - go and bug the suppliers!
It doesn’t work like that. Framework I’m sure has a lot of other things there are working on. Which Nrp mentioned in his response. Only so much time and resources to go around. If direct Framework staff reading every random thing here, bad use of time imo. You’ll notice many of the Community Moderators are volunteers / not Framework staff. And also, good will with your contacts at suppliers is a also limited resource. Don’t want to annoy them too much about lower priority projects.
It’s understandable that people are impatient with the time it’s taking, but random posts just to attempt to cause pressure doesn’t help. For anyone truly willing to wait, as you mentioned, Josh did offer refunds.
For anyone who hasn’t seen it, do take a look at this potential option which community members are experimenting with: OcuLink eGPU works with the Dual M.2 expansion bay module - #32 by TheLPeink. Do consider though, due to the winding path and added connections the signals need to run through, cable length might be more limited, or reliability when paired with iffy eGPU enclosures might not be as good. So this would not truly replace a dedicated board, I imagine that would still be the best option in the end
3 Likes
That is called a “kludge”, routing cables like that. Its not a plug&play (yes I know it needs shutdown/reboot), what is my intent.
I do understand that there are priorities. I’m a 25 years veteran of tech industry. As an old project manager, I know you need to “kick” a bit sometimes to get your stuff’s priorities up in the ladder. There are always stuff deemed extremely urgent, so that is not an excuse.
If you do understand the issues involved with projects like this, yet are saying you want to continue giving random “kicks” in the thread just because you want your personal top priority to be higher on everyone else’s list, perhaps consider if that’s in the spirit of the community guidelines.
But regardless, I think I’ll disengage.
I would counter, in part jokingly, that with a right set of 3D-printed parts that solution could feel quite robust. And a solution in your hand is better than the one that’s not materialized yet 
1 Like
Also part jokingly, guys, don’t mind us in the other thread, it is the “whatever works” engine room ;). I am not proud of it, but I am super happy that it finally works. And it takes away the pressure from waiting here. PS: I do, however, understand if people want to ask for status updates once in a while.
2 Likes
I’ve got an external GPU running via Oculink off of the M.2 expansion board. Yeah, it’s not quite ideal, I’m sure with some JB Weld and a Dremel it could be made to feel a bit more idea. Mostly though I’m just using it to experiment, long-term I want a real solution that’s not some kludgy cable twisted in a knot off of the M.2 port.
3 Likes
Put of curiosity, what is your experience? Yesterted I tested mine (6900XT off a Minisforum DEG1) and am experiencing crashing when games enter fullscreen or when alt-tabbing…
I’ve got 2 external monitors through a Dell WD19S180W and I’m running with the cover closed. I’ve got a GTX 960 plugged in externally, and I’ve been mostly running Satisfactory full-screen now, with no problems at all. I didn’t even know I was running on the external GPU (I’d plugged in for some AI testing) until I noticed the fans were spinning. It’s not the fastest card by any means, it was just what I had laying around after the Tesla K80 I tried didn’t show up (Apparently they’re picky). I do have an M.2 SSD in the other socket of the expansion too, so both halves of the bus are being used.
With the various discussions about signal integrity and various bifurcation issues, I wonder if adding a pcie switch would make sense in a module like these? That should give you some local signal termination and regeneration on the module, plus you could add multiple 4-8x lanes off that back to the board.
This approach seems to work well enough for little flash nas systems cramming 12x SSD’s into a low power atom chip with minimal pcie bus, this could add much versatility in what and how many things can be connected to the framework, so long as they don’t mind some oversubcription.
Edit: Something like these: PLX Technology rolls out low-latency PCIe switches - Embedded
They’re used in Highpoint flash raid controllers and such too.
1 Like
Those chips cost a whole lot of money. The Dual M.2 board from Framework is 45€, and that one chip is 35-55€ before the profit margin by the manufacturer is added on top, so would make the board 2-3 times more expensive.
Understood, but I think most of us clamoring for this thread would probably find another double or more in change to pay for it regardless. Build it and the geeks will come!
I thought about use of those switches as something like the dual m.2 expansion, but imagine it had another 4/8x occulink or two on the outside as well, maybe another usb4 port or two at the rear.
Another concept I had was use the use module for a dock-sort of assembly that when docked, could extend the 8x externally with more a hard dock-sort of connector (think old docks with some interface pins), but after adding to the dock usb4/thunderbolt dock connectors, a couple m.2 ports, maybe still an external occulink or two, it’s far more than the 8x bus, in which case putting a pcie switch would make more sense.
Most USB4/TB3-4 docks are $200-400USD anyways, for a module chock full of goodness, it could open up some options for any/all of the above functionality into a module.
1 Like
The PCIe switch seems more like it would be ideal in the device the Oculink connects TO. Keeps the interface on the laptop cheap and efficient, and lets people buy what capacity they need. There are some PCIe expansion chassis out there, and some probably work with Oculink one way or another.
Agree, as a more premium option this sounds compelling.
I was mostly referring to my pre-order from last year. I know there were some delays. The last bit of news was about Frameworks not responding to his emails.
so it seems that framework and josh might have faced some difficulties. I will tried something a bit different. since I got a 5090, and I noticed that the pcie4x4 is bottlenecking the hell out of it. although
oculink 4i is very good, Oculink 8i itself is not really a reliable connector. My friend and I are trying to design a mcio or slimsas version of egpu module. However, after the evaluation, the size of the connector is too big to put inside of the bay itself. So we will make a really ugly adaptation to solve the issue. I will tell you guy’s when I made more progress.
5 Likes