sorry for the spam, is the 8pcie version just a preorder?
Is there a difference if you use the 8 pcie slot with an 4pcie dock?
PCIe does transfer 75W, no matter how many lanes - not sure if OCULink transfers power at all though.
the x8 version is on preorder.
Using the x8 module with a x4 dock is possible. It does not provide any more bandwidth than the x4 module.
2 Likes
Oculink has no power lines (there might be some, but they are most definitely hard limited and wouldn’t be able to power anything remotely like a GPU), its all just data and some stuff like cable presence. The device on the other end needs its own power supply.
3 Likes
@Josh_Cook Great job making this.
I don’t have a framework yet, but have been researching/deciding on the 13 or 16.
Your project will probably push me into a 16. I will likely buy everything all together once pictures/ect or a guide comes out in how to use properly lol. And if someone finds out exactly what cables, adaoter boards to plug the GPU into.
I still need to figure out how I am going to repurpose the 16 mainboards, these things have wonky-shaped heatsinks.
2 Likes
Just another small update, after getting my 8x design reviewed by Framework, I found out that enabling bifurcation will mean that the DP input will not work since the DP and PCIe clock lanes are shared, so you will not be able to split off my 8x connection to 2 4x links. I’m still working out how this will work for the 4x + NVMe module, but there is a chance that the M.2 socket will have to be removed since removing the DP input port will make this board useless for eGPUs
7 Likes
will this means that 8x is ready for production?
Any possibility of leaving the M.2 for people who don’t need the DP input port, and are just going to run directly off of the GPU? So those who want it can use the DP input, and those who don’t can get access to another M.2?
I was just about to buy a GPD laptop for oculink support, I guess I should check this out first.
What’s the status of the 8x oculink module?
Can someone explain the graphics modulle Interposer? I bought the graphics interposer from framework but it doesnt even fit because I didn’t buy the external gpu that framework also has. So do I need the interposer for the Oculink Module to work? Again I have the 16 ’ frameowork and i bought the grapgics module interposer, unless there is another one I’m a little lost.
Yes, you need it. The Graphics Interposer has more connector which exposes PCIe lanes to the expansion module.
But I don’t need the external gpu right? I’m unable to have the interposer fit properly
You need the interposer, the Oculink board and the expansion bay shell.
1 Like
Getting very close to prototype ready
6 Likes
I’ll have to see if there is any interference, and also from a signal integrity standpoint, but this may be possible
1 Like
The interposer should fit on the oculink module. Since you don’t have it, it does not fit now.
So it’s an oculink piece that goes on the back of the framework?
Hi Josh, I’ve been lurking here for several months now waiting patiently for the oculink pcie x4 expansion to be ready for prime-time (actually I’ve been delaying upgrading from my older framework 12th gen intel entirely because of the lack of this product). It wouldn’t be a dealbreaker to me to lose the extra m.2 slot, since we still have 40gbit usb4 connections on the laptop itself, and those fast m.2 enclosures have become more and more affordable recently (saw one this morning on amazon US for $75). Right now, the only laptops I know of that support oculink built-in are GPD Win and some China-only lenovo models. I’m extremely excited about this product.
It’s a PCB that goes inside the Expansion Shell
My question is other than the preorder for the expansion oculink that Josh cook has, and the graphic interposer I got from framework, do I need anything else to be able to have the oculink work?